Law in Context, Vol 37, Issue 1; 2020 LA TROBE EDITORIAL BOARD

Pompeu Casanovas Jianfu Chen Emma Henderson General Editor Chair of the Editorial Board Book Review Editor

Kerstin Steiner Savitri Taylor David Wishart Deputy General Editor Deputy Executive Editor Executive Editor

Dina Afrianty Fiona Salisbury Nicholas Morris Jeff Barnes Magda Karagiannakis Andre Oboler Susanne Davies Fiona Kelly Lola Akin Ojelabi Louis de Koker Patrick Keyzer Tarryn Phillips Anne Wallace Commissioning Editor: Darren O’Donovan Treasurer: Lisa Egan (La Trobe Law School Manager) Editorial Assistants: Meredith Jones, Daniel McIntosh, David Venema Journal Administrators: Kallirroi Stavrianou, Mustafa Hashmi

INTERNATIONAL EDITORIAL BOARD

1. Andrews, Penelope (La Trobe Univer- 10. Francesconi, Enrico (IGSG-CNR, Italy, 20. Tranter, Kieran (Queensland University sity, Australia) and EUR-Lex, Luxembourg) of Technology, Australia)

2. Araszkiewicz, Michal (Jagiellonian 11. Frug, Sara (Cornell University, USA) 21. Vallbé, Joan-Josep (University of Bar- University, Poland) celona, Spain) 12. Governatori, Guido (Data61, CSIRO 3. Baron, Paula (La Trobe University, and La Trobe University, Australia) 22. Valverde, Mariana (University of To- Australia) ronto, Canada) 13. Liu, Sida (University of Toronto, Canada)

4. Bennet Moses, Lyria (UNSW, Australia) 14. Montiel, Elena (Polytechnic University 23. Van Engers, Tom (University of Am- sterdam, The Netheralands) 5. Chen, Albert Hung-yee (University of of Madrid, Spain) Hong Kong, Hong Kong) 15. Noriega, Pablo (IIIA-CSIC, Spain) 24. Wang, Zhiqiong June (Western Sydney University, Australia) 6. Contini, Francesco (IGSG-CNR, Italy) 16. Pagallo, Ugo (University of Torino, Italy) 25. Wang, Jiangyu (National University of 7. Czarnota, Adam (UNSW, Australia) 17. Poblet, Marta (RMIT University, Australia) Singapore, Singapore)

8. Dignum, Virginia (Umeå University, 18. Rodríguez-Doncel, Víctor (Polytechnic 26. Wheeler, Sally (Australian National Sweeden) University of Madrid, Spain) University, Australia)

9. Fabri, Marco (IGSG-CNR, Italy) 19. Schultz, Ulrike (German National Dis- 27. Zeleznikow, John (La Trobe University, tance Teaching University, Germany) Australia)

Law in Context, Vol 37, Issue 1, 2020 INTERNATIONAL ADVISORY BOARD

1. Abel, Richard (University of California Los Angeles, USA) 16. Greenleaf, Graham (UNSW, Australia)

2. Antons, Christoph (University of Newcastle, Australia) 17. Kirby, Michael (The Hon., AC CMG, Australia)

3. Arup, Christopher (Monash University, Australia) 18. Mathews, Robert (University of Hawaii, USA)

4. Ashley Kevin (University of Pittsburgh, USA) 19. Mendelson, Danuta (Deakin University, Australia)

5. Bourcier, Danièle (CNRS, France) 20. Mendelsohn, Oliver (La Trobe University, Australia)

6. Brigham, John (University of Massachussets Amherst, 21. O’Malley, Pat (Australian National University, Australia) USA) 22. Pérez-Perdomo, Rogelio ( Metropolitan University of 7. Bruce, Tom (Cornell University, USA) Caracas,Venezuela)

8. Chanock, Martin (La Trobe University, Australia) 23. Pitt, Jeremy (Imperial College, UK)

9. Di Federico, Giuseppe (IGSG -CNR, Italy) 24. Rotolo, Antonino (University of Bologna, Italy)

10. Ferrari, Vincenzo (University of Milan, Italy) 25. Sartor, Giovanni (University of Bologna, Italy)

11. Fitzpatrick, Peter † (Birkbeck, University of London, UK) 26. Scheiber, Harry N. (University of California Berkeley, USA) 12. Friedman, Lawrence (Stanford University, USA) 27. Schweighofer, Erich (University of Vienna, Austria) 13. Galanter, Marc (University of Wisconsin-Madison, USA) 28. Toharia, José Juan (Autonomous University of Madrid, 14. Garth, Bryant (University of California, Irvine, USA) Spain)

15. Goodrich, Peter (Cardozo School of Law, USA) 29. Tomasic, Roman (University of South Australia, Australia)

Law in Context, Vol 37, Issue 1; 2020 ISSN: 1839-4183

TABLE OF CONTENTS

Editorial: The Digital World We Will Live By 6 By Pompeu Casanovas, David Wishart and Jianfu Chen

Sovereigns, Viruses, and the Law: The Normative Challenges of Pandemics in Today’s Information Societies 11 By Ugo Pagallo

Knowledge Graphs: Trust, Privacy, and Transparency from a Legal Governance Approach 24 By Daniel Schwabe, Carlos Laufer, and Pompeu Casanovas

Electronic Australian Elections: Verifiability of Accuracy is a Design Goal, which Must be Mandated by Law and Deliberately Designed into Electronic Electoral Processes 42 By Vanessa Teague and Patrick Keyzer

Human Rights Issues in Constitutional Courts: Why Amici Curiae are Important in the U.S., and What Australia Can Learn from the U.S. Experience 66 By H. W. Perry Jr and Patrick Keyzer

The Adultification of the Youth Justice System: The Victorian Experience 99 By Natalia Antolak-Saper

At Play in the Field of Dreams: Theorising Attitudes, Perceptions and Practices of Law Students in conjunction with the Reflections of Early Career Commercial Lawyer 114 By Barry Yau and David Catanzariti

A Law School Course in Applied Legal Analytics and AI 134 By Jaromir Savelka, Matthias Grabmair and Kevin D. Ashley

Law in Context, Vol 37, Issue 1, 2020 4 ISSN: 1839-4183

Research Note: LYNX: Towards a Legal Knowledge Graph for Multilingual Europe 175 By Víctor Rodríguez-Doncel and Elena Montiel-Ponsoda

Research Note: Rules as Code 179 By Matthew Waddington

Book Review: Advanced Introduction to Law and Artificial Intelligence 187 By Stephanie Falconer

Book Review: Anne Wesemann: Citizenship in the European Union Constitutionalism, Rights and Norms 190 By David Wishart

Book Review: Can We Still Afford Human Rights? Critical Reflections on Universality, Costs and Proliferation 194 By Emma Henderson

Book Review: A Lesser Species of Homicide. Death, Drivers and the Law 197 By Robert P. Brown

5 Law in Context, Vol 37, Issue 1, 2020 Received: December 26, 2020, Date of publication: January 5, 2021, DOI: http://doi.org/10.26826/law-in-context.v37i1.136

Editorial The Digital World We Will Live By

By Pompeu Casanovas, La Trobe University, Melbourne, Australia, Orcid: http://orcid.org/0000-0002-0980-2371 David Wishart, La Trobe University, Melbourne, Australia, Orcid: http://orcid.org/0000-0003-2281-4745 Jianfu Chen, La Trobe University, Melbourne, Australia, Orcid: http://orcid.org/0000-0002-4711-6524

ABSTRACT This editorial provides a brief description of the digital turn experienced at the beginning of this century, the side effects of Covid-19, and the twelve commitments recently laid down by the United Nations. It introduces the concepts of ‘Global Syndemic’ (or synergy of epidemics), the ‘Internet of Things’, ‘RegTech’, and ‘Knowledge Graphs’. Finally, it

describesKeywords briefly – Covid-19, the articles Global includedSyndemic, in Digital this issue. turn, Internet of Things, RegTech, Knowledge Graphs

Disclosure statement – No potential conflict of interest was reported by the author.

License – This work is under Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) https://creativecommons.org/licenses/ by-nc-sa/4.0/

Suggested citation: P. Casanovas, D. Wishart, and J. Chen (2020). “Editorial: The Digital World We Will Live By”, Law in Context, 37 (1): 6-10, DOI: http://doi.org/10.26826/law-in-context.v37i1.136

Summary 1. The Digital World We Will Live By 2. On the Contents of this Issue: LiC 37 (1) 3. References

Law in Context, Vol 37, Issue 1, 2020 6 ISSN: 1839-4183

4. THE DIGITAL WORLD WE WILL LIVE BY read them in the light of Covid-19 we will soon realise that LiC 37 (1) is the third Issue of the new Online Law in what is at stake is not just a pandemic but what health Context. It comes at the end of a year subject to all possible experts have called a global syndemic, i.e. a synergy of evaluation criteria. The year leaves a host of questions epidemics. This was already advanced by some analyses awaiting answers. In March 2020, when Covid-19 was published at The Lancet in February 2019 before Covid-19 already in evidence all over the world, Yuval N. Harari appeared in October/November of the same year, focus- (2020) wrote in the Financial Times that humanity was ing on obesity, undernutrition, and climate change. This - global syndemic “affects most people in every country and ian surveillance and citizen empowerment; the second region worldwide” and their elements “co-occur in time betweenfacing two nationalist important isolation choices, theand first global between solidarity. totalitar The and place, interact with each other to produce complex US elections in November and their aftermath seem to sequalae, and share common societal drivers” (Swinburn - et al. 2019, p. 791). From this perspective, Covid-19 and tions than explanations or answers. the global response to it has accelerated the effects of confirm this judgement but have thrown up more ques Many declarations, manifestos, predictions and com- a global change that was already happening, focusing mentaries have followed one another over the course of especially on the Internet of Things. Both threats and challenges are taking place in a dig- present crisis has been felt as an urgent need. The United itised world that is not what we expected it to be only Nationsthe past celebratedtwelve months. its 75th Reflection Anniversary on the in effects 2020. ofHow the- ever, as the UN meeting coverage underlined, what world democracy, and John Perry Barlow’s dreams in his widely twenty-five years ago. The Internet alone cannot foster leaders actually said was “we are not here to celebrate” spread Declaration of the Independence of Cyberspace (1994) (UNGAa). Indeed, the Declaration on the Commemoration did not come true. Web governance will not emerge from of the Seventy-Fifth Anniversary of the United Nations, “ethics, enlightened self-interest, and the commonweal” submitted by the President to the General Assembly on (Barlow 1994, ibid.). He was visionary about the changes September 16th 2020 asserts that that cyberspace would bring. But he fell short about the “Our world is not yet the world that our founders envisaged 75 years ago. It is plagued by growing inequality, with a radically different environment. The IoT embraces manyevolution cyber-physical of technology systems and the difficultyautomatically to adapt connected, and cope climate change and pandemics. People in different corners using sensors and other technologies, enabling people poverty, hunger, armed conflicts, terrorism, insecurity, of the world are forced to make dangerous journeys in and all kind of devices to interact with each other and search of refuge and safety. The least developed coun- the environment on real time, and to keep track of it, i.e. tries are falling behind, and we still have not achieved collecting, storing and sharing data on an unprecedented complete decolonization.” (UNGA 2020b) scale. How are we going to manage, monitor and control The text lays out twelve commitments to reanimate global resolve: leave no one behind, protect the planet, the information processing flows that occur in the IoT? promote peace, abide by international law, place women in 2021. The Internet of Things blends the physical and Gartner has identified IoT applications as trending and girls at the centre, build trust, improve digital coop- digital worlds and is transforming the way we live and eration, upgrade the United Nations, ensure sustainable work. According to it, workplaces of the future will evolve in response to the pandemic “regardless of whether they - financing, boost partnerships, work with youth, and, facturing plants” (Gartner 2020). This holds for educa- These ideals matter and should not be ignored. We finally, be prepared. (UNGA 2020a, b) tion,are offices universities, for knowledge research workers centres, or and automobile what has manu been would like to stress from our side that they cannot be termed “digital twins”. Digital twins—digital replicas of treated separately, for they are as connected to each other a living or non-living entity in the IoT—are a potential as the threats they face. They are closely tied, and if we transformative technology for everything from devices to

7 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 cities, but standards or common integration frameworks and legal education should acknowledge what is happen- do not yet exist for digital twin data, models, analytics or ing and react accordingly. security (Gartner 2020). We are only now coming to understand that something 5. ON THE CONTENTS OF THIS ISSUE: LIC 37 (1) else is required to produce a fair society, the effective Law in Context online, LiC 36 enactment of civil rights, and a political acceptable world (1), we set a common ground for Law & Society and AI in digital environments. Embedding the rule of law into & Law.In the We first pointed volume at the of concept of legal knowledge as the structures of the Internet of Things is not enough a research subject (Casanovas, Chen and Wishart 2019), either, as the IoT is able to change the market, the way and we contended the urgent necessity of a joint tight co- how contracts are drafted and executed, the way how laws can be drafted and enacted, and our own understanding represented another step in this direction, strengthening of what comprises law and regulations. someoperation LiC valuesbetween and the beliefs two fields. on the The relevance second Issue of gender, 36 (2) According to a recent report of the Cambridge Centre history, and human rights (Savitri 2020). for Alternative Finance, industry forecasts for RegTech This third volume is a new step. It is not a special is- companies point to year-on-year growth of between sue on technology, rather it gathers some relevant works 23% and 25% between 2018 and 2023 (Swinburn et which consistently try to shed light on some contemporary al. 2019, 11). RegTech —LawTech, FinTech, SupTech, In- problems of this hybrid world arising from human, insti- suTech tutional, and technological changes. LiC 37 (1) contains and due diligence, dynamic compliance by design, risk seven original Research Articles, two relevant Research —companies and web services are offering profiling analysis and data analytics to corporations, companies, Notes, and four Book Reviews. Ugo Pagallo opens the Issue with an article on the the emergence of these new specialised legal markets legal and political impact of the Covid-19 crisis, draw- arepublic impacting administrations, on the structure, and law form firms. and Technology content of legal and ing the attention to fundamental questions on authority practice to an extent that has yet to be determined. For and political legitimacy, coercion and obligation, power instance, GPT-3 is trained on 410 billion tokens and has and cooperation. He depicts the regulatory environment 175 billion parameters. Would it replace lawyers in a of the Web of Data and the Internet of Things, which is non-distant future? Daniel Schwabe, Carlos Laufer and Pompeu Casano- to this kind of suggestions. But there are some authorised significantly pertinent to present the volume. vas introduce the notion of Knowledge Graph as a legal voicesThe that legal are field betting usually on reactsit. GPT-3 with learns extreme how toprudence predict governance framework to support trust, privacy, trans- the next word from reading literally billions of texts. If it parency and accountability. Stemming from the Web of can be fed and trained with cases and legal arguments Data, Knowledge Graphs (KG) represent a graph-theoretic (and it can), and it can be easily enriched with technical representation of human knowledge such that it can be annotations (and it can), then it will attract investors and ingested with semantics by a machine (Fensel et al. 2020). The authors show that it is an especially suitable tool for purposes, and subsequently other computer programs will it will be used in the legal field for professional and legal regulatory purposes on the Internet of Things. Vanessa Teague’s and Patrick Keyzer’s work comes as follow. Legal professionals are “like traffic cops just before a hot topic, after the American presidential elections and the roll-out of traffic lights”, just before being replaced by its aftermath in November and December 2020. Their “AutoLaw, where machines guide us through conflicts the work is an example of felicitous cooperation between a way traffic lights help us avoid collisions” (Blijd 2020). computer scientist and a constitutional lawyer. The au- certain is that legal doctrine (either in private, public, thors delve on the conditions that would make electronic internationalA fictional and dystopic transnational future perhaps? law), socio-legal Maybe. research, What is Australian elections safe and acceptable. Electronic voting

Law in Context, Vol 37, Issue 1, 2020 8 ISSN: 1839-4183 and counting have been adopted in a number of Australian attitudes and perceptions of law students, contrasted jurisdictions, but there is evidence that e-voting systems lack transparency. The authors argue that source code about their school experiences. Pierre Bourdieu’s (1984) and the voting data should be made available for testing notionsagainst the of powerreflections and ofdistinction early career are commercial used as focal lawyers lens and recommend several appropriate legislative reforms. to understand students’ perceptions and to contend that The next article also relates to Australia, focusing on their assorted ambitions would require a Law School human rights and identity groups. In its nature, it is a piece of comparative law. H. W. Perry Jr. and Patrick Keyzer ex- perspective of what ‘real practice’ consists of. environment promoting a more diversified and accurate amine the role of amicus curiae procedures in the United Jaromir Savelka, Matthias Grabmair and Kevin D. Ashley States Supreme Court and consider whether they can be describe the structure and contents of a new Law School adopted by the High Court of Australia. They contend that human rights cases raise questions of minority rights The course took place in the University of Pittsburgh, in that should not be adjudicated without input from those Pennsylvania,course in applied USA, legal and analyticsthe authors and thoroughly Artificial Intelligence. explain how minorities. It is then aligned with the Constitutional Open Machine Learning (ML) and Natural Language Processing Rights interpretation pointed out by Keyzer (2010). The (NLP) techniques can be taught to combined groups of authors recommend admitting written submissions and law students and students from technical departments. allowing applicants to make oral submissions. In the same practical way used in Ashley (2017), this Likewise, Natalia Antolak-Saper’s article focuses on course explores some effective strategies and exercises some social practices and institutional actions that affects to furnish to the students the required skills and knowl- minorities. This time, it is about youth offenders in the edge to deal with the new RegTech toolkit. As the authors Victorian Juvenile Justice system. The author examines explain, the course culminated in joint projects engaging - joint small teams in applying machine learning and data torical examination of the Victorian model. The article analytics to legal problems. The results of two of them showsthe process the role of of‘adultification’ the media in advocatingof the system, the afterimposition a his have been accepted for presentation at JURIX 2020 and of a ‘tough on crime’ approach that have been assumed already published (Gretok, Langerman and Oliver 2020; and implemented by the Victorian State Government Xu, Šavelka and Ashley 2020). without a sound evaluation, and contends that this kind This Issue also apprises the readers about two inter- of policy is not only ineffective but inconsistent with the esting research initiatives that are just emerging in the best interests of the child. The two last Research articles can be presented alike, legal field. They are published as Research Notes, i.e. although they have been submitted separately, and they short papers explaining processes, results and findings apparently deal with two very different subjects in the of specific research trends. Montiel-Ponsoda, deploys the results of the ongoing H2020 Project,The first LYNX, one, Legal by Víctor Knowledge Rodríguez-Doncel Graph for Multilingual and Elena waysfield ofof legalteaching education law and in practicesthe contemporary stemming world. from both It is Compliance Services. This is an example of what can be aan critical example reading of how from beneficial a socio-legal it could approach, be to explore and new the done to facilitate legal compliance through a collection of new technological toolkit that it is required to understand services, implementing natural language techniques, better and operate into the Internet of Things. Both articles deal legal information retrieval and cross-lingual answering with present true teaching experiences, drawing lessons of questions and information discovery. learned from them. The second one, by Mathew Waddington, is an elabo- The experience of Barry Young and David Catanzariti - is located at the Australian National University, in Can- lished at The Conversation (Governatori et al. 2020) on berra. The authors carry out a qualitative analysis on the therated recent and reflectivemovement response called Rules to the as brief Code recently. It is also pub an

9 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 invitation to discussion and dialogue with legal scholars 7. Governatori, G., Barneys, J., de Koker, L., Poblet, M. Hashmi, and legal professionals.To our opinion, this is one of the M., Zeleznikow, J. Casanovas, P. 2020. “‘Rules as Code’ will let computers apply laws and regulations. But over-rigid and development of this recent government trend that interpretations would undermine our freedoms.” The Conversation, 26 November, https://theconversation. attemptsbest descriptions to draft and the serve reader laws can using find bothabout coding the aims and com/rules-as-code-will-let-computers-apply-laws- natural language. and-regulations-but-over-rigid-interpretations-would- Finally, LiC 37 (1) closes with four reviews of books undermine-our-freedoms-149992 8. Gretok, E., Langerman, D., and Oliver, W.M. 2020. “Trans- (Stephanie Falconer), by Anne Wesemann (David Wishart), formers for Classifying Fourth Amendment Elements bypublished Jan Wouters in 2020: and by members Woodrow of Barfield the Leuven and Ugo Centre Pagallo for Global Governance Studies (Emma Henderson), and by P. (eds.) Legal Knowledge and Information Systems JURIX and Factors Tests”, In Villata, S., Harašta, J., and Křemen, Kerry King (Robert P. Brown). Again, these volumes are a 2020: The Thirty-third Annual Conference, IOS Press: Amsterdam, pp. 63-72. good sample of the researches covered by our Journal on Human Rights, International Law, Law and Technology, 9. Harari, Y.N. 2020. “The world after coronavirus”, Financial Times, March 20. https://www.ft.com/ and Criminology. content/19d90308-6858-11ea-a3c9-1fe6fedcca75 Finally, we would like to thank, in alphabetical order, 10. Keyzer, P. 2010. Open Constitutional Courts. Melbourne: Nishali Paththamperuma Arachchige, Alessandra Burato, Federation Press. Samuel Fenton, Madeleine E. Gome, Kelsey Kropf, and 11. Schizas, E., McKain, G., Zhang, B.Z., Garvey, K., Ganbold, A., Christina Michael for their valuable work with this edi- Hussain, H., Kumar, P., Huang, E., Wang, S. and Yerolemou, tion. We deem the support of our recently created LiC N. 2019. “The Global RegTech Industry Benchmark Report”. Student Editorial Board essential to help both authors Available at SSRN 3560811. and readers with Issues to come. 12. Swinburn, B.A., Kraak, V.I., Allender, S., Atkins, V.J., Baker, P.I., Bogard, J.R., Brinsden H., Calvillo, A., de Schutter, O., Devarajan, R., Ezzati, M. 2019. “The Global Syndemic 6. REFERENCES of Obesity, Undernutrition, and Climate Change: The 1. Ashley, K.D. 2017. Artificial intelligence and legal analytics: Lancet Commission report”, The Lancet. Feb 23; 393 new tools for law practice in the digital age. Cambridge: (10173):791-846. Cambridge University Press. 13. Taylor, S. 2019. “Editorial”, Law in Context, 36 (2): 3-4. 2. 14. United Nations General Assembly (UNGA). 2020a. Declara- when.” September 7. https://www.legalcomplex.com/2020/09/07/ tion on the Commemoration of the Seventy-Fifth Anniversary Blijd, R. 2020. “Will lawyers be replaced by GPT-3? Yes, and here’s will-lawyers-be-replaced-by-gpt-3-yes-and-heres-when/ of the United Nations. https://undocs.org/A/75/L.1 3. Bourdieu, P. 1984. Distinction: A social critique of the judge- 15. United Nations General Assembly (UNGA). 2020b. Meeting ment of taste. Cambridge, Mass.: Harvard University Press. Coverage. Seventy-Fifth Session, 3rd Meeting. GA/12267, 4. Casanovas, P., Chen, J. and Wishart, D., 2019. “Law in 21 September, https://www.un.org/press/en/2020/ Context for the Digital Age”, Law in Context. A Socio-legal ga12267.doc.htm Journal, 36 (1): 3-11. 16. Xu, H., Šavelka, J., and Ashley, K.D. 2020. “Using Argument 5. Mining for Legal Text Summarization”. In Villata, S., Harašta, Panasiuk, O., Toma, I., Umbrich, J. and Wahler, A., 2020. Legal Knowledge and Information KnowledgeFensel, D. Şimşek, Graphs. U.,Methodology, Angele, K., ToolsHuaman, and SelectedE., Kärle, Use E., Systems JURIX 2020: The Thirty-third Annual Conference, Cases. Cham: Springer International Publishing. J.,IOS and Press: Křemen, Amsterdam, P. (eds.) pp. 184-193. 6. Gartner. 2020. “Interactive hypecycle for the Internet of Things”. https://www.gartner.com/document/3987602

?ref=solrAll&refval=272631259

Law in Context, Vol 37, Issue 1, 2020 10 Received: December 11 May, Date of acceptance: August 22, 2020, Date of publication: September 30th 2020, DOI: https://doi.org/ 10.26826/law-in-context. v37i1.117 Sovereigns, Viruses, and the Law: The Normative Challenges of Pandemics in Today’s Information Societies

By Ugo Pagallo, Professor of Jurisprudence at the Department of Law, Orcid: https://orcid.org/0000-0001-7981-8849 University of Turin, Italy

ABSTRACT The paper examines the legal and political impact of the Covid-19 crisis, drawing the attention to fundamental questions on authority and political legitimacy, coercion and obligation, power and cooperation. National states and sovereign governments have had and still will have a crucial role in re-establishing the public health sector and addressing the colossal challenges of economic reconstruction. Scholars have accordingly discussed the set of legal means displayed during this crisis, such as emergency decrees, lockdowns, quarantines, travel bans, and more. The aim of this paper is to stress the limits of such discussions on powers of national governments and sovereigns, in order to illustrate what lies beyond such powers. Focus should be on the ontological, epistemic and normative constraints that affect today’s rights and duties of national states. These constraints correspond to a class of problems that is complex, often transnational, and increasingly data-driven. In addition, we should not overlook the lessons

coordinationlearnt from such and fields,cooperation as environmental complement law the and traditional internet powers governance, of national anti-terrorism governments, and eventransnational in the times business of the motherlaw, up toof theall pandemics. regulation Theof Artificial Covid-19 Intelligence crisis has been (AI). often Such interpretedfields show asthat if this legal were co-regulation the last chapter and mechanisms of an on-going of

of today’s information societies. history about the Leviathan and its bio-powers. It is not. The crisis regards the end of the first chapter on the history Keywords – Authority; Biopolitics; Coercion; Complexity; Coordination; Emergency; Governance; Information Societies; Legitimacy; Power; Sovereignty; Transnational; Virus Disclosure statement – No potential conflict of interest was reported by the author. License – This work is under Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) https://creativecommons.org/licenses/by-nc-sa/4.0/ Suggested citation: U. Pagallo. “Sovereigns, Viruses, and the Law: The Normative Challenges of Pandemics in Today’s Information Societies”, Law in Context, 37 (1): 11-23, DOI: https://doi.org/ 10.26826/law-in-context.v37i1.117 Summary 1. Introduction 2. On Life and Death 3. Back to the Leviathan? 4. The Governance of Pandemics 4.1 Requirements 4.2 Functions 4.3 Solutions and Open Problems 5. Conclusions 6. References

11 Law in Context, Vol 37, Issue 1; 2020 ISSN: 1839-4183

7. INTRODUCTION end of the cold war (1989), September Eleven (2001), and From time to time, there are historical events that now the Covid-19 pandemic (2020 onwards). deeply affect our understanding about the world and about This paper deals with this latter pandemic. It invites ourselves. Such events may concern political upheavals, once again to think big and out of the box. It is worth mentioning, in particular, today’s contrast between sound revolutions. One of these events marked the life and work ofnatural the English catastrophes, philosopher scientific Thomas discoveries, Hobbes, or technological the father of systems, and the confusion of most current political and modern political and legal thought. In his autobiography legalscientific debates research on the in Covid-19 e.g. epidemiology crisis. Confusion and complex entails from 1679, Hobbes refers to the Spanish attempt to invade chaos and bewilderment, turmoil and disorientation, England in July 1588, few months after his own birth in much as occurs when something new and big hits for the April of that year: “And hereupon it was my Mother Dear / Did bring forth Twins at once, both me and Fear.” Fear of today’s legal and political crisis, such as the exercise plays indeed a crucial role in Hobbes’s description of hu- offirst “emergency time. Some powers” have focused (Gostin on andthe short-termHodge 2020); features lock- man nature and the interaction in a state in which man to downs (Stock 2020); and their impact on international man is an “arrant Wolfe.” As a political and legal thinker, law (Bogdandy and Villareal 2020). Others have intended Hobbes’s aim was to sort out a rational solution for this to stress the long-term effects of the pandemic and the state of nature, much as occurred during the civil war political essence of this crisis. According to David Runci- man, for example, “at the heart of all modern politics is a powers of a Leviathan was Hobbes’s grand response. trade-off between personal liberty and collective choice... thatNatural afflicted catastrophes England from and 1642 technological to 1651. The failures, sovereign in Though the pandemic is a global phenomenon, its impact addition to political crises and revolutions, have inspired is greatly shaped by decisions taken by individual govern- Philosophische ments” (Runciman 2020). An alternative view proposes Lehrjahre from 1977, Hans-Georg Gadamer recalls his mechanisms of coordination, in order to tackle the “two reactionin-depth asreflections. a 12 years In oldhis autobiographykid to the sinking of the RMS particularly important choices” we face in this time of Titanic on 15 April 1912. The belief that science and crisis, that is, the choice between totalitarian surveillance technological innovation would have brought only roses and citizen empowerment on the one hand, and between but no thorns, faded away. Some claim that Gadamer’s nationalist isolation and global solidarity on the other later “rehabilitation of authority and tradition” in his (Harari 2020). masterpiece Truth and Method “may have begun as early One of the main assumptions in this paper is that as 1912 with the sinking of the Titanic” (Thiselton 2009: such traditional formulas, as ‘emergency powers,’ ‘state of exception,’ or the balancing between ‘personal liberty’ some interviews. Big events call for thinking out of the box. and ‘collective choice’ risk to be misleading, or even short- Interestingly,219). Gadamer Gadamer seems developed to confirm his this concepts interpretation of authority in sighted in the times of Covid-19. Most national legal systems and tradition at odds with the formalisms of Hobbes and have adopted extraordinary measures, such as massive the Enlightenment’s political thoughts. Authority entails lockdowns and forms of mass tracing; and yet, it would obedience and still, pace Hobbes, Gadamer presents the be deadly wrong to assume that such emergency powers latter as an act of reason that should be grasped in ac- exist because a sovereign is at work. This is only part of the cordance with its own historical dimension. This would story. What is at stake revolves around something deeper have been the framework, within which we should grasp than the traditional mechanisms of modern politics and such events as the sinking of Titanic. The “legitimacy of how international cooperation, or the empowerment of prejudices”––which Gadamer examines in the introduc- national citizens have aimed to address the limits of every tion to the section on authority and tradition in Truth and Hobbesian approach. Both alternatives lack perspective, Method––complements the problems of political power as if they were drawing a two-dimensional picture of that are at stake in the headquarters of sovereignty with “historical hermeneutics.” to speak, the three-dimensional relations between the In more recent times, there have been further histori- differenttoday’s legal elements and political of this crisis. crisis, In we order therefore to flesh need out, soto cal thresholds, such as Hobbes’s Invincible Armada, or properly set our level of abstraction. We have to formalize Gadamer’s Titanic. In my lifespan, I recall four of such the interface that makes an investigation of some crucial historical events, i.e. the missile crisis in Cuba (1961), the aspects of this crisis possible, through a set of observables

Law in Context, Vol 37, Issue 1, 2020 12 ISSN: 1839-4183 and variables of the analysis, the result of which provides 8. ON LIFE AND DEATH

The paper suggests that we should pay attention to the subject matter. Viruses are somehow in between what a model for the field. informational features of this crisis. The informational countsVirologists as life andstill non-living disagree on physical the definition entities. ofViruses their perspective has to include both the epistemological status are indeed equipped with genetics, reproduction and of the virus and the normative counterpart of the pandemic. By grasping the most relevant observables of the evolutionary functions through natural selection, and crisis in terms of information, the limits of most attempts yet, they don’t possess some of the features, which are to understand the legal and political facets of today’s typically conceived of as necessary conditions for life, pandemic in accordance with the traditional notions of such as cells (although viruses possess genomes). One sovereign powers, emergency measures, restriction of solution has been an ad hoc categorization of such entities, rights and liberties, will be clearer. Accordingly, the paper much as the biological status of a fungus is conceived as is divided into three sections. something of its own between animals and plants. The Section 2 sets the analysis of this paper in informational terms, as a matter of “life and death.” After all, in the of Viruses (ICTV) follows this strategy: “a virus species is phrasing of David Runciman, “as Hobbes knew, to exercise adefinition polythetic of class the International of viruses that Committee constitute on a replicatingTaxonomy political rule is to have the power of life and death over citizens” (Runciman 2020). The aim of this part of the lineage and occupy a particular ecological niche.” Experts examination is to stress what the epistemological status are still divided about the consistency of the formula (van of viruses and the normative counterpart of pandemics Regenmortel et al. 2013). Some suggest removing any may have in common. reference to the polythetic nature of the class (Gibbs and Section 3 presents the part destruens of the investiga- Gibbs 2006); others have proposed to substitute the poly- tion, namely, the limits of all traditional outlooks on the current political crisis and its legal boundaries. The limits as “a monophyletic group” (King et al. 2012). However, as concern either the lack of perspective that depends on a thethetic ICTV reference webpage to theis keen class to with warn the us, definition we should of virusesfurther short-term view, or on a short-sighted analysis. Bio-politics distinguish between what virus species and higher taxa applied to the information era is a good example of such are as “abstract concepts produced by rational thought limits of today’s debate (Aylmer 2020). and logic,” and the status of viruses as “real physical Section 4 illustrates the pars construens. We should add the notion of informational governance to the alternative entities produced by biological evolution and genetics.” between government and “governmentality of bio-politics” The “abstract concepts” particularly fruitful in the times (Foucault 1991). On the basis of this threefold viewpoint of a pandemic can be illustrated with a basic notion of on governments, governmentality, and governance, the complexity of the crisis should be grasped vis-à-vis the The modern debate about what life is should be traced the ICTV definition of virus, that is, biological evolution. epistemic, ontological, and normative dimensions of the back to the 1940s, when independently one from another, pandemic. Previous work on information as algorithmic great minds paved the way to the new paradigm. The compression of complexity (Chaitin 2005; Pagallo 2005), and the complexity of its normative challenges (Floridi connection between “real physical entities” and “biologi- 2015; Pagallo 2015a), casts light on how the elements of cal evolution and genetics” was formalised in terms of today’s crisis are in proportion one to another. The vanish- information. This is what Erwin Schrödinger’s What’s ing point of this informational perspective on sovereigns Life from 1944, John von Neumann’s 1948 lecture at the and viruses is given by the legal facets of their interplay. Hixon symposium on “cellular automatons,” and Norbert The intent is to illustrate a sort of Gadamer’s historical Wiener’s 1950 The Human Use of Human Beings have in hermeneutics for our current Titanic. common. Schrödinger conceived life as an encoded program, such as an “aperiodic crystal” with all the genetic

bonds (Schrödinger 1944). Von Neumann’s theory of automatainformation hinges for the on theconfiguration idea that the of laws covalent of physics chemical and

13 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 of biology can appropriately be understood as computer politics and legal systems in informational terms, has al- programs, that is, sets of mathematical instructions that lowed us to point out what is unique to the challenges of embed a plan of behaviour––which includes the ability “information ethics” (Floridi 2008), “techno-regulation” to reproduce itself––into an automaton (von Neumann (Pagallo 2015a), or the moral principles of AI (Floridi et 1951). In The Human Use of Human Beings, Wiener, the al. 2018). This work appears particularly fruitful in this father of cybernetics, expanded this informational view context, because the current Covid-19 crisis has put in to the realm of the traditional Aristotelian “practical sci- the spotlight three of the most relevant legal and political ences,” that is, ethics and politics. He warned against the features of today’s information societies and their agents. had contributed to create. “Information is information, is ‘complex.’ This means that such problems affect the normativenot matter challenges or energy. brought No materialism forth by the which same does field not he wholeThe first organization feature has and to environmentdo with a class of currentof problems societies. that admit this can survive at the present day” (Wiener 1954). This complexity is one of the main reasons why national A crucial element of this informational perspective is lawmaking activism has increasingly been short of breath given by the notion of ‘agent.’ The attention is not only over the past decades, and why constitutional powers of drawn to what an agent is, but to how different kinds national governments have been joined––and even re- of informational agents may interact in a given system. placed––by the network of competences and institutions Drawing on work by Allen, Varner and Zinser (2000), summarized by the idea of governance (Pagallo 2015b). further developed by Floridi and Sanders (2004), and The more an issue impacts on the whole infrastructure Pagallo (2013a), a physical entity can be understood as of the system, the more such issue is complex; but, the an agent if such entity is adaptable, interactive, and au- more an issue is complex, the less traditional notions of tonomous. Adaptability means that agents can improve legal and political thought can tackle such complexity in the rules through which their properties or inner states terms of physical sanctions, national jurisdictions or self- change. Interactivity stands for the agent’s perception referential rule of law. Classical examples of this complexity of the environment and its ability to respond to stimuli by changing the values of its inner states, or properties. governance (Pagallo 2015b). It seems fair to admit that theabound governance in the fields of pandemics of environmental should be law added and ofto internetthis list. or properties without external stimuli, so that agents exert Second, the class of problems we’re dealing with is someAutonomy degree concerns of control the overmodification their actions. of such The inner notion states of agent includes, in informational terms, both living beings information on the internet, go beyond i.e. trans the boundaries‘transnational,’ set up because by national viruses, and international much as the lawmak flow of- agent conceivable in between. We may also refer to the ers. National legislators have often attempted to stop this notionand artificial of “multi-agent entities, muchsystems.” as the Such affordance MAS concern of every the transmitted by biological viruses, or data by cyber empires. national sovereign state, an association, or a corporation Inflow both of transnationalcases, from a legal information, and political be it point the information of view, the (Floridibehaviour 2014). of traditional By grasping artificial the status legal of agents,an agent such in infor as a- dynamics does not revolve much around how national mational terms, we comprehend how different kinds of law and international law may interact, but rather, how agents may interact at different levels of abstraction. For today’s problems of national and international law present a transnational dimension. Scholars have extensively those informational agents that are made of (a piece of) scrutinized this dimension in today’s lex mercatoria, such codeexample, capable we may of replicating be interested with in detrimental some affinities effects. between This as the set of rules regulating the eBay trading system, or the powers of the internet corporation for the assignment The focus of this paper is restricted to the normative of names and numbers (ICANN). Transnational forms definition of agent includes both natural and cyber viruses. impact of such informational agents, and their systems. of data governance in e.g. the web of data (Pagallo et al. The representation of tenets of ethics and economics, 2019), and the health sector (Floridi et al. 2019), help us

Law in Context, Vol 37, Issue 1, 2020 14 ISSN: 1839-4183 clarify how national sovereign states fare in the times of a Others reckon, “in a lockdown, we can see the essence of pandemic. Several weeks of lockdowns and self-isolation politics is still what Hobbes described: some people get were and still are intertwined with this transnational to tell others what to do” (Runciman 2020). The times of dimension of our lives with its problems, e.g. the new pandemics make clear that our problems are quite more wave of cyber viruses and cyber attacks triggered by the complex. The next section aims to put things in perspec- massive use of video conference systems in Spring 2020. tive, by examining what role sovereign powers play in the Third, the transparent governance of a complex multi- Covid-19 crisis. It’s a matter of life and death, after all. agent system––such as today’s information societies––can ultimately hinge on the technicalities of design mechanisms. 9. BACK TO THE LEVIATHAN? This has been the bread and butter of two generations Sovereign powers of national states have played a major law, such as privacy and data protection, copyright and to Carl Schmitt’s 1922 formula, “Sovereign is he who of experts in the fields of information technology (IT) e-commerce, cyber crimes and so forth (Reidenberg 1998; decidesrole in the on firstthe exception”months of the(Schmitt Covid-19 1985). crisis. Since According January Lessig 1999). Legal constraints embedded into spaces, 2020, national governments, i.e. the executive powers of places, or digital interfaces, can either aim to nudge people national states have declared and enforced emergency (Thaler and Sunstein 2009), or to decrease the impact of measures, including orders and decrees for severe restric- harm-generating behaviours (e.g. risk management), or tions of rights (e.g. privacy), and liberties (e.g. association to prevent such harm-generating behaviours even from and economic freedoms). As a corollary of the sovereignty occurring (Pagallo 2011). Work on the ethical and legal principle, each state has made decisions in accordance with challenges of self-enforcing technologies should thus help us tackle current discussions about whether and to what both inwards (e.g. national lockdowns), and outwards extent new IT solutions for virus mass tracing, such as (e.g.its own borders specific control). laws. Emergency The President and martialof the US, laws Donald apply the Pan-European Privacy-Preserving Proximity Tracing Trump, for example, authorized entry restrictions from model (pepp-pt.org), or the Apple-Google design system, China by the late January 2020, later dubbing the crisis should be adopted. The role of design mechanisms in the as a “Chinese virus.” Then, travel bans were imposed by Trump on continental Europe by mid March 2020 and, of information and of transnational agents can hardly be governance of today’s information societies with their flux overestimated. European country imposing a “national quarantine” on However, as mentioned above in the introduction, 9a fewMarch days 2020, later, i.e. on Italy the UK soon and found Ireland itself as well.progressively The first scholars have often overlooked these facets of today’s legal isolated. The government had to take back control of the and political crisis. In some cases, this lack of attention is - on 17 March 2020. The most common remark of such over, can legitimately be––on the short-term aspects of daysnational was darling, ‘this is likethat a is, war.’ the flag carrier of Italy, Alitalia, justifiable since focus of the experts has been––and more the pandemic, e.g. how US emergency laws should strike Like any war, we got deaths. Some countries, as Spain a balance between public health and civil liberties (Gostin and Italy, announced the dramatic toll of the crisis in a and Hodge 2020). Yet, a considerable amount of comments formal and ritual way (at 6 p.m. of every given day in Italy, on the legal and political facets of the Covid-19 crisis has in the morning in Spain, whereas the cheer for frontline been simply short-sighted. Scholars often dismiss such workers was at 7 p.m. in New York City). Since 21 Janu- crucial aspects of today’s information societies, as their ary 2020, the website of the World Health Organization complexity, transnational dimension, and dependency on (WHO) delivers its “situation reports.” By 16 August, we different design solutions for the functioning of IT tech- already had 209 reports and, alas, 761.779 deaths (of nologies. Some claim the alternative is between “citizen which, 214.092 from the European region, 414.326 from empowerment” and “totalitarian surveillance,” between the Americas, etc.). The collapse of some national health “global solidarity” and “nationalist isolation” (Harari 2020).

15 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 systems was––and still is in certain cases––a concrete current analyses in the times of pandemic. Going back to possibility. As the Ancient Roman lawyer Cicero would the words of David Runciman, it is as if today’s information have suggested in our times of crisis, Salus Rei Publicae societies could simply offer “a choice between rival forms suprema lex esto, “the health of the people should be the of technocracy. In China, it is a government of engineers highest law” (De Legibus, book III). Most people have backed up by a one-party state. In the west, it is the rule indeed reacted in admirable and outstanding ways, and of economists and central bankers, operating within the still, we should ask how national governments exercise constraints of a democratic system” (Runciman 2020). their powers and how people will respond to this ex- Yet, there is the other side of the Covid-19 crisis. In ercise. Some grasp the current crisis as “the Faustian addition to the aims of “governmentality” and the devel- in the middle of the 17th century, when the country was be on the “empowerment of citizens” (Harari 2020), and bargain identified by the philosopher Thomas Hobbes being torn apart by a real civil war,” namely, the Faustian moreover,opment of onspecific how the technologies protection of of power, people’s focus rights should may political bargain between personal liberty and collective relate to further issues of authority, coercion, and politi- choice (Runciman 2020). Others suggest that we should cal legitimacy. In particular, as regards today’s state-of- add to the emergency powers of national governments, the-legal-art, it is worth mentioning that, over the past the analysis of that which Michel Foucault called “gov- decades, the European Court of Human Rights has ruled ernmentality” (Foucault 1979). This is the outlook that over the limits of state powers in times of emergency. was much discussed in the aftermath of the previous Some crucial provisions of the 1950 Convention on hu- crisis, that is, with September Eleven (Mbembe 2003; man rights refer indeed to “health” as the legal basis for Agamben 2004). What would “governmentality” add to restricting the exercise of such rights and freedoms as Art. 2’s “freedom of movement”; Art. 8’s “right to respect Governmentality focuses on power techniques that aim for private and family life”; Art. 9’s “freedom of thought, our previous analysis? to govern populations. Current lockdowns demonstrate that conscience and religion”; Art. 10’s “freedom of expression”; down to Art. 11’s “freedom of assembly and association.” of government that Foucault summed up as “biopolitics” Against the framework of the ECtHR case law, we may (Foucaultpopulations 1991). are the There subject would of specific be “two forms major or features” techniques of expect that e.g. its proportionality test will help us tackle this governmental rationality: the centralization of politi- some of the balances that shall be struck in the times of a cal power and the development of power technologies pandemic. Other legal sources may include the doctrine that aim to govern individual conduct continuously and of the European Court of Justice in Luxembourg, together permanently (Gutting and Oksala 2019). Foucault talked with statutes and case law of national jurisdictions. The about the latter as the individualising power, or “pastoral previous sections have already mentioned work on how to power” of modern life. His bio-examples on people’s diet, balance public health necessities and civil liberties in the mental health, or sexual practices can properly be comple- US legal system (Gostin and Hodge 2020), and the debate mented with today’s use of facial recognition software on data protection by design solutions in EU law. Crises do not only have to entail ‘zero sum’ games, but may also surveillance practices. Work on the “new surveillance” end up with ‘win win’ solutions (Pagallo 2013b). for social score ranking, methods of profiling, and mass society since the early 2000s (e.g. Marx 2002), and later, Whether or not current legal safeguards will be good the age of “surveillance capitalism” (Zuboff 2019), can be enough to tackle the Covid-19 crisis remains of course a understood as the footnotes, or the updating, of Foucault’s mid-term issue. The evaluation depends––also but not analysis on a new historical form of “governmentality” only––on the second phase of the crisis with the reacti- vation of essential economic activities. Much as occurs the nature of society and the markets. This account of in Plato’s Republic (369 C), when Socrates explains to with specific technologies of power and beliefs about neoliberal societies, which hinge on competitiveness Glaucon “the origins of the state,” crises force us to trace and the self-interest of economic subjects, frames many the basic principles of mutual need and division of labour,

Law in Context, Vol 37, Issue 1, 2020 16 ISSN: 1839-4183

- est of necessities is food, which is the condition of life the crisis mistakes two different kinds of problem. The andthereby existence... providing The for second priorities: is a “Now dwelling, the first and and the great third to the first extraordinary legal measures in response to- clothing and the like.” Although Socrates and his pupil litical authority and legitimacy, that is, the Gordian knot Plato have been often depicted as idealistic thinkers at offirst coercion issue regards and its counterpartthe aforementioned obedience: questions in chapter of po 21 odds with the realism of modern legal philosophy, e.g. of Leviathan, for instance, Hobbes concedes, “the obliga- Hobbes and the Realpolitik approach of several current tion of subjects to the sovereign is understood to last as commentators, both natural and positive law traditions long, and no longer, than the power lasteth by which he have to address the basic question on who should use is able to protect them” (Hobbes ed. 1999). Such ability, coercive power. In the phrasing of Socrates, “nothing can duty, or self-interest of sovereigns begs however a further be more important than that the work of a soldier should question. In addition to the extent to which a sovereign’s be well done” (Republic, 373 E). subjects believe their sovereign is able to “protect” them, As stressed above in connection with the tenets of e.g. the opinion of the US population on Donald Trump’s “bio-politics” and “governmentality,” such a use of coer- management of the Covid-19 crisis, we may wonder to cive power involves problems of political authority and what extent current sovereigns can really protect “their” legitimacy, i.e. who controls the soldiers in the story of population, and furthermore, what kind of powers a sover- Socrates in the Republic, much as the evolution of the eign should have, in order to attain such end today. Could techniques, according to which the work of a soldier can be considered as “well done.” After all, one of the crucial the problems we’re coping with (Floridi 2020), or should current debate on “digital sovereignty” find the answer for features of today’s information societies has to do with we reconstruct our rights and institutions from the very the fact that physical coercion has increasingly been joined––and even substituted––by new forms of digital Admittedly, since the French legal philosopher Jean beginning (Poblet et al. 2019)? coercion. Military techniques have evolved since Plato’s times, much as Foucault’s forms of governmental ratio- the right of he who only depends on his “own sword,” nality have increasingly been transformed into matters theBodin meaning first defined of sovereignty the notion has of evolvedsovereignty and inadapted 1576, asto of access to and control over data in the information era. multiple political contexts. Sovereignty embraces forms Work on cyber force and cyber wars (Pagallo 2015c), of direct democracy (Rousseau), dictatorship (Schmitt), digital retaliation (Taddeo 2017), workplace and techno- or the power of constitutional republics (e.g. Art. 1, 7 regulation (Brownsword and Yeung 2007), up to current and 11 of the Italian constitution). From a theoretical view, scholars have widely discussed whether the set of 2009), illustrate this trend. The digitalization of current traditional powers attributed to sovereigns represents a societies“pastoral canpowers” be grasped of profiling as a set(Hildebrandt of constraints and Gutwirthand pos- “polythetic class” or a “monophyletic group,” similar to sibilities that transform or reshape the environment of that which happens with the epistemological debate on people’s interaction, such as their political institutions and the status of biological viruses. From a practical viewpoint, legal systems, with corresponding problems of legitimacy nevertheless, it seems fair to concede that, whether poly- and authority, trust and coercion, rights and obedience thetic or monophyletic, the list of sovereign powers has (Floridi 2014; Pagallo 2015a; Poblet et al. 2019). been shrinking over the past decades. Still, the lockdowns of today’s pandemic have dra- Section 2 illustrated some of the reasons for this trend, matically juxtaposed the digital and physical bodies of drawing the attention to the complex nature of problems that are more often transnational, and increasingly data- sight impression that what really matters when crises driven and ICT dependent. Next, the aim of Section 4 is to hitthe hardpopulation. is the use This of juxtaposition coercive power has overcreated the thephysical first- further explore this scenario, by examining the ontological, body of the population (Runciman 2020). Yet, this kind epistemic, and normative constraints of sovereign pow- of bio-political approach to the threats of pandemics and ers during the Covid-19 crisis. The centralized powers of

17 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 national states have to be examined in connection with better understood as an “anarchical society” (Bull 1977); the scale of the problems we’re dealing with, whereas or, as a form of “governance without government” (Rosenau sovereigns can only tackle part of this huge complexity. and Czempiel 1992). The current Covid-19 pandemic has adjourned this debate on the redistribution of powers IT law, much as anti-terrorism and transnational business that occurs through coordination mechanisms in the in- regulationsLessons learned can shedfrom lightthe fields on what of environmental is beyond the lawcontrol and ternational arena, e.g. the role and powers of the World of national governments with their emergency powers. A Health Organization in the current crisis (Buranyi 2020). model of governance for today’s pandemic should follow Still, one of the lessons learned over the past decades in as a result of this analysis. (Roach 2011), is that dealing with phenomena that cross some fields of international law, such as anti-terrorism law 10. THE GOVERNANCE OF PANDEMICS or do not know borders, any attempt at legal regulation The governance of some critical aspects of today’s inthat involves national law and does not include elements formation societies help us put current emergency powers of international law, comparative law, or transnational of national sovereign states in perspective. In addition to law, is bound to fail. that which national governments and governmental ra- In order to understand why this is the case in the times tionality are supposed to do in this context, focus should of a pandemic, attention should be drawn to the sets of be on the wider framework, according to which we can constraints that arise during the decision-making process appreciate both the scale of the problems we are dealing of both national sovereigns and governance actors. Such with, and hence, the limits of the powers exercised by constraints regard either the nature of the issues under sovereign states. Such limits concern (i) mechanisms of scrutiny (ontological constraints), or the knowledge we coordination, in addition to the exercise of coercive power; have vis-à-vis its degree of validation (epistemic con- (ii) transnational and international dimensions of the law, straints), or the set of values, rights, or principles that in addition to the national boundaries of sovereign states; are at stake in the decision-making process (normative and, (iii) the set of constraints that both governance actors constraints). The cumulative effect of such constraints and national governments have to face vis-à-vis problems should make clear the limits of Hobbes’s sovereign pow- that affect the whole infrastructure and ecosystem of our ers in protecting “its” population. societies. Accordingly, this section is divided into three From an ontological point of view, we already stressed parts on (i) the requirements of governance, namely, what that the class of problems under scrutiny in this paper, governance is supposed to be; (ii) its functions, i.e. what e.g. the legal and political challenges of the Covid-19 governance is supposed to do; and, (iii) both solutions crisis tend to cross or ignore borders. In general terms, and open issues of today’s informational governance. environmental issues, internet protocols, terrorists, or viruses impact the whole organization and ecosystem of 4.1 REQUIREMENTS of science and societies affects how we should grasp and According to the jargon of international law experts, manageour societies. this complexity. From an epistemic In addition view, to currentthe datafication debates governance can be understood as the set of formal and on virus mass tracing, the information revolution has al- informal rules through which decisions are made and political authority is exercised (Pagallo 2015a). At the inclusiveness, aversion of risk, responsiveness, and data ready triggered new problems of adaptability, flexibility, which sovereigns as well as societal and economical actors the e-health sector (Blasimme and Vayena 2019; Pagallo interact.international Contrary law level,to some such realistic rules accountsdefine the of space, interna in- etmonitoring, al. 2019). that From are a particularly normative relevantperspective, in such it isfields, worth as tional law as a sort of Hobbesian state of nature, some mentioning how often transnational law and mechanisms insist on the role of coordination and even cooperation of coordination or cooperation complement national and in international affairs: the world political order could be international regulations through contracts, rather than

Law in Context, Vol 37, Issue 1, 2020 18 ISSN: 1839-4183 international agreements, or via “spontaneous orders,” such interaction between multiple regulatory systems rather than simple political constructivism (Murray 2007; Post 2009; Solum 2009; Reed 2012; Pagallo 2015a; etc.). in competition at different magnitudes of complexity? Whereas inter-national law refers to the legal relations 4.3 SOLUTIONS AND OPEN PROBLEMS “between” national states, trans-national law refers to We already mentioned that coordination and coopera- relations that go either “through,” or “beyond” national tion mechanisms, rather than techniques of coercion, are boundaries. Section 2 has already stressed that several more effective to tackle the ontological, epistemic, and weeks of lockdowns and self-isolation during the Covid-19 normative constraints of national governments. Some crisis were––and still are––incessantly accompanied by philosophers, or tycoons like Bill Gates, have insisted on this transnational dimension of our lives with its problems. the role of international cooperation and “global solidarity” in these cases (Harari 2020). Work on coordination 4.2 FUNCTIONS mechanisms illustrates the manifold ways in which new What governance is called to do in addition to national forms of governance, over the past decades, have already governments and their international organizations is to complemented––and even substituted––the top-down set the proper level of abstraction for the balances that regulations of sovereign powers for a class of problems shall be struck between multiple regulatory systems in that is complex, often transnational, and increasingly competition. As stressed above in Section 3, following data-driven. The list of legal sectors include also but not the story of Socrates about the “origins of the state,” such only data protection and internet law, anti-terrorism and regulatory systems in competition include the military transnational business law, ICTs and environmental regu- sector and politics, the forces of the market and of social norms. The Covid-19 crisis has made such competition, with the Web of Data and data protection (Pagallo et al. and the corresponding problems of balancing, apparent. lations. Each field has its model of governance, as occurs Sovereigns have had to strike dramatic balances between, socio-cognitive and socio-technical systems (Andrighetto for example, when reopening from the Covid-19 crisis––in et2019), al. 2013), linked human democracy computing (Poblet interaction et al. 2019), (Fredericks artificial et order to protect business and the economic sector––and the social acceptability for the number of deaths due to health sector (Casanovas et al. 2017; Pagallo et al. 2019). such reopening, which may trigger further problems of al. 2016), or the use of Artificial Intelligence (AI) in the- social cohesion. dination and cooperation mechanisms through which Still, each field of regulation relies on the power of coor Paradoxically, the exercise of sovereign powers with multiple forms of legal co-regulation govern individual and lockdowns, emergency laws, curfews, or quarantines coordination mechanisms include both multi-stakeholder social behaviour. For example, in the field of AI governance, unintended consequence. Dramatic drops in pollution mechanisms upstream for risk mitigation, e.g. unwanted andhas comegreenhouse with the gas eco-benefits emissions ofwent such hand-in-hand measures as anin consequences of human-AI interaction, and systems for many parts of the world with the beginning of the most user-driven benchmarking of marketed offerings that devastating economic crisis since World War II. Much allow trust in products and services as well as providers as occurs with national governments, we can thus say to be measured and shared (Floridi et al. 2018). that governance should be effective in striking balances In order to prevent another misapprehension in today’s between multiple regulatory systems, such as econom- debate, however, we should not overlook four different ics and social mores (Pagallo and Durante 2016). What ways in which the connection between the power of changes with the balances that shall be struck in the time coordination mechanisms and the power of sovereign of a pandemic is the scale at which national lawmakers states can be grasped. All these ways were already under scrutiny in this paper. They regard: corresponding moral dilemmas. How should we grasp and governance actors define such balances with their

19 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 i. Coordination as the condition and basis for any sov- of the most complex challenges of today’s information ereign’s act of coercion, e.g. Plato’s account on the societies are, or should be, addressed. Such different scales “origins of the state” in the Republic; of the issues that legal and political systems have to face ii. Coordination as the result of coercive sovereign powers, today have complemented the role of governments and e.g. Foucault’s techniques of governmental rationality the technicalities of “governmentality” through a network and the “pastoral power” of modern life; of coordination and cooperation mechanisms that often goes “beyond,” or “through” national sovereign borders iii. Coordination as opposed to such coercive sovereign powers, e.g. today’s research on the “spontaneous (Pagallo 2015b). Although the exercise of extraordinary orders” of the internet; and, powers during the Covid-19 crisis may end up with the disruption of some of such transnational or international iv. Coordination as complementary to coercion. networks, this scenario appears most of the time unlikely, The latter has been our stance in this paper. By stress- or self-defeating, because such networks exist and have ing the limits of the power of national governments and been developed over the past decades, due to the limits sovereign states in the current Covid-19 crisis, by no that national sovereign powers reveal, when coping with means the intent has been to underestimate such a state agents that do not know or simply cross borders: viruses, power. After all, this power has been in full display since internet surfers, money launderers, or terrorists. This is not to say that we can ignore the open problems still have a crucial role in re-establishing the public health of today’s environmental and transnational business regu- sector,the first and weeks addressing of the pandemic. the colossal Sovereigns challenges have of economic had and lations, together with the troubles with anti-terrorism, reconstruction. National governments and international data protection or internet governance. Cooperation institutions shall grapple simultaneously with the most and coordination mechanisms and new forms of legal devastating public health and economic crisis of a lifetime. co-regulation are not the magic bullet. As stressed in this According to the Letter to G20 governments signed by section, these forms of transnational and international Bertie Ahern and 200 other former leaders on 6 April governance complement, but not erase the powers of 2020, the global effort should include $ 1 billion urgently sovereign states. Their interaction should be grasped for WHO, $ 3 billion for vaccines, $ 2.25 billion for thera- at the different scale of the problems we are facing with peutics, and a further $ 35 billion to support countries the regulation of the internet and the big companies of with weaker health systems and vulnerable populations. Silicon Valley, the digital economy and the protection of Moreover, the international community should waive this year’s poorer countries’ debt repayments, including or of terrorism, down to the existential threats brought $44 billion due from Africa, whereas the World Bank and aboutthe environment, by viruses and the pandemics. transnational This borders realignment of finance, of the many regional development banks shall be recapitalized. legal sources does not wipe out the powers of sovereign The exercise of sovereign powers during the Covid-19 states. Should such powers disappear, there would be no legal space to go through, or beyond. ofcrisis national is however state powersa necessary, when yet dealing insufficient with the ingredient scale of 5. CONCLUSIONS problemsof the analysis. that are Section complex, 2 transnational,has illustrated and the data-driven. inefficacy We shouldn’t have ignored the risk of a new pandemic, Section 3 insisted on the legal limits of sovereignty, even after the SARS crisis from 2002-2004, and the 2009 Mexi- in the times of the Covid-19 crisis, pursuant to the doctrine of Courts and some basic principles of the rule of 2020, and most societies were taken aback, simply with law. Section 4.2 inspected the ontological, epistemic and theircan swine guard flu off. pandemic. Epic crises The entail bell huge however decisions, rang inand January hence normative constraints that affect both rights and duties force us to think hard. The paper has focused on the legal of national sovereigns, much as the scale at which some and political impact of the Covid-19 crisis, drawing the

Law in Context, Vol 37, Issue 1, 2020 20 ISSN: 1839-4183 attention to some fundamental questions on authority and rather added to the previous issues of transnational busi- legitimacy, coercion and obligation, power and coopera- ness regulation, anti-terrorism and internet governance, tion. National states and sovereign governments have had or the protection of the environment, that which Yuval and still will have a crucial role in addressing the colossal Harari calls the “two particularly important choices” of challenges of economic reconstruction. Scholars have this pandemic, that is, between totalitarian surveillance accordingly discussed the set of legal means displayed and citizen empowerment, and between nationalist isola- during the crisis: emergency decrees, lockdowns, travel tion and global solidarity (Harari 2020). bans, or quarantines. Against this kind of debate, the These big choices, as an act of reason, have to be grasped aim of this paper has been twofold, namely, (i) to stress in accordance with their historical dimension. This was the limits of current perspectives on governments and “governmentality techniques” of bio-politics; and, (ii) to on the catastrophe of the Titanic. The historical dimen- illustrate what goes beyond or through such boundaries thesion lesson suggests that that Hans-Georg we should Gamader similarly learned, grasp the reflecting current of national sovereign states. crisis in accordance with the lessons learnt in a world that On the one hand, we shed light on the limits of today’s already is interdependent from an ecological, technologi- debate with the set of ontological, epistemic and norma- cal, and economic viewpoint. The legal and political bal- tive constraints that affect rights and duties of national ances that shall be struck in the times of this pandemic, sovereigns. The governance of this pandemic has shown between liberty and public choice, empowerment and the nature of these limits. First, the class of problems we surveillance, solidarity and isolation, are going to take are addressing is complex, because such issues affect the place in societies that already have systemic problems that whole organization and environment of current societ- regularly go through, or beyond national legal boundaries. ies. Second, the dimension of the crisis is transnational The Covid-19 crisis has dramatically put such properties because it concerns agents and legal systems that do not of today’s information societies in the spotlight: viruses know or cross national borders. Third, the normative affect their whole infrastructure, do not know borders, constraints of the crisis reminded us of how, more often, and should be tackled with the best of our data-driven the legal problems of our societies revolve around matters technologies. Every particularly important choice of this of access to––and control and protection over––data and pandemic––which does not take into account the scale information in digital environments. Current discussions and complexity of the problems we are confronted with on IT solutions for virus mass tracing, such as the Google in the information era––is simply doomed to fail. and Apple initiative to trace people’s contacts with the virus, stress the fact that our societies are not simply re- 6. REFERENCES lated to the use of ICTs, but rather, they are data-driven 1. Agamben, Giorgio (2004) State of Exception, translated by and ICT-dependent. Kevin Attell. The University of Chicago Press. On the other hand, the paper aimed to stress that many 2. Allen, Colin, Gary Varner, and Jason Zinser (2000) Prolegomena crucial features of the current crisis can be addressed through Journal of Experimental the lessons learnt from previous work on environmental and Theoretical Artificial Intelligence 12: 251–261. law, internet governance, anti-terrorism and transnational 3. Andrighetto,to any future Giulia,artificial Governatori, moral agent. Guido, Noriega, Pablo and - Leon van der Torre, Normative Multi-agent Systems, Schloss oped over the past decades models of governance that Dagstuhl-Leibniz-Zentrum für Informatik, 2013, vol. 4. hingebusiness on differentlaw, or data forms protection. of co-regulation These fields and networks have devel of 4. Aylmer, Carey (2020) Coronavirus: A Political Tool of the coordination and cooperation mechanisms, as a response West, Guardian (Sydney), Issue 1906, March 9. Available at to the troubles of national states with problems that are complex, often transnational, and increasingly data-driven. https://search.informit.com.au/documentSummary;dn=0 Such sources of transnational or international law do not 24036551549530;res=IELAPA. entail that all problems are fixed. The Covid-19 crisis has

21 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

5. Blasimme, Alberto and Effy Vayena (2019) The Ethics of AI 20. Foucault, Michel (1991) Governmentality, trans. Rosi Braid- in Biomedical Research, Patient Care and Public Health, April otti and revised by Colin Gordon, in Graham Burchell, Colin Gordon and Peter Miller (eds), The Foucault Effect: Studies 6. Bogdandy, Armin von and Pedro Villareal (2020) International in Governmentality, pp. 87–104. Chicago, IL: University of Law9. Available on Pandemic at SSRN: Response: https://ssrn.com/abstract=. A First Stocktaking in Light of Chicago Press. the Coronavirus Crisis, Max Planck Institute for Comparative 21. Fredericks, Joel, Caldwell, Glenda and Martin Tomitsch, Public Law & International Law (MPIL), Research Paper No. Middle-out Design: Collaborative Community Engagement 2020-07. Available at https://papers.ssrn.com/sol3/papers. in Urban HCI. In OzCHI ’16 Proceedings of the 28th Australian Conference on Computer-Human Interaction, pp. 200-204, 7. Brownsword, R. and Karen Yeung (eds.) (2007) Regulating Launceston, TAS, 2016. cfm?abstract_id=3561650. Technologies: Legal Futures, Regulatory Frames and Techno- 22. logical Fixes, London: Hart. the “virus species”, Archives of Virology, 151: 1419–1422. Gibbs, A.J. and M.J. Gibbs (2006) A broader definition of 8. Bull, Hedley (1977) The Anarchy Society. A Study of Order in 23. Gostin, Lawrence O. and James G. Hodge Jr. (2020) US Emer- World Politics. Cambridge University Press. gency Legal Responses to Novel Coronavirus: Balancing Public 9. Buranyi, Stephen (2020) The WHO v. Coronavirus: Why It Health and Civil Liberties, Jama, February 13. Available at Can’t Tackle the Pandemic, The Guardian, April 10. Available doi: 10.1001/jama.2020.2025. at https://www.theguardian.com/news/2020/apr/10/ 24. Gutting, Gary and Oksala, Johanna (2019) Michel Foucault, world-health-organization-who-v-coronavirus-why-it-cant- The Stanford Encyclopedia of Philosophy (Spring 2019 Edi- handle-pandemic. 10. Casanovas, Pompeu, Mendelson, Danuta and Marta Poblet, edu/archives/spr2019/entries/foucault/>. A Linked Democracy Approach for Regulating Public Health 25. Harari,tion), Edward Yuval N.Noah Zalta (2020) (ed.), URLThe =world Routledge-Cavendish. Risks, Principles, and Recommendations, Minds and Ma- 33. Pagallo, Ugo (2005) Introduzione alla filosofia digitale. chines, 28(4): 689-707. Torino: Giappichelli. 18. Floridi, Luciano, Luetge, Christoph, Pagallo, Ugo et al. (2019) 34. Pagallo, Ugo (2011) Designing Data Protection Safeguards Key Ethical Challenges in the European Medical Information Ethically, Information, 2(2): 247-265. Framework, 29: 355–371. Minds & Machines, 35. Pagallo, Ugo (2013a) The Laws of Robots: Crimes, Contracts, 19. Foucault, Michel (1979) The Birth of Biopolitics, trans. Gra- and Torts. Dordrecht: Springer. ham Burchell, New York: St. Martin’s Press.

Law in Context, Vol 37, Issue 1, 2020 22 ISSN: 1839-4183

36. Pagallo, Ugo (2013b) Online Security and the Protection 47. Rosenau, J.N. and E.O. Czempiel (1992) Governance without of Civil Rights: A Legal Overview, Philosophy & Technology, Government: Order and Change in World Politics. Cambridge 26(4): 381-395. University Press. 37. Pagallo, Ugo (2015a) Good Onlife Governance: On Law, 48. Runciman, David (2020) Coronavirus has not suspended politics Spontaneous Orders, and Design. In Luciano Floridi (ed.), – it has revealed the nature of power, The Guardian, March 27. The Onlife Manifesto: Being Human in a Hyperconnected Era, Available at https://www.theguardian.com/commentisfree/2020/ pp. 161-177. Dordrecht: Springer. mar/27/coronavirus-politics-lockdown-hobbes. 38. Pagallo, Ugo (2015b) The Realignment of the Sources of the 49. Schmitt, Carl (1985) Political Theology: Four Chapters on the Law and their Meaning in an Information Society, Philosophy Concept of Sovereignty, trans. by George D. Schwab, MIT Press. & Technology 28(1): 57-73. 50. Schörindger, Erwin (1944) What’s Life?, Cambridge: Cam- 39. Pagallo, Ugo (2015c) Cyber Force and the Role of Sovereign bridge University Press. States in Informational Warfare, Philosophy & Technology, 51. Solum, Lawrence B. (2009), Models of Internet Governance, 28(3): 407-425. in L. A. Bygrave and J. Bing (eds.), Internet Governance: In- 40. Pagallo, Ugo and Massimo Durante (2016) The Pros and frastructure and Institutions, pp. 48-91. New York: Oxford Cons of Legal Automation and its Governance, European University Press. Journal of Risk Regulation, 7(2): 323-334. 52. Stock, James H. (2020) Coronavirus Data Gaps and the Policy 41. Pagallo, Ugo, Casanovas, Pompeu and Robert Madelin (2019) Response to the Novel Coronavirus, March 23. Available at The Middle-out Approach: Assessing models of legal govern- http://www.igmchicago.org/wp-content/uploads/2020/03/

of Data, The Theory and Practice of Legislation, 7(1): 1-25, 53. Taddeo, Mariarosaria (2018) The Limits of Deterrence ance in data protection, artificial intelligence, and the Web coronavirus_shutdown_dynamics_econ_v2.pdf. DOI: 10.1080/20508840.2019.1664543 Theory in Cyberspace, Philosophy & Technology, 31: 339–355. 42. Poblet, Marta, Casanovas, Pompeu and Victor Rodríguez- 54. Thaler, Richard H., and Cass R. Sunstein (2009) Nudge: Im- Doncel, Linked Democracy. Springer, Dordrecht, 2019. proving Decisions about Health, Wealth, and Happiness. Rev. 43. Post, David (2009) In Search of Jefferson’s Moose: Notes on and expanded ed. New York: Penguin Books. the State of Cyberspace, New York: Oxford University Press. 55. Thiselton, Anthony C. (2009) Hermeneutics: An Introduction. 44. Reed, Chis (2012) Making Laws for Cyberspace. Oxford: Eerdmans Publishing. Oxford University Press. 56. 45. Reidenberg, Joel R. (1998) Lex Informatica: The Formulation Calisher, Charles H. et al. (2013) Virus species polemics: of Information Policy Rules through Technology, Texas Law 14Van senior Regenmortel, virologists Marc oppose H.V., Ackermann, a proposed Hans=Wolfgang, change to the Review, 76(3): 553-594. Archives of Virology, 158: 46. Roach, Kent (2011) The 9/11 Effect: Comparative Counter- 1115–1119. ICTV definition of virus species. terrorism. Cambridge University Press. 57. Von Neumann, John (1951) The general and logical theory of automata, Cerebral Mechanisms in Behavior: The Hixon Symposium, edited by L.A. Jeffress, pp. 1-31. New York: John Wiley & Sons.

23 Law in Context, Vol 37, Issue 1, 2020 Received: November 25, Date of acceptance: December 15, 2020, Date of publication: December 23 2020, DOI: https://doi.org/10.26826/law-in-context.v37i1.126

Knowledge Graphs: Trust, Privacy, and Transparency from a Legal Governance Approach

By Daniel Schwabe1, Orcid: https://orcid. org/0000-0003-4347-2940 Professor at Department of Informatics (INF), Pontifical Catholic University, Rio de Janeiro, Carlos Laufer1, Associate Researcher, TecWeb Lab, at Dept. of Informatics, PUC-Rio, Orcid: https://orcid.org/0000-0002-2606-4067 and Pompeu Casanovas2, Research Professor at La Trobe University Law School, Bundoora, Melbourne, Australia, Orcid: https://orcid. org/0000-0002-0980-2371 1

2 La Trobe University, Melbourne, Australia Pontifical Catholic University, Rio de Janeiro, Brazil

ABSTRACT This paper presents the Knowledge Graph Usage framework, which allows the introduction of Knowledge Graph features to support Trust, Privacy, Transparency and Accountability concerns regarding the use of its contents by applications. A real-world example is presented and used to illustrate how the framework can be used. This article also shows how knowledge graphs can be linked to the elements of legal governance. Thus, it is an invitation to dialogue for legal and Law & Society scholars who might be interested in how the evolution of the web of data and computational sciences intersects with their own discipline.

Keywords – Knowledge Graphs, Privacy, Transparency, Trust, Legal Governance, Legal interpretation

Acknowledgements: A former version of this paper deprived from its present socio-legal dimension has been presented by Daniel Schwabe and Carlos Laufer with the title “Trust and Privacy in Knowledge Graphs” at The Web Conference (WWW’19), May 13, San Francisco, USA. See: Proceedings of WWW '19. ACM, New York, NY, USA. Daniel Schwabe was partially supported by a grant from CNPq. The present work by Pompeu Casanovas has been carried out for the EU H2020 Programme LYNX, Legal Knowledge Graph for Multilingual Compliance Services. The interested reader can find a description of this project as a Research Note at the end of this volume, LiC 37 (1). Disclosure statement – No potential conflict of interest was reported by the authors. License – This work is under Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) https://creativecommons.org/licenses/by-nc-sa/4.0/ Suggested citation: Schwabe, D., Laufer, C., Casanovas, P. (2020). “Knowledge Graphs: Trust, Privacy, and Transparency from a Legal Governance Approach”, Law in Context, 37 (1): 24-41 , DOI: https://doi.org/10.26826/law-in-context.v37i1.126 Summary 1. Introduction 5. A Summary of the KG Usage Framework 2. A Legal Governance Approach 5.1 KI Representation 2.1 Components of Legal Governance 5.2 Controlling Usage 2.2 A Preliminary Framework 5.3 Rules 3. Background Concepts 5.4 Conflict Resolution 3.1 Trust 3.2 Privacy 6. Example Scenario revisited 3.3 Transparency 7. Conclusions 4. An illustration—Disaster Relief Donation 8. References

Law in Context, Vol 37, Issue 1, 2020 24 ISSN: 1839-4183

4. INTRODUCTION data, which is a fundamental part of data in some domains The advent of the Web has enabled the generation of such as life-sciences (Kuhn et al. 2018). a vast amount of information, mostly in the form of in- KGs differ also on the way they are built and populated. terlinked pages with texts containing links (references) A few are curated (e.g., CYC), others rely on crowdsourced to other texts. Users accessed information by loading a information (e.g. Wikidata, and several, perhaps most, extract information from structured, semi-structured or textual information harvested from the Web). pagesocial - networks identified added by a end-user-generatedURL - in their browsers content, following further The multiplicity of sources and various extraction increasinglinks in the thepage amount to find relatedof textual information. information The available. advent of approaches naturally raises the issue of data quality and However, it quickly became evident that search function- confronts the user of the data in the KG with the issue of trusting, or not, the information contained in the KG. ushering the creation and growth of search engines. While For some types of information, for example in case of searchality was engines needed became to help hugely users findsuccessful, relevant it information,also became online reviews and online and social media, this trust can clear that searching strings contained within Web pages have a direct effect on commercial success (e.g. Angella was a limited approach to solve the actual problems that and Johnson 2016). This highlights the fact that data ultimately expresses a belief, opinion or point of view of a certain need. As stated by Google, people are interested some agent – the author. inusers “things, have, not which strings”. is to find1 For information example, users that iswould relevant like for to From a broader perspective, information (and knowl- distinguish between “Sydney” as a city and “Sydney” as the edge) is said to have become the prime resource in the name of a known person (e.g., “Sydney Pollack”, “Sydney Third Industrial Revolution, also called the Digital Age Poitier”, “Sydney Lumet”), a football club (“Sydney F..C”), – when digital technologies enabled new ways of generat- etc…, among many other possible meanings. ing, processing and sharing information (Castells 2010, the term (Ehrlinger and Wöß 2016, Hogan et al, 2020), we move into the Fourth Industrial Revolution (4IR) (Schwab adoptAlthough the view to thisthat daya Knowledge there isn’t Graph a precise (KG) definition represents of 2017).Rifkin 2011),The 4IR and is characterized is becoming by even a fusion more of central technologies, as we a network of interlinked descriptions of entities (objects, which is blurring the lines between the physical, digital, events, concepts etc.)-- a graph-theoretic representation and biological spheres. of human knowledge such that it can be ingested with Increasingly, systems and applications operate in semantics by a machine (Kejriwal 2019). Graph-based databases have been available for a long bearing on daily lives of billions of people, where two time2 and many more continue to be created on a regular fundamentala context in whichcharacteristics the flow of of the information use of such has informa direct- basis – see (Hogan et al, 2020) for a survey. The original tion emerge – Transparency and Privacy. Transparency vision for the WWW was later enriched to form the Se- is the quality that allows participants of a community to mantic Web, instances of which can also be regarded as know what the particular processes and agents are being a KG—see, for example, the survey by Gandon (2018). used in its functioning. It is generally regarded as a means While the graph model or some variant has been to enable checks and balances within this community, used in several KGs, it has already been observed that ultimately providing a basis for trust among participants using only atomic (indivisible) nodes as the “granule” of that community. When the community is regarded as being the entire society, these checks and balances are of information, such as events, or time-varying data. For of information is insufficient to express complex types of the parties involved. organized around Items described by a collection state- reflectedOne of inthe its mechanisms political system created to to prevent increase misuse transparency by any mentsexample, (Erxleben Wikidata et al.(Vrandečić 2014). Another and Krötzsch reason for 2014) having is in political systems is the enactment of regulations ensur- more complex “granules” is recording provenance (meta) ing the right of its members to access to information in a variety of contexts, ranging from government-produced

1 https://googleblog.blogspot.com/2012/05/introducing-knowledge graph-things-not.html 2 For instance, Wordnet (Miller 1995), DBPedia (Lehmann et al. 2013), Yago (Suchanek et al. 2007), CYC (Lenat 1995), NELL (Carlson et al 2007), ConceptNet (Speer and Havasi 2010)

25 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 information and data to consumer-related information should be introduced, and whether the Privacy Act should regarding goods and products, as well as the right of in- include a ‘right to erasure’. dividuals to freely create, publish and access information. It is our contention that these and similar needed As we will explain later, regulations do not comprehend reforms around the world should consider the technical only legislation, i.e. enacted laws and statutes, but all kinds developments occurring in computer science and seman- of legal instruments of governance—hard law, soft law, tic web studies. In order to deal with the myriad of often policies and ethics. and systems should incorporate adequate mechanisms toconflicting ensure compliance cross-cutting of concerns, both ethical Internet and legalapplications princi- (UniversalThe free Declaration flow of information, of Human on Rights the other 1948, hand, art. may12). ples. In order to be effective, we claim that the use of conflict with another basic human right, that of Privacy Knowledge Graphs ought to provide support for these and Zannone 2018), but in essence they all refer to the concerns—trust, privacy and transparency. In this paper rightThere of are an many individual definitions to control for Privacy how information (Paci, Squicciarini, about we propose a framework that enables this support. In her/him is used by others. Data Protection and Privacy Section 2 the regulatory framework approach and some have had, and still have, different regulatory regimes that are applied in a variety of jurisdictions in both Civil and Common Law legal systems. For instance, Privacy is background.concepts stemming Section from 3 backgrounds the work already three done main in thegeneral field considered a fundamental (human) right by the recent concepts—trust,will be defined. It privacy, will also and describe transparency—that the preliminary will legal be General Data Protection Regulation (GDPR) enacted in taken into account in our modelling. Section 4 presents Europe in May 2018. It is also assumed as a constitutional an illustration by way of example, and Section 5 discusses right in many EU countries. But this is not so in the USA, the proposed KG representation framework, showing its in which privacy and data protection do not qualify as previous example, and Section 7 draws some conclusions The situation in Common Law countries is nuanced, andapplication points to in future the illustration. work. Section 6 reflects on the varyingspecified with fundamental some subtlety rights between under the different Constitution. national jurisdictions. For instance, there is no general law right 5. A LEGAL GOVERNANCE APPROACH to privacy in Australia. Although Australia is a signatory In this section, we argue that the privacy legal regula- to the International Covenant on Civil and Political Rights tory framework does not solely lean on national legislation (adopted by UN General Assembly on 16 December 1966, but encompasses other legal instruments of transnational and in force from 23 March 1976), the international law nature (such as protocols, standards, best practices and right to privacy conferred under Article 17 of the ICCPR - has not been enacted into Australia’s domestic law (Watts cially suitable for a broader legal governance, in which and Casanovas 2018). The Privacy Act 1988 still regu- hetero-regulatory,ethical principles and co-regulatory values). The and field self-regulatory of privacy is spe in- lates information privacy in the Commonwealth public struments tend to coexist in different ways according to sector and the national private sector. It covers personal information and sensitive information (such as health stake in regional, national, international and transnational information, ethnicity, sexual preference, trade union economic,the specific social contexts and created political by spaces the normative (Pagallo et systems al. 2019; at membership). This situation is also evolving, partially Pagallo, Casanovas and Madelin 2019). fuelled by the palliative reaction against Covid-19. From 2019 the Federal Government has been engaged in several 3 practices of implementing the set of normative systems The recent Issues Paper (Australian Government 2020) Legal governance can be defined as the processes and ontrends the toreform update displays the definition a wide of set personal of issues, information. including i.e. the process of creating sustainable legal ecosystems. whether a statutory tort for serious invasions of privacy Fromput in this place standpoint, in specific legality contexts is thefor aresult variety of the of scenarios; coordina-

3 I.e. issuing several Recommendations for reform (Australian Government 2019). Amongtion others of (i)different updating thetypes definition of agency of ‘personal (artificial information’ and/or to capture human, technical data and other online identifiers (Recommendation 16(a)); (ii) strengthening existing notification requirements (Recommendation 16(b)); (iii) strengthening consent requirements and pro-consumer defaults (Recommendation 16(c)); (iv) and introducing a direct right of action to enforce privacy obligations under the Privacy Act (Recommendation 16(e)).

Law in Context, Vol 37, Issue 1, 2020 26 ISSN: 1839-4183 using socio-technical, cognitive-socio-technical systems no evidence so far of case-law nor out-of-court disputes or normative Multi-agent Systems). Some legal instru- regarding linked data resources. On the other hand, as ments—such as rules extracted from legal norms—can is well known, from 2000 onwards many data breaches be automated. Others, for instance, ethical values and have been reported—AOL, NETFLIX, Equifax, Cambridge principles or best practices regarding the monitoring of Analytica etc.5 — and the lawsuits that followed indicate regulatory systems, cannot be fully hardcoded, as they a certain alarm about a general state of uncontrolled require human intervention and decision-making (Koops surveillance (Norris et al. 2017, Zuboff 2019). Thus, the and Leenes, 2014) This is the case with trust, transpar- legal dimensions of security, intellectual property, pat- ency, privacy and data protection. Their implementation ents, licenses and, especially privacy and data protection requires building institutions to hold complex models of should be taken into account and applied to the use of legal governance. But this is not saying that they cannot legal resources and the building of new tools for the web be semi-automated. On the contrary, from this perspec- of data (Rodríguez-Doncel et al. 2016). tive, appropriate automation can facilitate human control Legal resources should be differentiated from legal and monitoring. sources. The former refers to the large number of exist- Accordingly, privacy and data protection can be con- ing legal vocabularies and documents on the Web of Data. sidered from the legal governance approach. This means that they can be implemented through the construction, legal’ at regional, national, international and transna- development and implementation of a technological tionalThe latter levels refers to be to effectively the specific implemented content that or ‘counts enforced. as toolkit, comprising data mining, data analytics and the Determining what is ‘valid’ law, what counts as legal, is linked open data tools of the later developments of the in itself a non-trivial theoretical operation that is usually Semantic Web—i.e. the Web of Data. The Knowledge performed through the concepts of doctrine, legal theory, Graphs approach that we embrace in this article is related and checks-and-balances (Sartor 2005, Peczenik 2011). to this dimension. For our purposes, it is worth noting that a taxonomy of a legal quadrant, or legal compass to classify and annotate a variety of sources (ii) that are 2.1 COMPONENTS OF LEGAL GOVERNANCE deemed necessary to produce the (Fig. ‘ecological 3) would validity’ suffice (i)of Since 2010, many large-scale RDF datasets have been a regulatory system—i.e. the condition of sustainability created. For instance, in 2017, Freebase 1 had 2.5 billion 4 of legal ecosystems for a cluster of stakeholders (Poblet, triples; DBpedia2 had more than 170 million triples. LOD Casanovas and Rodríguez-Doncel 2019). This legal compass (the Linked Open Data cloud) connects more than 3000 datasets, with more than 84 billion triples. The number of data sources doubles every three years (Zou and Özsu (i) reflects and endorses the two sides of the rule of law 2017). This has created a data space, strongly interconnected, in which law and government can have a leading role, as legal documents and, mostly, legal content and knowledge, are increasingly offered for public consumption (Casanovas et al. 2016). Linguistic resources for the legal annotated (Martín-Chozas et al. 2019, Rehm et al. 2020). Ontolex-lemon,domain are increasingly the vocabulary being identified, for lexical classified resources and in the Web of Data, have been extended to create databanks of legal terminologies that can be used automatically by dictionaries (Rodríguez-Doncel et al. 2015). However, new legal issues arise, such the use by LOD of crowdsourced vocabularies, where there is no authority imposing one interpretation over another. There is FIGURE 1. Legal compass for the rule of law. Source: Poblet, Casanovas and Rodríguez-Doncel (2019)

4 A triple is a set of three entities that codifies a statement about semantic data in the form of subject–predicate–object expressions. 5 See a short list of the most notorious cases at https://en.wikipedia.org/wiki/Data_breach

27 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

(binding power and social dialogue), (ii) and assumes as a pre-requisite that they can be partially coded through semantic representations. In the scheme of Fig. 1, institutional strengthening and trust are intended and eventually produced through a variety of sources that must be ordered beyond a determined threshold to build valid regulatory systems. Validity, and of compliance with several requirements—e.g. enforce- specifically legal ecological validity, emerges from the degree make the system acceptable and sustainable. Thus, validity ability, efficiency, effectiveness, fairness (justice)—that orqualifies norm). as This a second approach order is alsoproperty compatible that encompasses with recent FIGURE 2. Relationship between Anonymity, Unlinkability and surveysthe whole on regulatorybusiness languages system (notand compliance only a specific (Hashmi rule Disclosure. Source: Torra (2017, p. 9). Quoted with permission 2018) but shows that legal compliance requirements can be more complex than those set by regulatory languages. stance, linking the main concepts to be modelled (i) to

2.2 A PRELIMINARY FRAMEWORK 2019b) and (ii) to the existing data protection engineer- As already stated, there are many differences be- ingthe middle-outapproaches. approach defined by Pagallo et al. (2019a, tween the USA and EU approaches for regulating privacy. Whereas EU laws consider ‘privacy’ as a human right, organized their terminology for data minimization into a constitutional fundamental right, and a fundamental fourFollowing interrelated Pfitzmann categories, and Hansen depicted (2010), in Fig. Torra 3 as (2017) Venn EU right under the Article 7 and 8 of the EU Charter of Fundamental Rights (2000 and 2012) (González-Fuster original data minimization strategy. 2014, Blasi 2016),6 US “values it as a liberty over and diagrams. Henceforth, Torra extended and refined their against the state” (Blasi 2014). Thus, data protection of personal data should be differentiated from privacy tout Pfitzmann and Hansen (2010) originally provided the court. In contrast, as noticed by many scholars, the United following three definitions: (i) Anonymity of a subject States does not provide for an overall legal expectation of subjects, called ‘anonymity set’”; (ii) From an adversary privacy. The collection and processing of personal data is (intruder,means that attacker…) “the subject perspective, is not identifiable anonymity within of a subjecta set of regulated based on the type of data at stake. Thus, data means that “the adversary cannot achieve a certain level related to healthcare is subject to the Health Insurance Portability and Accountability Act (Kennedy–Kassebaum set”; (iii) Unlikability of two or more items of interest (IoI)of identification from an attacker’s for the perspectivesubject s within means the that anonymity “within is governed by the Financial Services Modernization Act the system (comprising these and possibly other items), (Gramm-Leach-BlileyAct, 1996) commonly known Act, 1999), as HIPAA, known and as financial GBLA.7 data

Legal requirements might be quite detailed, and differ- 8 ent according to national and jurisdictional frameworks. IoIs are related or not. Torra (ibid. 10) provides three the attacker cannot sufficiently distinguish whether those Related to privacy, to link the knowledge graph approach additional ones: (iv) Disclosure, that “takes place when to legal governance, we will endorse a broader conceptual attackers take advantage of the observation of available

6 The Charter of Fundamental Rights of the EU Union (2000-2012) states in Art. 7: “Everyonedata tohas improvethe right to respect their for knowledge his or her private on and some family confidentiallife, home and communications”. Art. 8.1 reads: “Everyone has the right to the protection of personal data concerning him or her”, and Art. 8.2: “Such data must be processed fairly for specified purposes and on the basis of the consent of the person concerned or some other legitimate basis laid down by law. Everyone has the right of access to data which has been collected concerning him or her, and the right to have it rectified.” 7 https://www.hipaajournal.com/comparison-of-european-and-american-privacy-law/ 8 It is worth noting that unlinkability is deemed a sufficient non-necessary condition, as it implies anonymity. However, Torra (2009, 9) points out that there might be a case in which linkability is possible, but anonymity is not.

Law in Context, Vol 37, Issue 1, 2020 28 ISSN: 1839-4183

TABLE 1: Fair Information Principles Practices. Source: Langheinrich (2001). There should be no secret record keeping. This includes both the publication of the existence of such 1. Openness and transparency collections, as well as their contents. 2. Individual participation The subject of a record should be able to see and correct the record. 3. Collection limitation Data collection should be proportional and not excessive compared to the purpose of the collection. 4. Data quality Data should be relevant to the purposes for which they are collected and should be kept up to date. 5. Use limitation

6. Reasonable security DataAdequate should security only be safeguards used for their should specific be put purpose in place, by according authorized to personnel.the sensitivity of the data collected. 7. Accountability Record keepers must be accountable for compliance with the other principles.

information about an IoI”; (v) Identity Disclosure, “when TABLE 2. Privacy Design Strategies. Source: from Hoepman the adversary can correctly link a respondent to a particular (2014) and ENISA (2015) record in the protected data set”; (vi) Attribute Disclosure “when the adversary can learn something new about an attribute of a respondent, even when no relationship can be established between the individual and the data”. These concepts stand on the shoulders of the Fair Information Practice Principles (FIPPs) proposed by US Secretary's Advisory Committee on Automated Personal Data Systems in a 1973 Report, Records, Computers and the Rights of Citizens. This Report was followed by the US Privacy Protection Study Commission Report on Personal Privacy in an Information Society (1977). Legal experts—Alan Westin (1967)—, AI and law experts—such as Layman E. Allen—, and computer scientists—Willis H. Ware (RAND Corporation)—were involved in their development. Table 1 summarises the so-called FIPPs. In 2004, Kim Cameron, Chief Identity Architect of Microsoft, wrote and blogged what he would call the “7 FIPPs. The result was the proposal of Privacy by Design Laws of the Internet”. His approach was to set a metasystem (PbD), the process for embedding privacy principles into identity layer, i.e. “to develop a formal understanding of the dynamics causing digital identity systems to succeed - or fail in various contexts, expressed as the Laws of Iden- itdesign happen specifications was—and still architectures. is—a model However, to be assembled computa or tional modelling, the specific engineering paths to make metasystem that can offer the Internet the identity layer to be built. Hoepman (2018) puts it in the following way: ittity. so Taken obviously together, requires. these9 laws define a unifying identity “Privacy by design is a system development philoso- In the following years, Ann Cavoukian (2006, 2007) phy that says that privacy should be taken into account aligned this Internet Identity Metasystem Layer with throughout the full system development lifecycle, from

9 “We need a unifying identity metasystem that can protect applications from the internal complexities of specific implementations and allow digital identity to become loosely coupled. This metasystem is in effect a system of systems that exposes a unified interface much like a device driver or network socket does. That allows one-offs to evolve towards standardized technologies that work within a metasystem framework without requiring the whole world to agree a priori.” Cameron (2004).

29 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 its inception, through implementation and deployment, formalisms to represent the relevant dimensions. Rather all the way until the system is decommissioned and no longer used. In software engineering terms this makes need to be addressed in any representation that is incor- privacy, like security or performance, a software quality thanporated a final in Knowledge solution, we Graphs show tothe support essential the aspects regulatory that attribute or non-functional requirement.” framework around Privacy and Transparency. Hoepman (2014) represented the existing engineering strategies (see also ENISA 2015) as design strate- 6. BACKGROUND CONCEPTS gies that describe a fundamental approach to achieve a certain design goal. Thus, a privacy design strategy could be understood as a "design strategy that achieves detailedBefore each detailing of them our in other framework, publications, we briefly as referenced present (some level of) privacy protection as its goal". The PbD inour each definition sub-section. for each of the basic concerns. We have analytical framework can be mainly set through eight strategies: minimize, hide, separate, aggregate, inform, 3.1 TRUST control, enforce and demonstrate (Table 2). The strate- The issue of trust has been prevalent in the Internet since its popularization in the early 90s—see the Grandi- are data-oriented strategies and (ii) the latter four are son and Sloman (2000) survey. Attention has focussed on process-orientedgies are assembled strategies. into two The classes: European (i) Unionthe first Agency four the lower layers of the Internet Architecture, emphasizing for Network and Information Security (ENISA) embraced this approach in 2015, as there was an increasing need to identity. More recently, with the advent of the Web and create a common framework for policy makers, legisla- socialauthentication, networks, which the cybersphere deals primarily and with society verification generally, of tors, service providers, data protection authorities and - standardization bodies. It is similar to the framework eventually assumed in GDPR provisions. Colesky, Hoep- thehave Internet. become Thereheavily are influenced many studies by information carried out (and in several misin man and Hillen (2016) correlated and map the strategies disciplinesformation) thatattempting flows in tonews characterize sites and social and understandnetworks in against privacy patterns, adding the notion of ‘tactics’ as the spread of information in the cybersphere, as well as an additional level of abstraction between them. how this affects society—see Marwick and Lewis (2017) - for an overview. The original vision for the Semantic Web included (2010)Nevertheless, minimization this strategyremains byan Torra open (2017).field. We There have are al a “Trust” layer, although its emphasis was more on au- manyready quotedother possibilities. the extension For of the instance, Pfitzmann Rastogi, and Hansen Gloria thentication and validation with static trust measures for and Hendler (2017) have developed a general method data. There have been many efforts in representing trust, to enhance privacy in the cloud, deploying mobile appli- including computational models.10 As proposed initially cations dynamically on a scalable on-demand hardware in Almendra and Schwabe (2006), and later in Laufer and software platform. From a legal governance perspec- and Schwabe (2017) and Schwabe, Laufer and Busson tive, Casanovas et al. (2014) have proposed an indirect (2019), the approach used here is based on the work of strategy for security platforms in which coding should Gerck (1997) and Castelfranchi and Falcone (2001), tak- be combined with the construction of ad hoc ‘anchoring’ ing the view that trust is “knowledge-based reliance on institutions to monitor and control the outcomes of the received information”, that is, an agent (i.e., a person or a software program) decides to trust (or not) based solely In the following sections, we illustrate how these information processing flow. on her/his knowledge, and the decision to trust implies considerations can be supported using Semantic Web

10 A general survey can be found in Pinyol and Sabater-Mir (2013); Artz and Gil (2017) presents an excellent earlier survey for the Semantic Web; and Sherchan, Nepal, and Paris (2013) surveys trust in social networks. In the Linked Data world, it is clear that facts in Semantic Web should be regarded as claims rather than hard facts (Bizer and Cyganiak 2009), which naturally raises the issue of trust on those claims.

Law in Context, Vol 37, Issue 1, 2020 30 ISSN: 1839-4183 the decision to rely on the truth of received or on already • known information to perform some action. Q3: How to resolve conflicts between applicable rules? In terms of a Knowledge Graph, an agent wishing to 3.3 TRANSPARENCY Generally speaking, according to Meijer (2013, 430) and items it deems “trusted”, i.e., it will use them to perform perform an action must first filter those information the intended action. Since it is not possible to “half-act”, “the availability of information about an actor that allows in this sense trust is binary—either the agent trusts the otherothers actors (Schudson to monitor 2015), the transparency workings or can performance be defined asof information, or it does not. A more extensive discussion can be found in Schwabe, Laufer and Busson (2019). to obtain valid and timely information about the activities Trusted information as the basis for supporting privacy ofthe government first actor.” Itor contemplates private organizations”. the “capacity12 Transparency of outsiders and transparency is discussed next. presupposes the involvement of an observed and an observer (Berstein 2017). In contrast to privacy, that is con- 3.2 PRIVACY cerned with information about individuals, transparency concerns any type of information, although it may make for privacy.11 a difference if the producer (author) is an organization intoAs four discussed broad types:in section the 2.2,nonintrusion, there are many seclusion, definitions limitation, and control Tavani theories. (2007) Forclassified our research, privacy we theories adopt Transparency is also related to privacy. In the so-called or an individual (Heimstäd and Dobusch 2018). the Restricted Access/Limited Control (RALC) Theory “Attention Economy” (GDPR, CE 2016), for example, the proposed by Moor and Tavani (2001). RALC presupposes information about users and consumers are a primary that an adequate theory of privacy needs to differentiate source of value, and companies actively seek to obtain as much information about users as possible. This can be in management of privacy. Accordingly, the RALC framework hasthe conceptthree components: of privacy anfrom account both theof the justification concept of andprivacy, the right to control how the information about them is used bydirect others. conflict In this with state privacy of affairs, rights transparency of users, who can have support the of the management of privacy. “RALC requires that one mustan account have ofprotection the justification from intrusion of privacy, and and interference an account the disclosure of information about the company’s pro- and information access, it addresses concerns not only cessesthe resolution and procedures of potential associated conflict withto the privacy individual policies: can about protecting informational privacy (as described in contribute to foster trust, showing that it is compliant the control and the limitation theories) but also about with these regulations and thus to stimulate authorization protection against the kinds of threats described in the on the part of the users. This has been subject to recent nonintrusion and the seclusion theories as well.” (Tavani regulations such as the General Data Protection Regula- 2007, 10). tion enacted by the European Union. From a more abstract approach, both Privacy and access to information related to an agent”. In order to Transparency relate to controlling actions over informa- ensureBased privacy, on this it theory, is necessary we define to answer Privacy three as “controlled questions: • Q1: What types of Actions are allowed (and controlled) be regarded as two points in the same control dimension. Privacytion, and tends who tocan limit define or restrict such controls. actions Asover such, information they can items, whereas Transparency tends to allow (in some • Q2: What are the relation types between some Agent over Knowledge Items (KIs)? cases, mandate) actions over them, which explains the and a KI that entitle this Agent to establish a Privacy natural tension that exists between the two.

Rule governing Actions over that KI? 11 See for example Smith, Dinev and Xu (2011). Paci, F., Squicciarini, A., Zannone (2018) and Such and Criado (2018) present surveys on multi-party privacy. 12 https://www.britannica.com/topic/transparency-government#accordion-article-history

31 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

7. AN ILLUSTRATION - DISASTER RELIEF DONATION between transparency and privacy rules. George’s relation The goal of this example is to illustrate the complex with ReliefOrg should be accessible in the KG, according and interdependent nature of trust, privacy and transparency; it will later be used to show the expressive power privacy rules would prevent this access. Since transpar- to not-for-profit legislation. On the other hand, George’s of the framework and how it incorporates the various ency legislation in this case has higher precedence than concerns at play. personal rules, this association can then be used by Ed’s Consider a scenario in which a disaster has occurred, rules, and therefore ReliefOrg would not be accepted as a recipient of donations by Ed. Consider now a slightly different scenario where George severalsuch as demands, the fire that Ed wantsburned to the donate roof some of the money Notre to Dame help ReliefOrg, and Ed wants to check the withCathedral, the disaster or the bushrelief firesactions. in Australia. Ed has received In response several to ReliefOrg is not an officer of donation requests from different organizations and needs report for ReliefOrg from the KG, but wants to make sure financial integrity of . He retrieves the financial to choose one of them to make the donation. However, given his past experience, Ed wants to make sure these Au- it has been audited, so he looks for a certification of the are legimate request as opposed to frequently occurring ditInc, which is unknown to him, so he needs to verify that financial report. He finds out it has been audited by scams. In other words, he wants to make sure that the donated money will actually be used for the relief actions, it has an accreditation certificate from a public authority. rather than being misused, e.g., funding the organization’s If such a certificate is available, Ed analyses the financial basic infrastructure, or employed in another action, or However, tipped by a friend, Ed learns that George may report finding nothing wrong in principle. in fact be one of the owners of AuditInc. He then checks organization. AuditInc to see if its owners are listed, and whether George even pocketed by unscrupulous officers of the receiving Ed formulates a rule saying that he only trusts organiza- is one of them. Given George’s privacy rule plus the fact that AuditInc - vacy rule would prevent access to the owner relationship, is not a not-for-profit organization, his pri validatedtions that by openly accredited publish audit who organizations. are their financial Furthermore, officers, so Ed would not know that George is one of the owners, becauseand their of financial personal reasons,records. EdFinancial does not records want to must contrib be- thus deciding to contribute. Note that in this scenario, George’s privacy rule would apply not only to AuditInc’s rule can be regarded as an application-related rule, akin information, but also to information furnished by others, toute what to an is organization referred to asin whicha “business George rule” is an in officer. traditional This for example, a photo in a social network where George software development. appears in the annual Christmas party of AuditInc with Let us assume that there is a law that stipulates that a caption mentioning his role as one of the partners, or perhaps with a badge on his neck identifying him as such. ReliefOrg, an NGO We next describe our proposed KG Usage framework dedicatednot-for-profit to raising organizations funds for must and helpingpublicly disaster identify relieftheir and subsequently analyse this example showing how it efforts.officers George, and that being George a very is an reserved officer person,of has a privacy can be represented by it. rule that stipulates that his association with any organization, including ReliefOrg, should not be made public. 8. A SUMMARY OF THE KG USAGE FRAMEWORK. Ed receives a request for donation from ReliefOrg and Figure 3 shows a diagram of the use of information within a KG. “Using a KG” is represented as a Request made by ReliefOrg are some Agent for an Action over a Knowledge Item (KI). We publishedneeds to decide in the whether KG.13 he should donate or not. The first assume the existence of an underlying RDF graph, which step Ed follows is to verify if the officers of Here we can see a potential conflict 13 See for instance https://permid.org or https://opencorporates.com as examples of KGs with this type of information.

Law in Context, Vol 37, Issue 1, 2020 32 ISSN: 1839-4183

and their roles, location information, date information, An RDF graph is simply a collection of statements, each depictions (photos, videos, …) and so on. would be equivalent to the “traditional” definition of KGs. The provenance graph of the nanopublication will contain provenance information about the assertions in the form of a triple such14 as The in the assertion graph (e.g.; what image or natural lan- . An RDF named graph guage processing software was used, recorded location is one for which there is an identifier to refer to it. over this underlying RDF graph as a way to structure info, whether the assertions were inferred using some actual KG is formed by defining several named graphs the RDF triples into Knowledge Items, similarly to Items, inference engine, etc.). The provenance graph can be used to represent, to the desired level of detail, the sup- 2014). Thus, a Knowledge Graph represents a collection porting information for the assertions. For example, if Statements and Qualifiers in Wikidata (Erxleben et al. of interlinked descriptions of KIs – real-world objects, an automated face recognition algorithm was used, the events, situations or abstract concepts – where: provenance information represented in the provenance • Descriptions have a formal structure that allows graph of the nanopublication may inform which algorithm both people and computers to process them in an was used, which parameters were used in this particular

• Entity descriptions refer to one another, forming a correctness of the extraction. Another use of provenance efficient and unambiguous manner; case, and a confidence factor of the algorithm about the network, where each entity represents part of the can be seen in the case of a statement stating that, for description of the entities related to it.15 provenance information may include documentation to We propose to represent the KG as a collection of supportexample, its such placeOfBirth as a reference . to a birth The Knowledge Items (KIs), each of which as a nanopublica- tion16 (Groth, Gibson and Velterop 2010). A nanopublication “offers a supplementary form of publishing alongside certificate that states that indeed the place of birth of traditional narrative publications”, consisting of three parts representable by RDF graphs: “(i) an assertion (a small, unambiguous unit of information), (ii) the provenance of that assertion (who made that assertion, where, when, etc.), (iii) the provenance of the nanopublication itself (who formed or extracted the assertion, when, and by what method).” (Golden and Shaw 2016)

5.1 KNOWLEDGE ITEM REPRESENTATION A KI, as all nanopublications, comprises an assertion graph, a provenance graph and a publication info graph. FIGURE 4. Request Evaluation Algorithm The assertion graph of a KI contains a set of assertions about its content. The assertions in this graph are Barack Obama is Hawaii. a subset of the assertions in the underlying RDF graph. The publication info graph will contain metadata about As an example, if a KI refers to an event, the assertion the creation of the KI itself (as opposed to the information graph would contain statements about the participants

14 https://www.w3.org/TR/rdf11-concepts/#section-rdf-graph 15 https://www.ontotext.com/knowledgehub/fundamentals/what-is-a-knowledge-graph/ 16 http://nanopub.org . “A scholar can promote small pieces of information within her work using the practice of nanopublication. Nanopublications include useful and usable representations of the provenance of structured assertions. These representations of provenance are useful because they allow consumers of the published data to make connections to other sources of information about the context of the production of that data.” (Golden and Shaw 2016)

33 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 contained in its assertions sub-graph), such as its author, creation date, etc…

5.2 CONTROLLING USAGE In our framework, Privacy and Transparency refer to Request for actions over some information (in a KI), for which an Authorization must be granted, according to the Rules set forth by the relevant stakeholders. Stake- holders include persons “related” to the KI, as well as institutional agents such as “the State” (whose rules are stated as laws). Rules may be based (make use of) on any information available in the KG. We refer to both Privacy and Transparency rules collectively as Usage Rules. Figure 4 introduces the basic algorithm to evaluate a the action is to determine which KIs in the KG are trusted by,Request. the Agent The requestingfirst step in the granting Action. an This Authorization is realized by for a Trust Engine that collects applicable rules and includes them in a Trusted sub-graph of the KG. We refer to the set of trusted KIs by Agent α as the Trusted KG(a). Once the Trusted KG has been determined, a set of usage rules are evaluated, which entails determining applicable FIGURE 5. Usage process for information in a KG rules (UsageRuleSet), evaluating them (EvalUsageRuleSet) - usage rule, it is also necessary to evaluate the trust rules formation items by Agent in the KG (stored in TGA), these collected rules to find out the set of trusted in ofand the resolving author of conflicts that usage if they rule, arise. which In order needs to to evaluate be based a and perform the requested Action on the KI, using the trusted information in TGA one for each Author ai of an applicable usage rule. The In order to apply a set of usage rules RS for a request on trusted KIs. Thus, a set of Trusted KG (ai) is defined, Action in the Request that was made. If the Authorization above), one must perform the following steps: result of this process defines the final Authorization for the for an Action on a KI by an Agent (as specified in step 2. is “Allowed”, the Action is carried out using the TrustedKG 4. Sort (reorder) the rules in RS in decreasing order of of the Agent that requested the Action. precedence (i.e., the rule with highest precedence Stated in words, the algorithm stipulates that, to evaluate a request for an Action by an Agent over a Knowledge 5. Assume there is a default authorization – i.e., if noth- comes first) Item (KI), the following steps should be performed (line ing is said, an Action is allowed (or denied) by default. numbers correspond to those in Figure 4). 6. Consider each rule R in the set RS in turn. For each one, 1. Collect the applicable rules referring to the KI and • Determine the authorization value by applying this store them in variable RS rule R for the Action on the KI requested by Agent. 2. Apply the collected rules in RS for the requested Action • If the authorization value can be determined, then by Agent on the KI. return it as the result of the algorithm. It is possible that the relevant rules are unable to determine an 3. If the Action was allowed, then collect the trust rules for Agent and store them in variable TS, then use information in the KG. authorization value, e.g., because of insufficient

Law in Context, Vol 37, Issue 1, 2020 34 ISSN: 1839-4183

7. If no value could be determined, return the default relations. Furthermore, the rules may (or may not) allow authorization as the result of the algorithm. the use of inferred relations in the KG. resolution strategy must be employed, which is in turn of which are sets of statements (Almendra and Schwabe subjectTo determine to Governance the final Rules. Authorization This algorithm value, abstracts a conflict the 2016).Rules The are antecedent of the form of antecedent privacy rules => consequent, may refer to both any essential decisions that must be made, to wit: statements in the KG, including 1. • Any statements in the subgraphs in the KI’s nanopublication (assertions, provenance and publication info); 2. Who can formulate a rule for a given KI? – in line 1; • The identity of the agent requesting permission; 3. How are conflicts between rules resolved? – in line 5-8 • The type of Action; We detail possible answers to these questions in the What are the allowed actions over KIs? – in line 4. following sub-sections. • Information in the KG serving as contextual information, such as • Date/time of the request; 5.3 RULES • Current location of agents involved; The UsageRuleSet(KI) function call determines what are the applicable rules given a KI. For trust rules, it is the rule is given in its assertion graph, using a notation determined by the author of the request. For privacy suchSeen as N3Logicas nanopublications, (Berners-Lee the et actual al. 2006) specification or SWRL. of17 rules this corresponds to answering the question “who Governance Rules are Rules that include other Rules (in either of antecedent and consequent), so in this sense they are meta-rules – i.e. rules about rules. thathas the it must right be to defineany agent a Privacy that is Rule somehow that controls related actions to the informationover this KI?” contained The definition in the KI. of AnyPrivacy useful itself instantiation indicates of the framework must spell out what are the accepted 5.4 CONFLICT RESOLUTION relation types, which can include: The Sort-by-Precedence (RS, decreasing) function call sorts the enabled rules in descending precedence • in the KI – for example, some person appearing in a order, typically combining several kinds of information postedAn identification photo or property video; for any agent that is included to establish order relations among rules. Some possible complementary order relations that • Any relation denoting referral to a person included can be employed are: in the KI – for example, some person cited in a post; • • Any creation or authorship relation; by a higher-ranked user take precedence over rules • Any agent related to the creation of the KI – for ex- Hierarchical relations between users – rules defined ample the author of a video posted by someone else; established by laws (authored by state agents) take • Any agent who has legal jurisdiction over an agent precedencedefined by lower-ranked over rules stated ones. by For individuals example, (com rules- mon citizens); • The Agent representing the legal system(s) that • Hierarchical relations between relation types. Rules identified or mentioned in the KI; has(have) jurisdiction over the KG, over the Agent or over the Action request. ranked relation type take precedence over rules defined by users related to the KI through a higher- The presence of such relations can directly occur in the KG (i.e., as a typed edge), or as a composition of valid type. For example, one may state that relation type defined by users related via a lower-ranked relation

“identifies” takes precedence over type “mentions”. 17 https://www.w3.org/Submission/SWRL/

35 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

person B is mentioned in a conversation, person A’s Donate subClassOf Action. Thus, in a video where a person A is identified, and CertificationAgency subClassOf org:Organization. rules would take precedence over person B’s rules. ReliefOrg AuditInc> type Since hierarchies are partial orders (which means that < > type NonProfit. < there may be items without an ordering between them), ReliefOrg AuditInc . < > officer . < > officer RuleEd1 they may not define precedence completely, so further . - canconflict be employed. resolution Most strategies require are user still involvement needed. Such at run and- Criado (2018) identified six categories of strategies that time, but the aggregation-based class of strategies can be {?O type org:Organization; hasOfficial: ?Ofc; hasFinan easily incorporated into an algorithm. Strategies in this cialRecord ?FR. ?Ofc type foaf:Person. log:includes ?FR assertions ?FRa. ?FRa log:semantics ?FRaS; ?FRAS class define an aggregation function such as consensus, majority, minimum fixed number of votes, permit-overrides, { ?FR auditedBy ?Aud. ?Aud type AuditCo}. results by a single aggregated result. - deny-overrides, etc, and replace the set of conflicting rule An alternative to the aggregation approach is to de- ?FR provenance ?PFr. ?PFr log:semantics ?PFrS. ?PFrS log:includes {prov:hasPrimarySouce ?DOCS}. ?DOCS as one is subject to only a single rule. This makes sense sertions ?DOCSa. ?DOCSa log:semantics ?DOCSaS. whencompose the the Action KI into to befiner-grained performed elements can be stated so that as each the ?DOCSaS log:includes {AuditCo certifiedBy ?CA. }. composition of the same operation performed on each author element independently. For the elements where the Ac- log:semantics ?TGEd. ?TGEd log:includes {?CA type tion is denied, the resulting composite object is altered. CertificationAgency}} For example, in a group photo showing several people, => { :add {?O type:NonProfit}. } some may allow publication other may deny publication. Stated in plain language, the KG contains statements In this case, it would be possible, to blurr out the face of the persons who have denied publication. are a kind of Organization, and Donate is an Action. It also affirming thatReliefOrg NonProfit, AuditCo and CertificationAgencyAuditInc 6. EXAMPLE SCENARIO REVISITED bothaffirms ReliefOrg that and AuditInc is a NonProfit. organization; In this section we re-examine the illustrative example is a Certification Agency and that George is an officer of scenario in light of the KG usage framework, showing how the most important aspects are represented. The RuleEd1 reflects Ed’s trust criteria – Ed’s Trusted Graph- will include a NonProfit organization only if its financial he only trusts organizations that openly publish who records include a list of its officers, and the financial re first rule captures Ed’s requirement about Non-Profits: This is a trust rule since it refers to KIs which are cord itself is audited by a certified Certification Agency. Financial records must be validated by accredited audit needed as input for Ed to allow taking the “Donate” action. organizations.are their financial officers, and their financial records. Ed’s second rule is in fact a business rule – he does not The trust rule below captures this, expressed using want to contribute to an organization in which George is N3Logic with extensions. We state under KG some statements we assume to be present in the KG: code using the KG, we express it here as a rule as well, to an officer. While this could be embedded in application KG facilitate the discussion. - RuleEd2 ClassOf org:Organization. NonProfit subClassOf org:Organization. AuditCo sub { author ; log:semantics ?TGEd. ?TGEd log:includes

Law in Context, Vol 37, Issue 1, 2020 36 ISSN: 1839-4183

{?O type:NonProfit, officer:}. { type Authorization; rule ; action thor }. ; value “Allowed”}}. - ruleEd1 assertions ?ARule1. ?Arule1 log:semantics provenance ?PTransp. ?PTransp ?Arule1S.?ARule1S log:includes log:semantics ?PTranspS. ?PTranspS log:includes { type Donate, recipient . intends Transp> prov:HasPrimarySource } Stated in words, RuleTransp (which assumes there is a } => { type Authorization, rule , action Stated in words, RuleEd2 stipulates that if Ed’s trusted - , value “Denied”}} statement affirming that NonProfitAct is a Law) says that graph contains an Organization for which George is an statesany operation that the toprovenance read who are graph the of officers RuleTransp of an organiza contains tion of type NonProfit will be allowed. Furthermore, it also whose recipient is this Organization, then the request a reference to the proper legal document, the text of the willofficer, be anddenied. Ed intends to take an action of type “Donate” Next, we look at George’s privacy rule. It is a privacy law () that is being interpreted by this rule. rule because it refers to an action over personal informa- and RuleTransp, there is a meta-rule stipulating that the In order to manage the conflict between RuleGeorge1 tion about George. latter has precedence over the former. This precedence relation is used in the Sort-by-Precedence function call RuleGeorge1 in line 5 in Fig. 4. -

{G :is {?O type:NonProfit; officer:. author ; log:semantics ?TGG. PersonalPrivacyRule subClassOf Rule. Legislation ?TGG log:includes {?O officer:. subClassOf Rule. author }. MetaRule1 ruleGeorge1 assertions ?AGeorge1. ?AGeorge1 log:semantics ?AGeorge1S. ?Ageorge1S log:includes { type Read; object ?G. ?A intends } { assertions ?AMR1. ?AMR1 log:semantics => { type Authorization; rule ; ?AMR1S. ?AMR1S log:includes action ; value “Denied”}} {{?R1 type PersonalPrivacyRule. ?R2 type Legislation} Stated in words, RuleGeoge1 says that if a read operation => {?R2 precedes ?R1}} is intended over the KG containing the information that Stated in words, this rule says that rules of type Legislation - have precendence over rules of type PersonalPrivacyRule. tion for that operation will be denied. - George is an officer of some organization, the authoriza The stipulation expressed in the transparency legislation PrivacyRule – it is any rule that uses a PersonalInformation We must also define more precisely what is a Personal KG for non-profits to divulge its officers can be expressed as property in its antecedent, which must also be defined.

RuleTransp NonProfitAct type Law. ?RuleP assertions ?RulePA.?RulePA log:semantics - ?RulePAS;?RulePAS log:includes {{?p1 ?r ?p2. { author . as (?p1 rdf:type Person OR ?p2 rdf:type Person)}} sertions ?ATransp. ?ATransp log:semantics ?ATranspS. => {?r rdf:type PersonalInformationRelation} Stated in words, this rule says that any relation involv- ?ATranspS log:includes{?G is {?O type Nonprofit; officer ?Ofr. ing a person is a PersonalInformationRelation. This is type Read; object ?G. ?A intends }

37 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

application that uses the KG, can be incorporated into the relations that characterize “personal information” should KG itself. The various nuances and interdependencies of beadmittedly further elaborated. a simplification, as in actual situations these these concerns were illustrated in a running example. One interesting point in the example is the fact that in spite of careful policies, loopholes in the regulations could allow undesired actions to take place. {?RulePP assertions ?RulePAS. ?RulePPAS log:semantics - As ongoing and future work, we are investigating ?RulePPASS; - ?RulePPAS log:includes{?RulePP antecedent ?RuleP able usage control over existing KGs, as well as exploring PAA. ?RulePPAA log:semantics ?RulePPAAS theimplementation applicability architectures to various domains. to allow Thisefficient architecture and scal ?RulePPAAS log:includes {{?p ?r ?q.(?r rdf:type PersonalInformationRelation)} yearscan also and be into fitted more into general wider Online legal governance Dispute Resolution models => {?RulePP type PersonalPrivacyRule}} Stated in words, this rule says that a Personal Privacy governancethat have been models developed (Ebner in and the Zeleznikow literature in 2016). the last While five Rule is a rule that involves (uses) a Personal Information the algorithm and the example presented here consider Relation in its antecedent. Notice that this is also a meta- that the authorization for an action would be determined rule, since its objects are rule themselves. in an automated fashion, it is entirely possible a version The second scenario does not require any additional where the algorithm could reach a state in which it is rules. It simply results in an “allowed” authorization, it could refer the information and their supporting facts not able to resolve conflicts between rules. At this stage, AuditInc>, since it is not of to a human being, to allow them to bring in additional because no conflict arises between RuleGeorge1 and criteria which would be outside the scope of the rules in ThisRuleTransp illustrates when a possible applied loophole to < in the legislation which type NonProfit, and therefore RuleTransp is not applicable. extension of the framework to accommodate this mode. the KG to reach a final decision. We are investigating the organization from being audited by another organization could be avoided if the NonProfit legislation prohibited an 8. REFERENCES 1. A. Hogan, E. Blomqvist, M. Cochez, C. d’Amato, G. de Melo, C. having common officers. 7. CONCLUSIONS Gutierrez, J. E. L. Gayo, S. Kirrane, S. Neumaier, A. Polleres, R. Navigli, A. N. Ngomo, S. M. Rashid, A. Rula, L. Schmelzeisen, We have presented a usage framework explicating J. F. Sequeda, S. Staab, and A. Zimmermann. “Knowledge Graphs”. arXiv. abs/2003.02320. 2020. capture privacy, transparency and trust concerns. The 2. Almendra, V. D. S., and Schwabe, D. 2006. Trust policies for frameworkthe various typesalso provides of specifications a better understandingthat must be made of the to semantic web repositories. In Proceedings of 2nd Interna- relations between these concerns. tional Semantic Web Policy Workshop (SWPW’06), at the 5th Trust entails determining which data items will be used International Semantic Web Conference, ISWC, pp. 17-31. to perform an action; privacy and transparency involve 3. Angella J., Kim, K. and Johnson, K.P. 2016. “Power of consum- controlling who can perform an action over a data item. related user-generated content on Facebook”, Computers in Trust is thus more fundamental, as privacy and transpar- Humaners using Behavior, social media: 58: 98-108. Examining http://dx.doi.org/10.1016/j. the influences of brand- ency rules must be based on trusted data. Furthermore, chb.2015.12.047 . privacy and transparency are approached as being differ- 4. Artz, D., Gil, Y. 2007. “A survey of trust in computer science ent points along the control dimension, thus explaining and the Semantic Web”, Web Semantics: Science, Services and the natural tension between the two. Agents on the World Wide Web, 5 82): 58-71. http://www. We have shown how legal requirements, and other types sciencedirect.com/science/article/pii/S1570826807000133 of norms, which ultimately regulate the functioning of any

Law in Context, Vol 37, Issue 1, 2020 38 ISSN: 1839-4183

5. Australian Government. 2019. Department of the Treasury. Deception in Virtual Societies. Dartmouth: Springer-Verlag Regulating in the digital age: Government Response and (2001). Implementation Roadmap for the Digital Platforms Inquiry 18. Castells, M. 2010. The Rise of the network society (2 ed.). (Government Response, December). https://treasury.gov. Cambridge, MA, USA: Blackwell Publ. Inc. au/publication/p2019-41708 19. Cavoukian, A. 2006. “7 Laws of Identity: The Case for Privacy- 6. Australian Government. 2020. Attorney-General’s Depart- Embedded Laws of Identity in the Digital Age”, Technology, ment. Privacy Act Review. Issues Paper. October, https:// Ontario Information and Privacy Commissioner, October, pp. 1-24. apo-nid309253.pdf 20. Cavoukian, A. 2010. “Privacy by Design. apo.org.au/sites/default/files/resource-files/2020-10/ The 7 Foundational 7. Berners-Lee, T., Connolly, D., Kagal, L., Scharf, Y., Hendler, J. Principles. Implementation and Mapping of Fair information 2008. “N3Logic: A Logical Framework for the World Wide Practices. Information and Privacy Commissioner, Ontario, Web”, Theory Pract. Log. Program. 8: 249–269. Canada. 8. Bernstein, E. 2017. “Making Transparency Transparent: The 21. Colesky, M., Hoepman, J.H. and Hillen, C. 2016, “A critical Evolution of Observation in Management Theory”, Academy analysis of privacy design strategies”. In 2016 IEEE Security of Management Annals 11 (1): 217 – 66. and Privacy Workshops (SPW), IEEE, pp. 33-40. 9. Bizer, C. and Cyganiak, R. 2009. “Quality-Driven Information 22. Council of the European Union, European Parliament, Filtering Using the WIQA Policy Framework”, Web Seman- 2016. Regulation (EU) 2016/679 of the European Parlia- tics: Science, Services and Agents on the World Wide Web. 7: ment and of the Council. General Data Protection Regula- 1-10. http://www.websemanticsjournal.org/index.php/ tion”, https://eur-lex.europa.eu/legal-content/EN/TXT/ ps/article/view/157/155 10. Blasi-Casagran, C. 2014. “The EU Role in Shaping Global 23. Davenport, T.H. and Beck, J.C. 2001. The attention economy: PDF/?uri=CELEX:32016R0679 . Data Privacy Standards”, PrivOn2015 https://sites.google. Understanding the new currency of business. Harvard Busi- com/site/privon2015/accepted-papers ness Press. 11. Blasi-Casagran, C. 2016. Global data protection in the field 24. Ebner, N. and Zeleznikow, J. 2016. “No sheriff in town: gov- of law enforcement: an EU perspective. London: Routledge. ernance for online dispute resolution”, Negotiation Journal, 12. Cameron. K. 2004. December- “The Seven laws of the Inter- 32(4): 297-323. net”. http://www.identityblog.com/stories/2004/12/09/ 25. thelaws.html Knowledge Graphs”, SEMANTiCS, CEUR 1695, http://ceur- 13. Carlson, A., Betteridge, J., Wang, R.C., Hruschka Jr., E.R., ws.org/Vol-1695/paper4.pdfEhrlinger, L. and Wöß, W. 2016. “Towards a Definition of Mitchell, T.M. 2010. “Coupled semi-supervised learning for 26. information extraction”. In: Proceedings of the Third ACM D. 2014. “Introducing Wikidata to the Linked Data Web”, International Conference on Web Search and Data Mining, Proc.Erxleben, ISWC F., Günther,2014, Part M., I.Krötzsch, LNCS, vol. M., 8796.Mendez, Cham: J., and Springer. Vrandečic, pp. 101–110. 27. European Union. 2012. Charter of Fundamental Rights of 14. Casanovas, P., Arraiza, J., Melero, F., González-Conejero, J., The European Union. Molcho, G. and Cuadros, M. 2014. “Fighting Organized Crime Journal of the European Union C 326/391. https://eur-lex. Through Open Source Intelligence: Regulatory Strategies of (2012/C 326/02) 26.10.2012 Official the CAPER Project”. In Hoekstra, R. ed., Legal Knowledge and 28. Danezis, G., Domingo-Ferrer, J., Hansen, M., Hoepman, J.-H, : Amsterdam: IOS Press, europa.eu/legal-content/EN/TXT/?uri=CELEX:12012P/TXT Information Systems JURIX 2014. Metayer, D.L., Tirtea, R. and Schiffner, S. 2014. Privacy and pp. 189-198. Data Protection by Design – From policy to engineering. 15. Casanovas, P., Jorge González-Conejero, J., de Koker, L. 2017. Technical report, ENISA, December. ISBN 978-92-9204- "Legal compliance by design (LCbD) and through design 108-3, DOI 10.2824/38623. https://www.enisa.europa. (LCtD): preliminary survey." TERECOM@JURIX 2017, eu/activities/identity-and-trust/library/deliverables/ 16. Casanovas, P., Palmirani, M., Peroni, S., Van Engers, T. and privacy-and-data-protection-by-design Vitali, F. 2016. “Semantic web for the legal domain: the next 29. Gandon, F. 2018. “A Survey of the First 20 Years of Research step”. Semantic Web, 7(3): 213-227. on Semantic Web and Linked Data”, Revue des Sciences et 17. Castelfranchi, C., Falcone, R. Social Trust: A Cognitive Ap- Technologies de l’Information - Série ISI : Ingénierie des proach. In: Castelfranchi, C., Yao-Hua Tan (Eds.) Trust and

39 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

Systèmes d’Information, Lavoisier, https://hal.inria.fr/ hal-01935898/document abs/1809.06532. https://arxiv.org/pdf/1809.06532.pdf Resource of Provenance-Centric Scientific Linked Data. CoRR, 30. Gerck, E. 1998. Toward Real-World Models of Trust: Reliance 43. Langheinrich, M. 2001. “Privacy by design—principles of on Received Information. Report MCWG-Jan22-1998, http:// privacy-aware ubiquitous systems. In G.D. Abowd, B. Brumitt, www.safevote.com/papers/trustdef.htm and S. Shafer (Eds.), International conference on Ubiquitous 31. Golden, P. and Shaw, R., 2016. “Nanopublication beyond the Computing, LNCS 2201, Berlin, Heidelberg: Springer, pp. sciences: the PeriodO period gazetteer”. PeerJ Computer Sci- 273-291. ence, 2, p.e44. https://peerj.com/articles/cs-44/ 44. Laufer, C, Schwabe, D. 2017.” On Modeling Political Systems 32. González-Fuster, G. 2014. The emergence of personal data to Support the Trust Process, Proceedings of the Privacy protection as a fundamental right of the EU, LGT Vol. 16. Online”, PrivON 2017, co-located with the 16th Interna- Cham: Springer Science & Business. tional Semantic Web Conference (ISWC 2017) Vienna, 33. Grandison, T., and Sloman, M. 2000. “A survey of trust in Austria, Out. 2017, p.1 – 16 http://ceur-ws.org/Vol-1951/ internet applications”, IEEE Communications Surveys & 3 (4): 2-16. 45. Lehmann, J., Isele, R., Jakob, M., Jentzsch, A., Kontokostas, Tutorials PrivOn2017_paper_7.pdf 34. Groth, P., Gibson, A., and Velterop. J. 2010. "The anatomy of a D., Mendes, P.N., Hellmann, S., Morsey, M., van Kleef, P., Auer, S., and Bizer, C.. 2013. “DBpedia-A large-scale, multilingual nanopublication." Information Services & Use 30 (1-2): 51-56. knowledge base extracted from Wikipedia”, Semantic Web 35. Hashmi, M., Casanovas, P. and de Koker, L. 2018. “Legal Com- Journal 6(2): 167–195. pliance Through Design: Preliminary Results of a Literature Survey”. TERECOM2018@JURIX, Technologies for Regulatory 46. Lenat, D.B. 1995. “CYC: a large-scale investment in knowl- Compliance http://ceur-ws.org/Vol-2309/06.pdf edge infrastructure”, Communications ACM 38(11): 33–38. 36. Hashmi, M., Governatori, G., Lam, H.P. and Wynn, M.T. 2018. 47. Martín-Chozas, P., Montiel-Ponsoda, E. and Rodríguez-Doncel, “Are we done with business process compliance: state of V. 2019. “Language resources as linked data for the legal domain” In, G. Peruginelli, S. Faro (Eds), the art and challenges ahead”, Knowledge and Information Knowledge of the Amsterdam: IOS Press, pp. 170-180. Systems, 57(1): 79-133. Law in the Big Data Age, 37. 48. Marwick, A. andLewis, R.. 2017. “Media Manipulation and Organizational Transparency as Multiactor Negotiation”, Disinformation Online.” Data & Society Research Institute, New York. https://datasociety.net/pubs/oh/DataAndSoci- PublicHeimstäd, Administration M., Dobusch, Review L. 2018. 78 (5): “Politics 727-738 of Disclosure: 38. Hoepman, J.-H. 2018. “Privacy Design Strategies” (The Lit- tle Blue Book). https://repository.ubn.ru.nl/bitstream/ 49. McCrae, J.P., Bosque-Gil, J., Gracia, J., Buitelaar, P. and Cimiano, P.ety_MediaManipulationAndDisinformationOnline.pdf 2017. “The Ontolex-Lemon model: development and applications”. In Proceedings of eLex 2017 Conference, pp. 19-21. 39. Hoepman, J.-H. 2014. "Privacy Design Strategies (extended handle/2066/195397/195397.pdf?sequence=1 50. Meijer, A. 2013. “Understanding the Complex Dynamics of abstract)", in N. Cuppens-Boulahia et al. (Eds.), ICT Systems Transparency”, 73 (3): 429 – 439. Security and Privacy Protection - 29th IFIP TC 11 International Public Administration Review Conference, SEC 2014, Marrakech, Morocco, June 2-4, 2014. 51. Miller, G. A. 1995. “WordNet: Smith, H.J., Dinev, T., Xu, H A Proceedings. Springer 2014 IFIP Advances in Information lexical database for English”, Communications ACM 38 (11): and Communication Technology, pp. 446-459. 39–41. 40. Kejriwal, M. 2019. Domain-Specific Knowledge Graph Con- 52. Norris, C., De Hert, P., L'Hoiry, X., Galetta, A. (Eds.) 2017. The struction. Cham: Brief Springer International Publishing Unaccountable State of Surveillance. Cham: Springer. 41. Koops, B.J. and Leenes, R. 2014. “Privacy regulation can- 53. Paci, F., Squicciarini, A., Zannone, N. 2018. “Survey on Access not be hardcoded. A critical comment on the ‘privacy by Control for Community-Centered Collaborative Systems”. design’provision in data-protection law”. International ACM Comput. Surv. 51 (6): 1–6. Review of Law, Computers & Technology, 28(2): 159-171. 54. Pagallo, U., Aurucci, P., Casanovas, P., Chatila, R., Chazerand, 42. Kuhn, T., Meroño-Peñuela, A., Malic, A., Poelen, J.H., Hurlbert, P., Dignum, V., Luetge, C., Madelin, R., Schafer, B. and Valcke, A.H., Centeno, E., Furlong, L.I., Queralt-Rosinach, N., Chichester, P., 2019. AI4People-On Good AI Governance: 14 Priority Ac- C., Banda, J.M., Willighagen, E.L., Ehrhart, F., Evelo, C.T., Malas, tions, a SMART Model of Governance, and a Regulatory Tool- T.B., & Dumontier, M. 2018. Nanopublications: A Growing box. https://www.eismd.eu/pdf/AI4PEOPLE%20On%20 Good%20Ai%20Governance%202019.pdf

Law in Context, Vol 37, Issue 1, 2020 40 ISSN: 1839-4183

55. Pagallo, U., Casanovas, P. and Madelin, R. 2019. “The middle- 68. Schwabe, D., Laufer, C., and Busson, A. 2019. “Building out approach: assessing models of legal governance in data Knowledge Graphs About Political Agents in the Age of The Misinformation, http://arxiv.org/abs/1901.11408 Theory and Practice of Legislation, 7 (1): 1-25. https://www. 69. Sherchan, W. Nepal, S., and Paris, C. 2013. “A Survey of trust in protection, artificial intelligence, and the Web of Data”. tandfonline.com/doi/full/10.1080/20508840.2019.1664543 social networks”, ACM Comput. Surv. 45, 4, Article 47 , August. 56. Peczenik, A. 2004. Scientia Juris. In A Treatise of Legal Philoso- 70. Smith, H.J., Dinev, T., Xu, H. 2011. “Information Privacy phy and General Jurisprudence, Vol. 4. Dartmouth: Springer. Research: An Interdisciplinary Review”, MIS Quarterly 35: 57. - 989–1015. ing about privacy by data minimization: Anonymity, unlink- 71. Speer, R. and Havasi, C. 2012. "Representing General Re- Pfitzmann, A. and Hansen, M. 2010. “A terminology for talk ability, undetectability, unobservability, pseudonymity, and lational Knowledge” In ConceptNet 5, Proceedings of the identity management”. V.034. http://www.maroki.de/pub/ Eight International Conference on Language Resources and Evaluation (LREC'12), ELRA, pp. 3679-3686. 58. Pinyol, I.; Sabater-Mir, J.; Computational trust and reputation 72. Such, J.M. and Criado, N. 2018. “Multiparty Privacy in Social dphistory/2010_Anon_Terminology_v0.34.pdf Media”, Communications ACM. 61: 74–81. Intelligence Review 40:1–25 DOI 10.1007/s10462-011- 73. Suchanek, F.M., Kasneci, G., and Weikum, G.. 2007. “YAGO: a models for open multi-agent systems: a review. Artificial 9277-z (2013) core of semantic knowledge unifying WordNet and Wikipe- 59. Poblet, M., Casanovas, P. and Rodríguez-Doncel, V. 2019. dia”. In 16th International Conference on World Wide Web, Linked Democracy: Foundations, tools, and applications. pp. 697–706. Springer Law Brief n. 750. Cham: Springer Nature. 74. Tavani, H.T. and Moor, J.H. 2001. “Privacy Protection, Control 60. Rastogi, N., Gloria, M.J.K. and Hendler, J. 2017. “Security and of Information, and Privacy-enhancing Technologies”, SIGCAS Privacy of performing Data Analytics in the cloud-A three- Comput. Soc. 31: 6–11. way handshake of Technology, Policy, and Management”. 75. Tavani, H.T. 2007. “Philosophical Theories of Privacy: Im- arXiv preprint arXiv:1701.06828. plications for an Adequate Online Privacy Policy”, Metaphi- 61. Rehm, G., Galanis, D., Labropoulou, P., Piperidis, S., Welß, M., losophy 38: 1–22. Usbeck, R., Köhler, J., Deligiannis, M., Gkirtzou, K., Fischer, J. 76. Torra, V. 2017. Data Privacy: Foundations, New Developments and Chiarcos, C., 2020. “Towards an Interoperable Ecosystem and the Big Data Challenge. Cham: Springer Nature. of AI and LT Platforms: A Roadmap for the Implementa- 77. United Nations, Universal Bill of Human Rights, Resolution tion of Different Levels of Interoperability”. arXiv preprint A/RES/217(III) http://unbisnet.un.org:8080/ipac20/ipac. arXiv:2004.08355. 62. power is transforming energy, the economy, and the world. jsp?session=14O243550E15G.60956&profile=voting&uri= 78. US Privacy Protection Study Commission. 1977. Personal NewRifkin, York: J. (2011). Palgrave The Macmillan.third industrial revolution: How lateral full=3100023~!909326~!67 Privacy in an Information Society. https://epic.org/privacy/ 63. Rodriguez-Doncel, V., Santos, C., Casanovas, P. and Gomez- ppsc1977report/ Perez, A. (2016). “Legal aspects of linked data–The European 79. US Secretary's Advisory Committee on Automated framework”. Computer Law & Security Review, 32 (6): 799-813. Personal Data Systems. 1973. Records, Computers and 64. Rodríguez-Doncel, V., Santos, C., Casanovas, P. and Gómez- the Rights of Citizens. https://aspe.hhs.gov/report/ Pérez, A. (2015) “A Linked Term Bank of Copyright-Related records-computers-and-rights-citizens Terms”. In JURIX-2015, Amsterdam: IOS Press, pp. 91-100. 80. - 65. Sartor, G. 2005. In A Legal reasoning. Treatise of Legal Philoso- tive knowledge base”, Communications ACM 57 (10): 78–85. phy and General Jurisprudence. Vol. 5. Dartmouth: Springer. Vrandečić, D., Krötzsch, M. 2014. “Wikidata: a free collabora 81. Westin, A. 1967. Privacy and Freedom. New York: Atheneum. 66. Schudson, M. 2015. The Rise of the Right to Know: Politics 82. Zou, L., and Özsu, M.T. 2017. "Graph-based RDF data manage- and the Culture of Transparency, 1945-1973. Cambridge, ment", 2 (1): 56-70. MA: Harvard University Press. Data Science and Engineering 83. Zuboff, S., 2019. The Age of Surveillance Capitalism: The Fight 67. Schwab, K., 2017. The fourth industrial revolution. Crown Barack Business. for a Human Future at the New Frontier of Power.

Obama's Books of 2019. NY: Profile Books.

41 Law in Context, Vol 37, Issue 1, 2020 Received: September 2, Date of acceptance: October 16, 2020, Date of publication: November 9 2020, DOI: https://doi.org/10.26826/law-in-context.v37i1.119

Electronic Australian Elections: Verifiability of Accuracy is a Design Goal, which Must be Mandated by Law and Deliberately Designed into Electronic Electoral Processes

By Vanessa Teague1, Orcid: https://orcid.org/0000-0003-2648-2565 and Patrick Keyzer2, Research Professor of Law and Public Policy at La Trobe University, Orcid: https://orcid.org/0000-0003-0807-8366 1 Thinking Cybersecurity Pty Ltd and the Australian National University 2 La Trobe University

ABSTRACT Electronic voting and counting are increasingly common and have been adopted in a number of Australian jurisdictions. Unfortunately, there is evidence that e-voting systems lack transparency. At present there are reasonable solutions for poll-site e-voting but none for remote paperless Internet voting. Although there are reasonable methods for statistical audits of electronically counted election results, Australian elections do not use them. The authors argue that a purposive approach should be taken to relevant electoral laws to ensure that genuine scrutiny of electronic electoral processes can be undertaken. This would require the source code and the voting data to be made available

reforms need to be undertaken to ensure that Australian elections are accurate, and consistent with the constitutional requirementfor testing. The of direct authors choice recommend by electors. a number of legislative reforms to ensure the verifiability of e-voting. These

Keywords – Elections, electronic voting, scrutiny, security, verifiability

Acknowledgments – We thank Associate Professor Jeffrey Barnes and Mr Joel Butler for their helpful comments on previous drafts. Disclosure statement – Vanessa Teague has engaged in several consulting projects with the Swiss Federal Chancellery, to examine their e-voting system and advise on their regulations. She is also on the board of advisors of Verified Voting. No potential conflict of interest was reported by Patrick Keyzer. License – This work is under Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) https://creativecommons.org/licenses/by-nc-sa/4.0/ Suggested citation: designed into electronic electoral processes” 37 (1): 42-65. DOI: https://doi.org/10.26826/law-in-context.v37i1.119 Teague, V. and Keyzer, P., Law“Engineering in Context, Australian Elections: Verifiability is a design goal, which must be mandated by law and deliberately Summary 1. Introduction: the unchanging requirements of electoral processes 5.2 Risk limiting audit 2. The many roles of computers in Australian elections 5.3 How could we verify the Senate scanning process? 2.1 Many experiments and no consensus 5.4 Who should verify the Senate scanning process? 2.2 Verifiability is a design goal that should be legislated 6. Is the iVote Protocol Verifiable? 2.3 The structure of this paper 7. The Secret Electronic Ballot 3. Scrutinising the software and the processes 8. Could electronically-dependent election results be challenged? 3.1 Switzerland: how good regulations can expose serious problems 9. Electoral Commission analysis and guidelines 3.2 Inadequate laws left the same problems hidden in NSW 9.1 The ECANZ Essential Guidelines 3.3 Recommendation: the secrecy laws in NSW should be replaced with transparency laws 9.2 The Wilkins Report on iVote 3.4 The ACT's EVACS System 10. Recommendations 4. Should the source code for the Senate count be open? 11. Conclusions 5. Genuine auditing, monitoring and scrutiny of the data 12. References 5.1 “Software independence”

42 Law in Context, Vol 37, Issue 1; 2020 ISSN: 1839-4183

6. INTRODUCTION: THE UNCHANGING Ballot Secrecy REQUIREMENTS OF ELECTORAL PROCESSES Nobody, including administrators, should be able to Australians have been engineering better electoral link a voter to their vote. This protects voters from processes for over one hundred and sixty years. In 1856, vote-buying and coercion. the colony of Victoria, newly separated from New South Transparency commenced on the 23rd of September that year (Brent It should be possible for observers to verify that the 2006;Wales, Chapman was defining 1967). the rules for its first election, which election is conducted properly through scrutiny, re- William Nicholson, a Member of the Legislative Coun- moving opportunities for undetectable error or fraud. cil, moved in late 1855 that the elections be conducted These goals are as critical now as they were in 1855. by secret ballot. It would prevent voter intimidation, he argued, particularly by the government, which was a very large employer; and it would stop the practice of “treat- challenges to the transparency of elections in Australia, ing” and hence make elections more orderly (Brett 2019; particularlyThe first aim relating of this toarticle electronic is to tease electoral out contemporary processes. see also Hasen 2000). Australian State and Territory and also Commonwealth Henry Chapman, another member of the Legislative electoral laws all contemplate scrutiny. (See for example Council then designed the new system: voters had their Commonwealth Electoral Act 1918, Part XVIII; Electoral names marked off on an electoral roll at a polling place, Act 1992 (ACT), ss 122-3; Electoral Act 2017 (NSW), Di- were presented with a printed ballot paper, and retired vision 7 and s 158 (analysed below); Electoral Act 2004 to separated stalls to mark their ballot paper, in secret, (NT), ss 46-7, 128 and Div 5 subdiv 2; Electoral Act 1992 before depositing their ballot paper in a locked box under (Qld), s 104; Electoral Act 1985 (SA), s 67 and Part 10 the watchful eye of a poll clerk (Sawer 2010). By the time (note Div 3A, which addressed “Computer vote counting the Victorian election took place, Tasmania and South in Legislative Council elections”; Electoral Act 2004 (Tas), Australia had followed suit (Brent 2006; Newman 2003). Parts 5 and 11, and s 172; Electoral Act 2002 (Vic), ss 76, The use of voting stalls 110, 111, 114, 116 and 119; Electoral Act 1907 (WA), ss 92, 99, 117, 134, 137, 144-6). But in general these laws was just as crucial to the success of the scheme as the have been written having regard to the technology of the ballot paper. Together, these practical measures made time, i.e., when standing in a room and watching what was secret voting a workable reality (Brent 2006).

These electoral processes were deliberately designed properly. Increasingly, Australian electoral processes are electronic.going on was In sufficient this article to ensure we demonstrate the process that was electoral running and requirements were written into law. law and practice has not evolved to achieve the crucial with specific security goals in mind, and these designs Different countries devised different solutions to objective of meaningful scrutiny of electronic electoral achieve the same goals. For example, French voters put processes. If anything, we have gone backwards. their (candidate-coloured) ballot paper into an enve- The second aim of this essay is to describe the most lope before depositing it in a transparent urn (Balinski important reforms that are required to ensure that our and Laraki 2010). Some jurisdictions, particularly some electoral law is consistent with the principles of the se- parts of the United States, have never adopted secure cret ballot and meaningful candidate-appointed scrutiny, and transparent electoral processes, and have suffered when a large part of the electoral process is electronic. decades of troubled elections as a result (Gumbel 2005; Drawing upon successful examples of good regulation in Jones and Simons 2012). other democracies, we will explain the highest reform There are two main election security requirements: priorities for Australia’s electoral law. We demonstrate below that these reforms are urgently needed when recent

Law in Context, Vol 37, Issue 1, 2020 43 ISSN: 1839-4183 developments are taken into account – the Australian 3.4). The Victorian Electoral Commission publishes the Electoral Commission recently argued that even the ex- source code, but not the voting data, citing vote privacy isting, very basic, scrutineering provisions did not apply to those parts of the electronic Senate counting process code for errors, but not to double-check the count itself. they had outsourced. Finally, we provide a plan for what concerns. This makes it possible to examine the official changes need to be made to achieve ballot secrecy, trans- requests to release the code, citing security concerns and parency and accuracy. alsoThe AECthe protectionpublishes the of informationvote data but that has is refused considered specific to

7. THE MANY ROLES OF COMPUTERS IN New South Wales Electoral Commission (NSWEC) is be “commercial-in-confidence” (see Section 3.3). The AUSTRALIAN ELECTIONS similar. West Australians can observe neither code nor data. Although there are academic projects that provide a general framework incorporating various options for 2.1 MANY EXPERIMENTS AND NO CONSENSUS STV counting, (Ghale et al 2018) neither a shared software Australia has nine electoral systems; one for each of the Commonwealth, the States and the two self-governing (and therefore transparency) seem likely. project nor a unified national approach to verifiability territories. (A tenth electoral jurisdiction, Norfolk Island, Voting by computer varies even more widely. Western operated as a self-governing Territory between 1979 and Australia and South Australia have conducted small ex- 2015. See the Norfolk Island Act 1979 (Cth) (repealed), and periments allowing computer-assisted completion of a the Norfolk Island Legislation Amendment Act 2015, and, paper ballot in a polling place for voters with disabilities. further, O’Collins 2002). The nine polities have conducted The Australian Capital Territory has had electronic voting interesting experiments in elections from time-to-time. in a polling place (in which the actual vote is electronic) They each have subtly different, but also broadly similar for many years, and both New South Wales and Victoria laws and practices. That said, there is very little logic or consistency in the use of computers across Australia’s electoral commissions. forhave use provided in a controlled this also. environment Victoria briefly (such experimentedas an embassy For example, consider the electronic counting of the Single orwith consulate) an end-to-end by voters verifiable living abroad. electronic The NSWvoting Electoral system Transferable Vote (STV) (Tudeman 1995). For this very Commission uses its iVote Internet voting system, which in complex vote-counting function, counting by computer the 2015 and 2019 elections received more than 200,000 has numerous obvious advantages. Unsurprisingly, many votes. They have tried to market it to other commissions, electoral commissions, including the Australian Electoral but so far only WA has adopted it, and on a much more Commission (AEC) (for counting the Senate), most of the limited scale. States (for counting their Legislative Councils) and the So we have a number of experiments, but no consen- Australian Capital Territory (for its unicameral Legislative sus. Neither the requirements (for privacy, transparency, Assembly) implement some form of electronic counting security, etc) nor even the basic question of what con- of STVs. stitutes an acceptable level of risk, are broadly agreed Publication of the source code and the voting data across Australia. is the minimum necessary for enabling public error- Except possibly electronically-assisted completion checking of processes and results of electronic electoral of a paper ballot, none of these systems truly replicate processes. The Australian Capital Territory Electoral the scrutiny opportunities of a traditional polling place. Commission publishes both the source code and the vot- Even if both the source code and vote data are available, scrutineers still need evidence that the electronic votes to rerun the count themselves (at least it did until this ing data, so interested parties can use the official code year – see Elections ACT “Electronic voting and counting”, for numerous reasons, even if the electoral commission accessed 20 August 2020, discussed further in Section accurately reflect the voters' intentions. This may fail

44 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 has published a record of the votes and advertised code that seems to be correct: there could be malware on the does not come automatically – naively designed electronic machines that saves a vote different from the one that the systemsIt is very may important be subject toto understandfraud that is thatundetectable verifiability by human observers. Just as paper-based processes need to that swaps two candidates' positions on the touchscreen, be designed with observability in mind—we do not let our orvoter there asked could for, be or athere problem could with be a configurationthe communication problem of electoral commissions count paper votes in secret, away results that causes some of them to be dropped in transit, from the eyes of scrutineers—electronic ones need to be or many other similar problems. There is not necessarily designed to generate evidence that they have correctly any way for scrutineers to detect any of these problems, handled the votes. The slogan must be ‘verify the election, even if they are standing at the polling place. This applies not the software.’ Successful examples, and important both to electronic counting systems, where we need to failures, are described in Section 4. check that the electronic vote data accurately digitizes a paper ballot, and to electronic voting systems, where The main idea we want to convey in this essay is this: we somehow need to solve the much harder problem of Verifiability is a design goal, which must be mandated by verifying that an electronically-captured vote accurately law and deliberately designed into electronic electoral processes. We strongly support both the open availability of the reflects the voter's intention. election outcome, is the subject of the next section. software and Verifying the entire data flow, from voter intention to in theory you could have one without the other, they the verifiability of the process. Although 2.2 VERIFIABILITY IS A DESIGN GOAL THAT SHOULD BE are complementary and both contribute substantially LEGISLATED to evidence that the system functions appropriately and The purpose of scrutiny in an election is to allow the achieves what we want elections to achieve: demon- representatives of candidates (scrutineers) to verify that strated accuracy. For one thing, it seems hard in practice the election has been conducted fairly and properly. without revealing details of how it works. There are two main ways to ensure proper scrutiny in to convince anyone that the verification process is sound

2.3 THE STRUCTURE OF THIS PAPER relatedan electronic documentation, process. The to understandfirst, described how in the Section system 3, is In Section 3.1, the next section of this article, we will intendedto examine to the work. software, This should specifically specify, the in source precise code detail, and - exactly what computations are being performed and what sPost Internet voting system. This provides a case study protections are in place. This is a good way of checking for thatexplain we usethe verifiabilityas a reference failures point. discoveredThen, in Section in the 3.2, Swis we accidental errors or security problems (and many have will consider the implications of identical errors in the been found, as we shall demonstrate below). NSW iVote system. The question of the availability of the That said, examination of the source code and related counting code in Australian Senate elections is considered documentation does not really prove whether the elec- in Section 3.3. In Section 3.4 we will critically evaluate the tion was properly conducted on the day. For one thing, a ACT's EVACS system. corrupt insider or external attacker could simply cause For each of these examples, we’ll describe what can be different software to run on the real election system on learned by examining the software, then discuss in Sec- election day. This insight has motivated many years of tion 4 how (or whether) the system provides adequate technical research into the question of how one could evidence to verify its outcomes and, where possible, verify an election result without trusting the electronic explain how it could be reformed to do so. In particular, system on which the election was conducted. through an analysis of the relevant provisions of the NSW Electoral Act, we will demonstrate that provisions

Law in Context, Vol 37, Issue 1, 2020 45 ISSN: 1839-4183 regulating audits, monitoring and scrutiny of electronic über die elektronische Stimmabgabe (VEleS) [Federal electoral processes cannot produce the required transpar- Chancellery Ordinance on Electronic Voting (VEleS)] (Switzerland) 13 December 2013 SR 161.116 art 7b). electronic voting systems have not been compromised, However, this requirement applies only to software thenency. this We raisesargue not that only if we cryptographic cannot feel concerns,confident butthat quite our - possibly constitutional concerns. byseeking software certification that is not for openly use by available up to 100% for public of voters. scrutiny. Cer tification for use by up to 50% of voters can be achieved what changes need to be made to electronic voting in Aus- In early 2019 with a team of other researchers, one of traliaIn tothe achieve final section ballot secrecy, of the paper transparency we outline and a accuracy. plan for us examined the source code made available to comply with that regulation (Lewis, Pereira and Teague 2020). 8. SCRUTINISING THE SOFTWARE AND PROCESSES Three serious cryptographic errors were found. It was This section examines what can be achieved by ex- - exploited to forge a cryptographic ‘proof’ that the elec- independently demonstrated that the first two could be mentation available for genuinely independent and open amining the source code, specifications and other docu when actually other votes had been substituted (Haines tronic votes had been correctly shuffled and decrypted, to understand more accurately what a system’s security et al 2020). Crucially, this proved that the system did not review. These activities make it possible to find errors and for use by up to 100% of voters. meet the verifiability criterion necessary for certification properties truly are. It is not, however, sufficient for the For universal verification, auditors receive proof that verification of the result (more on this later). 3.1. SWITZERLAND: HOW GOOD REGULATIONS CAN the result has been ascertained correctly. They must EXPOSE SERIOUS PROBLEMS evaluate the proof in an observable procedure. To do Switzerland has one of the oldest Internet voting this, they must use technical aids that are independent of projects in the world (Serdult et al 2014; Lust 2018). and isolated from the rest of the system. The proof must Although elections are administered by cantons, the Fed- confirm that the result ascertained: (a) takes account of all votes cast in conformity with the for Internet voting. Their regulations are strict, detailed, system that were registered by the trustworthy part eral Chancellery administers standards and certification of the system; (Barrat I Steve, Goldsmith and Turner, 2012). Openness (b) takes account only of votes cast in conformity with ofand the emphasise source code privacy, is required. verifiability Modalities and fortransparency publishing the system; the source code include: (c) takes account of all partial votes in accordance with 1. The source code must be prepared and documented the proof generated in the course of the individual according to the best practices. verification (Verordnung der BK über die elektronische Stimmabgabe (VEleS) [Federal Chancellery Ordi- 2. It must be easily obtainable, free of charge, on the internet. nance on Electronic Voting (VEleS)] (Switzerland) 13 December 2013 SR 161.116 art 9, Annex, art 5.4). 3. The documentation on the system and its operation must explain the relevance of the individual compo- Swiss, because the whole purpose of open and public nents of the source code for the security of electronic scrutinyThese had first been two discoveries to identify didand not correct overly errors concern of this the voting. The documentation must be published along kind. However, the third problem affected a property with the source code. 4. Anyone is entitled to examine, modify, compile and For the purpose of individual verification, voters must execute the source code for ideational purposes, and to called ‘individual verifiability.’ receive proof that the server system has registered the write and publish studies thereon (Verordnung der BK vote as it was entered by the voter on the user platform

46 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

as being in conformity with the system. Proof of correct difference was that Switzerland was opening a system to registration must be provided for each partial vote (ibid., scrutiny that they were considering using six months in Art 4.2). Unlike the earlier errors, this problem also the future; NSW was already running it for early voting, affected an earlier system already in use and already had already received votes, and was intending to decrypt certified for use by up to 50% of voters. It allowed mal- on election day, March 23rd, 2019 (see NSW Electoral ware on the voter’s computer to produce a ‘proof’ that Commission, Report on the Conduct of the 2019 NSW State allowed for apparently-valid vote verification for the voter, Election). In its report, the NSW Electoral Commission while actually submitting a nonsense vote that would not be counted. This was a far more serious matter for unlike the Swiss Post system, the machine on which the briefly remarked: the Swiss, because it indicated that an already-certified iVote mixnet runs was not physically connected to any other system did not meet its certification standards, which in computer systems either within or outside the Electoral turn demonstrated that their non-public certification Commission. The mixnet issue was assessed and rectified process had failed. before the relevant code was used for the 2019 NSW State Internet voting in Switzerland has been on hold since election. these discoveries, while the system designers attempt to The phrase “not physically connected” is a peculiar repair their cryptographic protocol and the Chancellery one. PWC’s audit report, which is redacted, says “**** on tightens their regulations further. There has been considerable public discussion and debate about these errors 2019). While one can only speculate on the identity of a and their implications for the future of Internet voting in wordair-gapped that has (offline) been redacted, computers it iswas possible not disabled” that the (PWCword Switzerland (Federal Council of the Swiss Confederation, was “WiFi”. This conclusion is reinforced by the use of 2019). The main point is that well-written regulations the word “was” in the sentence. If that is so, then it is that emphasise transparency brought serious technical conceivable that the data could have been corrupted by an external party. (Internet connectivity is not necessary security properties allowed researchers to demonstrate for insiders to manipulate the votes.) We may never know. problems to light. The clear definitions of the required unequivocally that the system did not meet those require- How did such a fundamental problem, so obvious to ments. If the earlier system had been made available for an expert observer that two other research groups in- completely open public scrutiny earlier (rather than being dependently discovered it in Switzerland, escape notice more likely that its errors would have been detected earlier. the product of a closed certification process) it is much helps create this problem. While the 2017 Electoral Act containscompletely provisions in NSW? that We appearbelieve tothat provide NSW electoralfor auditing, law 3.2 INADEQUATE LAWS LEFT THE SAME PROBLEMS monitoring and scrutiny of electronic electoral processes, HIDDEN IN NEW SOUTH WALES they ultimately fail to provide the proper preconditions Election software is a global business. The Swiss e- for effective auditing, monitoring and scrutiny. voting system was provided by a multinational corpora- We can start with s 156 of the Electoral Act 2017 (NSW), tion, Scytl, which also provides the NSW iVote Internet which states: voting system. Though the electoral systems and user 156 INDEPENDENT AUDITING OF TECHNOLOGY AS- experiences are quite different between countries, these SISTED VOTING (1) The Electoral Commissioner is to engage an independent person (the "independent auditor") to conduct wetwo found systems in the shared Swiss the system same back-endwas made code public, for shufflingthe NSW audits of the information technology used under the Electoraland decrypting Commission the votes. announced When the firstthat oftheir the systemweaknesses was approved procedures. affected by the same problem (NSW Electoral Commis- sion, 2019). This was March 12th, 2019. The important

Law in Context, Vol 37, Issue 1, 2020 47 ISSN: 1839-4183

(2) Audits under this section are to be conducted and the system may not be conducted, making the “audit” the results of those audits are to be provided to the meaningless. Electoral Commissioner: Furthermore, the auditor’s report is not public, but (a) at least 7 days before voting commences in each only made to the Commissioner. A request for information Assembly general election at which technology under freedom of information law, while conceivable, is assisted voting is to be available, and unlikely to be approved. When introducing the iVote sys- (b) within 60 days after the return of the writs for each tem in 2010, the NSW Government said (NSW Hansard, Assembly general election at which technology 24 November 2010): assisted voting was available. schedule 3 of this bill will amend the Government Infor- (3) Without limiting the content of the audit, the indepen- mation (Public Access) Act 2009 to protect sensitive infor- dent auditor is to determine whether test votes cast mation kept for the administration of elections, including in accordance with the approved procedures were software programs and codes for the iVote system. The bill will amend the Government Information (Public Access) Act papers produced under those procedures. 2009 to provide for a conclusive presumption of overrid- accurately reflected in the corresponding test ballot (4) The independent auditor may make recommendations ing public interest against disclosure in relation to certain to the Electoral Commissioner to reduce or eliminate provisions in the Act, specifically those concerning secrecy any risks that could affect the security, accuracy or relating to technology-assisted voting... secrecy of voting in accordance with the approved We acknowledge that the auditor is intended to be “in- procedures. dependent” but if the report need not be made public and cannot be disclosed via freedom of information laws, then provision requires compulsory audits, before and after it is impossible to effectively review the auditor’s work. elections,At first byglance, an independent the provision person seems (more fit for onpurpose. this later), The The auditor’s appointment is not characterized by the type of tenure that is typically associated with an independent may make recommendations relating to the “security, accuracywithin specified or secrecy time of limits. voting” By (s implication, 156(4)). The the auditing auditor Act 1997 (Cth)). The NSW Auditor-General, by way of public official (contrast Schedule 2 of the Auditor-General could involve testing of source code, for example, because Public that would arguably fall within the expression “informa- Finance and Audit Act 1983 (NSW), s 28). contrast, holds office for a period of eight years ( tion technology” in s 156(1) and the implied objectives of the audit (as we will demonstrate below). The auditor after the 2019 problems were detected, a number of in- How has section 156 worked in practice? Before and has to provide the results of the audit to the Electoral Commissioner, and this report could be disclosed under The second of SwissPosts’s three cryptographic problems ternal audits failed to find serious cryptographic errors. NSW freedom of information legislation (the Government was made public very near to NSW election day. The Information (Public Access) Act 2009 (NSW) (more on problem allowed the decryption service to fake a proof this below). that the votes had been properly decrypted, while actually Section 156 is, however, defective in material respects. substituting nonsense votes that would not be counted. There is no requirement that the auditor have suitable This time, NSWEC put out a press release claiming that knowledge and technical expertise (a point we return to the problem was ‘not relevant’ to the iVote system. This later). Knowing the right questions to ask is important. was implausible since the two errors affected the same The audit may not include analysis of the source code or part of the code (the mixing and decryption service). But other relevant data. In addition, s 156 does not require since the source code remained secret, there was no way that any particular tests be conducted. This means that to test this claim.

states: tests that are critical and necessary to expose flaws with Does section 157 improve upon s 156? Section 157

48 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

157 INDEPENDENT MONITORING OF TECHNOLOGY ASSISTED VOTING is now known – that electronic voting can be subverted in unqualified auditor or monitor is appointed. Given what (1) The Electoral Commissioner may appoint one or more ways that are not obvious to someone without the requisite independent persons (an "independent monitor" ) to technical expertise who simply “observes” the process monitor and observe the technology assisted voting – we believe that there is a substantial argument that process at an election, including the counting of votes auditors and monitors appointed under these provisions cast by means of technology assisted voting and the must have those skills. It is an unsurprising principle of general operation of the technology assisted voting administrative law that, if a person makes an administra- process. tive decision in circumstances where more than lay skills (2) An independent monitor is to report and may make are required, and the decision-maker clearly lacks the recommendations to the Electoral Commissioner required skills to make the decision, then their decision regarding the technology assisted voting process. could be made irrationally. If so, it could be susceptible to judicial review on the grounds that it is unreasonable (Fuduche v Minister for Immigration, Local Government it confers a discretion on the Electoral Commissioner. and Ethnic Affairs (1993) 45 FCR 515, 527 (Burchett J)). StrictlyThe firstspeaking, point tothat make means about that this the provision Commissioner is that need not exercise their power to appoint an independent system are both dire and obvious. monitor for the purposes of the provision. Assuming that The implications for public confidence in the electoral an appointment is made, s 157 goes further than s 156, We now come to section 158: because it covers “the general operation of the technology 158 SCRUTINEERS assisted voting process” (s 157(1)). There is provision for A candidate or registered party may appoint a scru- a report under s 157, an accountability measure that is tineer to observe: not stipulated by s 156. (a) any production of the printed ballot papers and bun- “Monitoring” goes beyond merely “observing”. “Moni- dling and sealing of those ballot papers in accordance toring” is also a verb that denotes systematic and ongoing with the approved procedures, and review, and implies that there is something to monitor, but (b) any other element of the technology assisted voting in practice it may not require more than observing what process that is approved for the purposes of this has been done. There is no indication in the provision section. what standards would be applied within the monitoring. The breadth of s 158(b) is notable in one respect. It Although possible, there is no indication that monitoring authorizes “a scrutineer to observe: any other element would require any testing, let alone the tests that cryptog- of the technology assisted voting process.” This is a very raphers might apply to ensure the voting was accurate. wide phrase indeed and a phrase commencing with the Like s 156, s 157 only requires the report to be given to perfectly absolute pronoun “any”. the Electoral Commissioner. There is no requirement of That said, the presence of the s 156 auditor and the tabling in Parliament, which would provide Westminster- possibility of a s 157 monitor together indicate that s style checks and balances (such as they are). There is 158 is describing a different process and role. Given the no requirement that the source code be made available. lengthy history of “scrutineers” it is certainly arguable that Again, as with s 156, if the source code or other relevant the role contemplated by this provision is the traditional data was not monitored via s 157, then any freedom of one, i.e., not a role requiring technical knowledge, per information disclosure would not assist. Like s 156, there se, but just the opportunity to “observe”. Observing is a is no requirement in s 157 that the independent monitor less intrusive activity than auditing or even monitoring. have suitable knowledge and technical expertise. This conclusion is reinforced by the presence of ss 156 It is important to dwell on this last point to identify a and 157, which cannot be redundant, consistent with the possible ground for judicial review in the event that an principle of statutory interpretation that Parliaments

Law in Context, Vol 37, Issue 1, 2020 49 ISSN: 1839-4183 do not use surplus language, and that every provision has a distinct purpose. There is a presumption in statu- votes cast and that it is secure. This will allow tests of the general election to ensure that it properly reflects the tory interpretation that parliaments do not use surplus iVote system software to ensure that it is accurate and Project that the secrecy of votes is protected, with the system Blue Sky Inc v Australian Broadcasting Authority (1998) resistant to hackers and any other malicious tampering”. 194language, CLR 382 and [71]). every provision has significance ( Our point though is that far from ensuring that audits, But s 158(b) clearly elaborates on s 158(a). Does it s 159 mandates secrecy of the source code and thereby better understand ss 156-7, we need to consider these restrictsmonitoring the and operation scrutiny that properly ss 156-8 reflect could thehave votes to enable cast, provisionshave a distinctive in context. purpose? So we turnTo understand, to s 159: and to also proper review. 159 SECRECY RELATING TO TECHNOLOGY ASSISTED Yet, where the source code for this system remained VOTING entirely secret until months after the 2019 election, this (1) Any person who becomes aware of how an eligible did not reduce the exploitability of its cryptographic elector, voting in accordance with the approved pro- problems. It is by sheer good luck that Switzerland’s cedures, voted is not to disclose that information to transparency laws happened to allow the exposure of any other person except in accordance with the ap- a problem with the New South Wales system. Without proved procedures. Swiss transparency laws, the citizens of NSW (or even the Maximum penalty: 20 penalty units or imprisonment NSWEC) would have been none the wiser. for 6 months, or both. It is remarkable that there is no detailed account of (2) A person must not disclose to any other person any this problem in the Commission’s review of the 2019 source code or other computer software that relates election. Regrettably, source code for the iVote system to technology assisted voting under the approved remained unavailable at election time, except under a procedures, except in accordance with the approved very restrictive non-disclosure agreement that prevented procedures or in accordance with any arrange- the deviation from the candidate-appointed scrutineer- ment entered into by the person with the Electoral sharing any findings with the public for five years. (Note Commissioner. ing mandated for paper-based electoral processes set out below). Ultimately, four months after the election, NSWEC chose to make its source code available under reasonable code” but penalizes disclosure without authorization. terms (NSW Electoral Commission, 2019). One of us signed SectionSection 158 159 cannot contains be read a specific in a way reference that makes to “source s 159 up for the scheme and checked the relevant part of the redundant. That means that (absent such authorization) code, for the decryption proof problem that NSWEC had scrutiny in s 158 cannot entail access to the source code claimed was 'not relevant' to the iVote system. iVote’s - which we argue elsewhere in this paper is necessary to proof was slightly different from the equivalent part of enable effective scrutiny. the Swiss code, but it was immediately apparent that the A court considering the meaning of these provisions problem was still present (Teague, 14 November 2019). may well have regard to the purpose of their predecessor In summary, the scrutiny provisions of the NSW Elec- provisions (see for example s 120AE of the Parliamentary toral Act (ss 156-9) are weak and inadequate in protecting Electorates and Elections Act 1912 (NSW)). Introducing against the problems with the source code that we have the predecessor to s 156 in 2010, the NSW Government indicated that its purpose was to “guard the integrity of • lack appropriate objectives by which the perfor- - been discussing. Specifically, the provisions: ing on the proposed s 156, the Minister said: mance of the auditor, monitor and scrutineers can the system” (NSW Hansard, 24 November 2010). Reflect be assessed (ballot secrecy, transparency, accuracy) ‘the bill will require an independent audit of the technology assisted voting system both before and after each

50 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

• exclude effective review and analysis of the source combines them with the electronically-cast votes, and code counts them. • do not give that role to the independent auditor or From 2001 to 2016, most of the system's source code the independent monitor was openly available for scrutiny, which permitted valu- • do not require the independent auditor or monitor - to have the required expertise lems. Scientists from the Australian National University discoveredable independent errors examination in the code usedand identification to conduct the of ACT'sprob • do not make their work accountable, as they do not STV count (Goré and Slater 2020). These were found by have to provide a public report about their work. inspecting the openly available source code and compar- • do not guarantee that the auditor or monitor are ing it with the legislation. The most serious errors have independent, i.e. free from real or perceived biases, now been corrected. In separate research, serious privacy either of a political nature or in the sense of prefer- problems were found by T Wilson-Brown, who noticed ring to understate or hide problems. that the system retained detailed timestamps for each vote

3.3 RECOMMENDATION: THE SECRECY PROVISIONS IN thus allowing for easy linking of individual voters with NSW SHOULD BE REPLACED WITH TRANSPARENCY theirand did electronic not shuffle votes votes (Wilson-Brown before they were 2018). posted We online,do not LAWS know whether these problems have been addressed. At Even after one of us had, with colleagues, alerted the the time, the Elections ACT denied the problem existed - (Hayne and Bogle 2018) and since then they have not ternal NSWEC process failed to correct it. Its continuing published any updates to their source code. NSWEC to the decryption-verification problem, the in presence was revealed only when the source code was Two crucial gaps in scrutiny for ACT elections are in the accurate capture of voter intent, when voters cast a after the election. vote on a computer, and when a paper ballot is digitised. finally made available to independent experts months The website of Elections ACT does discuss the ques- voting under reasonable assumptions is not presently tion of whether the process for digitising paper ballots We will argue later that sufficiently secure Internet feasible in NSW. We believe that, if there is a genuinely is accurate: transparent process, it will become evident that the system Following each election, the Commission surveys a does not meet the crucial security requirements of elec- random sample of scanned ballot paper batches from each tions and may well produce inaccurate results and none electorate and compares the final electronic interpretation of us will be the wiser. We also argue that transparency with the data included in the scanned ballot paper data about the system will help to drive better decisions about file. No errors have so far been identified, indicating that accepting its level of risk. electronic scanning and counting is highly accurate (ACT We will return to iVote in Section 4.3 and examine Elections 2020). This sounds comforting, until you realise that no cryptographic errors are corrected. whether iVote’s election data can be verified even if its the paper linked to as a source of 'More information' 3.4 THE ACT'S EVACS SYSTEM datessample from size 2001. or quantification Furthermore, of unless accuracy such is a given, sample and is The Australian Capital Territory runs one of Australia's conducted transparently in the presence of scrutineers, oldest e-voting systems. Voters in the ACT may vote either merely reporting it does not constitute real evidence for on traditional paper ballots or on computers in a polling those scrutineers. We return in Sections 5.3 and 5.4 to place, which store an electronic-only ballot record with audits of digitisation processes – we recommend for the no paper. A separate system digitises the paper votes, ACT the same process we recommend for the Senate.

Law in Context, Vol 37, Issue 1, 2020 51 ISSN: 1839-4183

Source code for the systems of 2001, 2004, 2008, 2012 providing voters and scrutineers with evidence that the and 2016 then follows, but it is written in C, not Ada, so it electronicThe most votes difficult have been gap toaccurately fill in the recorded ACT's system and con is- does not match the description of the 2020 system quoted veyed through the system. When a voter presses a button above. Furthermore, ACT Electoral law was amended on a touch screen, there is no direct evidence that the vote has been recorded as they intended – an electronic electronic counting – it must now round tallies down to 6slightly decimal in places 2019 ratherto change than the to thespecified nearest algorithm integer. This for is unlikely to make much difference, except that it must errors.verification The screen process can of be electronic easily spoofed transfer by throughmalware, the or have required an update to the code – it would not be localeven networkaccidentally at the rendered polling place,inaccurate onto electronicby configuration media, legal to use the 2016 system again in 2020. and to a central counting service, also allows scrutineers no equivalent of the direct visibility of a cardboard ballot were told by email on 24 September 2020 that the 2020 box. Again, electronic manipulation or even accidental codeWe was requested written in clarification Ada and, while from the Elections code could ACT be proand- vided "in the near future", it was not currently available in a paper ballot, or carefully examine a paper printout, because it was "currently continuing through an audit getdata much mishandling, better evidence could be that undetectable. the vote they Voters intended who has fill entered into the scrutiny process. (There is considerable controversy over whether human-readable printouts are departureand certification from their process." earlier They policy also of added making that it openlyaccess adequate, because people often do not check them, but availablewould require was not a confidentialityat all evident from deed. their This website, significant which we sidestep that here---certainly if there is no paper at continued to describe the election software as "open and all, the voter has no opportunity to check directly how transparent" until several days after voting had com- their vote has been recorded). But a careful reading of elections ACT's website indi- the email and was not available on the website but was cates an even more insidious problem: eventuallymenced. The made confidentiality public in response deed was to ournot FoIattached request, to In 2008, 2012 and 2016, an intelligent character recog- two days after voting had started. nition scanning system was used to capture preferences on We were not the only people confused: experienced paper ballots, with intensive manual checks used to ensure journalists described the code as "open source" in an a very high level of accuracy. This data was then combined with the results of the electronic voting, and the computer Apart from the confusing statements that the code is article written at the end of the first week of polling. program distributed preferences under the ACT's Hare- "open and transparent" when it is in fact closed under a Clark electoral system. This system will again be used in 2020. [Our emphasis] - The software for the electronic voting and counting larlyconfidentiality problematic deed, provision: the deed itself defines the researcher's findings as confidential information, and has one particu system is built in the Ada language, a coding language 5(2) "The confidant may publish its Findings only if it intended for the development of high integrity software provides a copy of the Findings to the Territory in writing used in systems where highly reliable operation is essen- at least 60 days before the intended publication". tial and uses open source Linux as the operating system. - sures is a normal convention and is quite independent of system to ensure the election software is reliable, open whetherA period the ofcode confidentiality was openly available.following securityHowever, disclo there This combination was chosen specifically for this electoral and transparent, and could be made available to scruti- are three reasons that this is deeply problematic in this neers, candidates and other participants in the electoral context. process (ACT Elections 2020). Neither the code nor the deed were available before voting started, so there was no way to inspect the code

52 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 and make conclusions public in time to warn voters or This shows why openly available source code is valu- explain possible errors to candidates, let alone to get errors corrected before the election. able for error-correction, but not sufficient for election - and scrutineers to verify that the election has been prop- verification. Genuine scrutiny requires a way for voters ity period overlaps the end of voting and includes the erly conducted without trusting that the code running periodIt should when also candidates be noted might that decide the 60-day whether confidential to accept on the computers is the code they have been shown. In or challenge election results. Section 259 of the Electoral the case of EVACS, a paper record showing voters how Act 1992 (ACT) requires that a dispute about an election their vote had been recorded would help, but only under within 40 days of the declaration of the result the assumption that voters checked carefully for errors. (see also ss 256(2)(e) and 258). Elections ACT would be The scrutiny process would then use the paper, not the must be filed electronic record. We examine methods for election veri- means that it is very unlikely that the Court of Disputed Returnsaware of this,would of course.be in a Theposition 60-day to confidentiality conduct “an inquiry period fication in Sections 5 and 6. into the accuracy of approved computer programs used 4. SHOULD THE SOURCE CODE FOR THE SENATE in electronic voting and the electronic scrutiny of votes” COUNT BE OPEN? (s 269(1)(b)). We now turn to the Federal system. In 2013 Michael Cordover tried unsuccessfully to use freedom of informa- remaining silent after election day – if the problem is no tion law to compel the AEC to publish the code it used for longerIn our exploitable, opinion there then thereis no reasonable is a strong obligation justification to tellfor counting Senate votes (Cordover and Australian Electoral the public, to allow for an honest assessment of the accuracy Commission (Freedom of Information) [2015] AATA 956, of the results. There is a substantial question whether the paragraphs [27]-[36]; see further Cordover 2014). The power of the Court of Disputed Returns would be impeded Senate passed a supporting motion ordering the Special Minister of State to table “the source code of the software displace it. Indeed, it is debatable whether a responsible by which Senate vote counts are conducted,” but this failed researcherby a confidentiality should remain clause, silent and a about legal unpatchedaction would security likely to pass the House when the Minister replied, “I am advised vulnerabilities that only became apparent during voting: that publication of the software could leave the voting disclosure could allow voters to protect themselves from system open to hacking or manipulation” (Sharma 2014). the problem by voting on paper. We accept, though, that This is precisely backwards: the system is highly likely disclosure could conceivably also increase the likelihood to be open to hacking or manipulation, or simply bugs and of exploitation by bad actors. made available. Bringing any problems out into the open be running a more transparent system than other Aus- configuration errors, whether or not the source code is tralianSo while electoral Elections authorities, ACT might there superficially are important seem gaps to corrected. is much more likely to result in them being identified and in the opportunity for scrutineers to verify the process, In July 2015 Cordover’s appeal to the AAT was rejected particularly the accurate capture and electronic transfer when the AEC argued that ‘the documents were exempt from of voter intent. We simply cannot share Elections ACT’s disclosure on the grounds that they contained information assessment of the system as open and transparent and we that had a commercial value that would be diminished if - disclosed’ (Australian Electoral Commission 2016). The ality clause, disingenuous. It is not even clear that errors ‘commercial value’ belonged to the AEC, who were argu- regard Elections ACT’s defence of it, given the confidenti ing that their fee-for-service elections arm would not be have been corrected. The system with openly-available able to compete effectively with commercial providers if in the source code identified by independent researchers source code is simply not the one that voters used. it was forced to reveal its code (Cordover and Australian Electoral Commission (Freedom of Information) [2015]

Law in Context, Vol 37, Issue 1, 2020 53 ISSN: 1839-4183

Technology assisted voting requires the operation of a understand why the AEC’s commercial incentives should complex information technology platform by, or on behalf trumpAATA 956, their paragraphs obligation [27]-[36]).to demonstrate We find that it theirvery hardSenate to of, the Electoral Commission as part of the crucial function counting code is correct. We are not arguing that the AEC of conducting elections for the New South Wales Parliament. is disentitled to have commercial interests or concerns. Accordingly, the Electoral Commissioner requires a degree Its powers and functions are broad (Commonwealth of flexibility in determining and approving procedures for Electoral Act 1918, s 7). But we question the priorities iVoting. However, constitutional integrity of the electoral system requires that any such flexibility is limited by refer- Surely accuracy should be the number one priority of an ence in the bill to principles of accuracy, accountability and electoralreflected commission.in its approach to Mr Cordover’s application. transparency (emphasis added). Indeed, the issue may have constitutional dimensions. (Unfortunately, ss 156-9 of the NSW Electoral Act, as Section 7 of the Australian Constitution states that the we demonstrated earlier, lack the rigour and protections “Senate shall be composed of senators for each State, directly chosen by the people of the State”. Likewise, NSW is accurate, accountable and transparent). to enable the public to have confidence that e-voting in section 24 of the Australian Constitution states that the We also doubt whether the code has commercial value, “House of Representatives shall be composed of members notwithstanding the decision of the AAT in the Cordover directly chosen by the people of the Commonwealth”. In case. The notion that it has commercial value seems ques- their unanimous judgment in Lange v Australian Broad- tionable when several free and open source alternatives casting Corporation (1997) 189 CLR 520 the High Court exist (Silicon Econometrics Pty Ltd 2020; Bowland 2014). of Australia observed that these two sections are part (We note in passing that the AAT’s decision in Cordover appears to have been based, at least in in part, on a con- that Australia has a system of representative government. cession by counsel for Cordover that the code might have of a suite of provisions in the Constitution that confirm Furthermore, the Court unanimously stated that it is “the commercial value: see Cordover and Australian Electoral manner of choice of members of the legislative assembly, Commission (Freedom of Information) [2015] AATA 956, rather than their characteristics or behaviour, which is paragraph [35]). generally taken to be the criterion of a representative form The practical value of using openly available source of government” (emphasis added) (p 559). The High Court has acknowledged that elections need to be administered demonstrated in other parts of Australia. The New South having regard to “competing considerations relevant to the Walescode to STV find count, errors like in the STV Australian counting Senate,systems uses has secret been code to count public data. Using that data, which includes choice by the people” (Murphy v Electoral Commissioner both NSW Legislative Council and numerous NSW local making of a free, informed, peaceful, efficient and prompt (2016) 261 CLR 28 at 88 [184]; cited with approval in government elections, Conway et al were able to compare Palmer v Australian Electoral Commission [2019] HCA - ment of accuracy veryour results high probability) with the official elected transcripts. the wrong This candidate uncovered to a 14, [8]). Nevertheless, we feel confident that a require mandate of direct choice in the constitutional provisions NSWseveral local errors council in the (Conway official code, et al 2017).including We one have that already (with flows directly from the constitutional described the errors in the ACT counting code, discovered Senate vote has not been compromised, then this not by researchers at ANU. So the existence of errors in the identified above. If we cannot feel confident that the only raises cryptographic concerns, it may well raise Senate counting code is highly plausible, though no errors constitutional concerns under s 7 of the Constitution. have ever been found. As the NSW Government noted when it introduced the The Senate code has become even more complex since predecessor provisions to ss 156-9 of the Electoral Act Mr Cordover’s unsuccessful FoI application. Now there is 2017 (NSW Hansard, 24 November 2010): a complex electronic process to digitise and then count

54 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 the paper ballots. The public preference data allows us to In 2008, Rivest and Wack coined the term Software In- double-check the counting phase, just as we can in NSW, dependent to describe a system that provides the evidence but the rest could fail undetectably. necessary to check whether the software has correctly Based on the foregoing analysis, we recommend that recorded, included and counted all the votes (Rivest and legislation mandate openly-available source code for all Wack 2008). A voting system is Software Independent if an electoral processes (readable and compilable so that a (undetected) change or error in its software cannot cause thorough examination can be conducted). an undetectable change or error in an election outcome. The terms Software Independent and verifiable are often used interchangeably, but more often refers 5. GENUINE AUDITING, MONITORING AND verifiable SCRUTINY OF THE DATA that their own vote is properly recorded, or an observer’s In this article we have argued that verifiability does opportunityto a specific test,to check such that as a allvoter’s votes opportunity have been correctlyto check not come automatically in electronic systems: it is a secu- included in the count, whereas Software Independent describes an entire electoral process. process. In this section of the article we explore different methodsrity property for genuine that must auditing, be specifically monitoring designed and scrutinyinto the The most obvious way to make an electoral process software independent is not to use any software. Aus- that the Senate voting system is, at present, not effectively tralia’s internationally celebrated process of a manual that can help ensure verifiability. Then we demonstrate count of hand-marked paper ballots obviously has this property. There are, however, innovative ways of engineer- verifiable. 5.1 “SOFTWARE INDEPENDENCE” computers, while still ensuring software independence. Although inspecting the source code can identify er- Foring electoralexample, Westernprocesses Australia to derive and much South of Australiathe benefit have of rors, only the election data (i.e. the real votes) can provide both provided computerised assistance to voters with genuine evidence of an accurate election outcome. This disabilities, allowing them to use a computer in a polling section examines how electronic systems could be adapted place to complete a paper ballot. This is software inde- to allow scrutineers to examine the votes and hence derive pendent because (except for voters with a severe vision direct evidence of the accuracy of the outcome. impairment) the voter can directly verify that the marks Many electronic electoral processes are not designed to permit genuine scrutiny of the data. The most notorious observe the paper going into a ballot box, where it enters are the Direct-Recording-Electronic (DRE) voting machines theon their normal paper scrutiny ballot process. reflect their instructions, and then that have been common in the USA since about the year Victoria’s now-discontinued vVote project used a 2000, but are gradually being phased out after a series complicated combination of voter challenges and cryp- of disastrous security and accuracy failures (Feldman, tographic proofs to achieve software independence for Halderman and Felten 2006; Aviv et al 2008; Checkoway voters in remote supervised locations such as overseas et al 2009; Ottoboni and Stark 2019). DREs are used in embassies (Parliament of Victoria 2017, p 162). polling places to make an all-electronic record of each person’s vote. Although often characterised as a security 5.2 RISK LIMITING AUDITS problem, the real problem with paperless electronic voting machines is a verifiability problem. It is not possible to Statistical methods can also be used to verify election discern whether the internal electronic record correctly outcomes. Election observers can observe (or participate encodes the voter’s intention or has instead been mis- directly in) the audit and hence gain evidence that the recorded because of a software bug, insider manipulation, election outcome is correct. or security failure. A Risk Limiting Audit - antee: that if the election outcome is wrong, the audit meets a specific statistical guar

Law in Context, Vol 37, Issue 1, 2020 55 ISSN: 1839-4183 will not pass except with some small probability (the risk evidence, scrutineers would need to see the electronic limit) determined in advance. digitised preferences in advance, the fair generation of Many US jurisdictions count their votes electronically random ballot selections, and the actual marks on the but conduct a rigorous statistical audit of the paper bal- randomly selected paper ballots. They would then need to be (authorized to) replicate the statistical computations California Law (California Election Code § 15367) speci- on their own computers if they wished. lots against the final count (Breedon and Bryant 2019). includes a provision for Risk Limiting Audits. The law 5.4 WHO SHOULD VERIFY THE SENATE SCANNING fies mandatory post-election audits of paper ballots and PROCESS? constitutes meaningful evidence of voter intent, and how It is entirely appropriate for an authority running an thegoes audit into considerableprocess should detail, be conducted. defining what sort of ballot information technology process to pay professionals to Note especially Section 15367(c): assess its security and reliability. Contracted code review, “The risk-limiting audit shall be a public and observ- intrusion testing, stress testing, et cetera, are all good ways able process”. to improve the process. However, these do not perform the same function as candidate-appointed scrutineering, even if they do improve the security of the system. To put 5.3 HOW COULD WE VERIFY THE SENATE SCANNING PROCESS? it bluntly: you have to trust the contractors, but many candidates have no reason to do so. Australian Senate ballots are cast on paper and then converted into electronic preferences in a complex process The Australian Parliament’s Joint Standing Committee that combines both automated character recognition and on electoral matters recommended in 2018 that the AEC human data entry. Unfortunately, the lack of transparency appoint its own technical ‘expert scrutineer’ to examine of the process means that random errors and deliberate the electronic Senate scanning system: manipulation could be possible and would not neces- Recommendation 3 The Committee recommends that a sarily be detectable by scrutineers. It is not “software non-partisan independent expert scrutineer be appointed independent”. to each Central Senate Scrutiny Centre in each state and The Australian Electoral Commission claims to per- territory and be responsible for: form some estimates of the accuracy of its Senate scan- • auditing the computer systems and processes used to ning process, but since these are conducted away from capture and count votes; • undertaking randomised checks between captured them. There is no provably risk-limiting technique for data and physical ballot papers throughout the count scrutiny it is extremely difficult to find any detail about at a level that provides surety as to the accuracy of assess whether a small change early in the elimination the system; and orderauditing has the cascaded Senate intocount, a completely because it differentis very difficult outcome. to • providing reports to candidate scrutineers about their Nevertheless, it would still be valuable to conduct a rig- findings on a regular basis during the count (AEC orous statistical audit comparing the electronic digitised 2018, p xxiii). preferences against the actual paper ballots. This would When asked afterwards whether they had implemented provide an estimate of the rate of scanning errors, could this recommendation, the AEC replied: detect some forms of deliberate fraud, and, if conducted in a way that allowed meaningful scrutiny, provide evidence ... there are a number of difficulties with implementing supporting the announced election outcome. Recommendation 3, not the least of which are This should be conducted in a way that allows scruti- 1. The difficulties of proving ‘non-partisan’, ‘independent’ neers to check both the algorithms and the data. In order and ‘expert’ in relation to a scrutineer. to truly “observe” in the sense of gaining meaningful

56 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

2. It is the AEC’s legislated role to deliver elections in an that Australian democratic traditions and respect for ‘independent, non-partisan’ manner. Essentially, the AEC the rule of law would tolerate delays in the production is not a participant in the election - it is the independent of results. Surely accuracy is a paramount consideration ‘umpire’. Accordingly, the appointment of (another) in an election. independent arbitrator would create significant confu- The AEC concluded: sion, and potentially prevent the AEC from fulfilling its For these reasons the AEC does not intend to take any statutory function...(AEC, 6 December 2019). action regarding this recommendation unless legislatively We agree. This is exactly why asking the electoral compelled to do so. commission to select someone ‘independent’ and ‘expert’ We believe them. For the reasons we have developed appears not to have worked in NSW. in this article, steps should now be taken to develop and This seems to us to be an invented problem to which enact such legislation. we already have the solution: scrutineers should be responsible for assessing whether they have been given surety as to the accuracy of the system. It is perfectly 6. IS THE IVOTE PROTOCOL VERIFIABLE? In this section we show that the NSW iVote protocol manual work necessary to undertake randomised checks betweenreasonable the tocaptured appoint data an and AEC the official physical to ballotperform papers, the The most challenging part of an Internet voting system is not genuinely verifiable. but there is no need for that person to be ‘non-partisan’, is what Swiss law calls individual verifiability, the oppor- ’independent’ or ’expert’ because all their actions should be tunity for each voter to verify that their electronic vote available to meaningful scrutiny by candidate-appointed scrutineers. Hence our amended version of the JSCEM’s uploaded (Guasch Castellò, no date). This step is the Recommendation 3, given above. (One other pedantic hardestaccurately to designreflects because their intention it needs andto work has beenfor voters, properly not just technical specialists or even scrutineers. It needs to ‘surety’. Certainty could be achieved only by examining defend against system errors, insider attacks, and malware allpoint: of the a statistical paper ballots). audit can provide confidence, but not on the voter’s computer, however insecure it might be. A rigorous, statistical audit that compares physical It cannot be delegated to others because the content of ballot papers with electronic derived data at a level that the vote should be private. In a polling place, the obvious become a normal part of the Senate vote count. Like every a paper printout; when voting electronically from home, solution is to ask voters to fill in a paper ballot or check otherprovides part confidence of the Senate as to count, the accuracy it should of bethe conducted result should by there is no corresponding simple check. If the method is too complex for voters to execute properly, then they may way that gives scrutineers full visibility of the process so thatAEC officialsthey can and verify it should that it be has designed been conducted and conducted correctly. in a not have checked their vote properly. think they have verified their vote when in fact they may At this point, the second part of the AEC’s objection You will recall from Section 3.1 that this was the aspect might still apply: already having been in use for some time. The Estonian Further, such a process would significantly impact the e-votingof Swiss verifiabilitysystem uses that a direct was showncryptographic to be unsound demonstra after- efficiency of electoral processes and the timeliness of pro- tion that the encryption has been properly conducted— ducing a result... (AEC 2019, p 12). voters have to scan a QR code and use another device This is also true, but we would rather have a slowly to recompute and check that the correct vote has been computed correct result than a speedy one that incorpo- encrypted. In some ways this is a better and more direct rated serious error or fraud. The sky did not fall in when the 1990 Federal election result took eight days to tally much more cumbersome and it opens up opportunities verification mechanism than the Swiss approach, but it is (Commonwealth Parliament 1990). We feel confident

Law in Context, Vol 37, Issue 1, 2020 57 ISSN: 1839-4183 for vote selling, because voters can prove the contents of voting, email voting, web-loading PDFs or any other form their encrypted vote. of remote paperless e-voting. When votes are cast and We do not know of any secure, reliable, usable method counted manually, being able to stand in the room and watch immediately allows a scrutineer to observe that the use anywhere in the world or proposed in the academic process is running correctly. For an electronic process, the literature.of individual voter verification for remote e-voting in process must be deliberately and very carefully designed NSW votes are much more complex than votes in ei- to provide this evidence to voters and scrutineers. If it does ther Switzerland or Estonia, which further increases the not provide such evidence, it is subject to undetectable fraud. There are reasonable solutions in a polling place, choose one candidate or party, or sometimes one option ondifficulty a referendum, of this crucial while step.NSW Swiss voters and list Estonian preferences voters of laws be changed to require verifiability for all (electronic) but verifiability remains crucial. We recommend that sometimes hundreds of candidates (Green 2019). electoral processes. Unfortunately, the NSW iVote protocol does not provide a reasonable solution to this problem. The designs 7. THE SECRET ELECTRONIC BALLOT have varied through different rollouts in 2011, 2015 and There is no real electronic equivalent of a physical bal- 2019, but the most recent version asks users to download lot box, which allows a voter to see that nobody else can see how they are voting. If there is malware, a key logger, separate from the computer they used to vote on. Voters or some other security problem on the computer that a casta Verification their vote App using onto the a deviceiVote website, (such as which a phone) then that pro is- person uses to cast their vote, then their vote privacy duces a QR code which the voter reads with the app on cannot be defended. Furthermore, these problems could their phone. The App is then supposed to query the iVote be undetectable by the voter. voting server to ask (given the voter’s login credentials) However, there are well-established cryptographic which vote is recorded on that voter’s behalf. The App techniques for protecting the privacy of the vote after it then displays the vote to the voter, who is supposed to has been cast. Since 2015, iVote has provided end-to-end check that it matches what they wanted. encryption of the ballot between the voter’s device and However, both the App and the Internet voting system the electoral commission. That means that the vote is are provided by the same company, which gives voters (supposed to be) encrypted in such a way such that even the electoral commission web server cannot decrypt it app’. So we are trusting the company that provided the (but see Culnane et al 2017). Some systems (including votingno opportunity software to attestreimplement to whether their its own voting ‘verification software the Swiss system) also provide a cryptographic mecha- accurately encoded the vote. This might be a useful check against malware on the voter’s computer, but it provides provide a guarantee of privacy that depends on electronic no defence at all against insider attacks of the electoral processes,nism for electronically such as the proper shuffling generation the vote. of These randomness, systems commission, the software provider, or anyone who compromises them. An attacker could simply alter the vote in depends on trust. - that cannot be independently verified. Ballot privacy tion App to present the voter with the vote they wanted, 8. COULD ELECTRONICALLY-DEPENDENT ELECTION ratherthe voter’s than web the browser,vote that then was useactually the subverted sent. Verifica RESULTS BE CHALLENGED? This is the central unsolved problem of remote e-voting, The impossibility of verifying an electronic process and it is even more complex for Australian preferential relates to the difference between proving that there was votes than it is in other countries. This motivates our a large enough problem to affect the outcome and prov- recommendation to avoid paperless Internet voting alto- gether. We recommend that we should not allow Internet In most jurisdictions in Australia a “court of disputed ing that the problem worked against a specific candidate.

58 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 returns” has power to make orders addressing defects in 352). That said, courts have wide powers which include voting processes.” (See e.g., Commonwealth Electoral Act s power to declare that any person who was returned was 360(2).The Electoral Act 1992 (ACT) allows ‘that the Court not duly elected; to declare any candidate duly elected may make any orders in relation to the application that who was not returned as elected; and to declare any the court considers appropriate’ (s 265), the Electoral Act election absolutely void. So, for example, when the AEC 2017 (NSW) allows the Court to ‘exercise all or any of its misplaced 1,375 ballots in the 2013 Western Australian powers … on such grounds as the Court in its discretion Senate election, the WA Greens argued that reversing a Electoral Act difference of 14 votes, and hence eliminating a differ- 1992 (Qld) allows that the Court may make any order ent candidate earlier in the count, would cascade into a orthinks exercise just anyand powersufficient’ in relation (s 225(2)), to the the application that different set of winners. They were able to produce an the court considers just and equitable (s 146(1)), the open-source reimplementation of the Senate counting Electoral Act 1985 (SA) requires the Court to be guided by algorithm (Bowland 2018), which allowed other interested the ‘good conscience and the substantial merits of each parties to redo the count. case’ (s 106(1)), the Electoral Act 2004 (Tas) requires the A challenge to iVote based on the magnitude of prob- Court ‘to be guided by the substantial merits and good lems (such as the registration site crashing) might fail in conscience of the case’ (s 212(2)(a)), in Victoria the Court NSW (because it is impossible to prove what the miss- must act fairly and according to the substantial merits ing or suspect votes should have been), but succeed in of the petition’ (Electoral Act 2002 (Vic) s 126), and in WA, the relevant provision, titled ‘Court must act fairly’ the problem was large enough to change the outcome. Of allows that the Court can exercise all or any of its powers course,a jurisdiction for security where vulnerabilities it suffices to prove or software that the errors size ofit - Electoral Act 1907 (WA) s 126(2)). Only lem, even when its existence can be clearly demonstrated. ‘on such grounds as the Court in its discretion thinks fit the Electoral Act 2004 (NT) has no such broad enabling Normay isbe a extremely formal challenge difficult to necessary assess the for size changing of the prob the and sufficient’ ( provision). These provisions give superior courts wide result: Conway et al's independent implementation was power to invalidate an election based upon any breach of the relevant Electoral Act. Under Commonwealth law, loss was improbable based on the randomised count—at a court would not invalidate an election “unless the Court leastsufficient one toof informthose candidates some candidates heard thatabout their our electoral results and demanded a recount, which, (as expected, with high affected” (s 363(3)) (by an illegal practice in the instance probability) gave him a seat. ofis satisfiedthis particular that the provision). result of Odgersthe election Senate was Practice likely (14th to be ed) states: 9. ELECTORAL COMMISSION ANALYSIS AND Recounts normally occur only when the result of an GUIDELINES election is very close. At any time before the declaration of The electoral commissions themselves have examined the result of an election, the officer conducting the election the relative merits of various kinds of electronic systems, may, at the written request of a candidate or on the officer’s own decision, recount some or all of the ballot papers. The The AEC submitted to the Commonwealth Joint Standing Electoral Commissioner or an Australian Electoral Officer Committeeconsidering on both Electoral their risks Matters and their(2018): potential benefits. may direct a recount. The AEC remains of the view that the Electoral Act and … related laws should be refined to remove unnecessarily There are time limits on petitions that cannot be set prescriptive language and to further streamline processes. aside (Rudolphy v Lightfoot (1999) 197 CLR 500) and a The language in the current Electoral Act impedes the Court cannot declare a whole general election void (Ab- AEC’s ability to innovate and to deliver services in the most botto v Australian Electoral Commission (1997) 144 ALR efficient manner possible.

Law in Context, Vol 37, Issue 1, 2020 59 ISSN: 1839-4183

We think this is a false dichotomy, because well-written principle almost entirely. As described above, the system provides very limited vote privacy, transparency only to those endorsed by the Commissioner, and no meaningful regulations,standards for for privacy example, and do verifiability not prevent could car stillmanufac allow- plenty of flexibility for innovative design. Auto emissions turers from innovating, they merely prevent them from For the foregoing reasons we recommend developing verifiability whatsoever. falling below an acceptable standard. They might even these principles into laws, rather than providing mere provide incentives to innovate to achieve further emis- guidance. It seems highly unlikely to us that these strongly sions reductions. worded requirements could be met by an Internet voting system, but the principles still have value for electroni- 9.1 THE ECANZ ESSENTIAL PRINCIPLES cally assisted voting in a polling place. For example, an The Electoral Council of Australia and New Zealand electronically assisted ballot marker could help a voter have published eleven “essential principles for an Austra- with a physical disability to complete their own ballot, but lian internet voting service” (NSW Electoral Commission still allow the person an opportunity to see their paper 2019). Although not technically detailed, the intention of ballot deposited in a physical ballot box. the guidelines seems mostly satisfactory. For example, the section on transparency is: Our recommendations? Implement the guidelines Transparency electronically assisted voting solution in a polling place forabout those verifiability voters who and need transparency this reasonable into accommodation law. Design an The service and processes be designed to enable scrutiny, and ensure that it meets rigorous mandatory standards to provide stakeholder confidence. The internet voting service and accompanying processes will be established with a focus on transparency. [...] Upon for transparency, security, privacy and verifiability. casting their vote, the service will verify to the voter that 9.2 THE WILKINS REPORT ON IVOTE his or her intention is accurately represented and that the The NSWEC commissioned a report into its iVote system vote has been submitted. Any alteration to the voter’s vote by Roger Wilkins AO (Wilkins 2018). It is often cited as a should be detected by the service. Voters and third parties should be able to observe the count of the votes and check reference was to assess “whether the security of the iVote reason to trust in iVote’s security, because its first term of that only eligible voters’ votes are included in the results. The service will provide evidence that only eligible voters’ respect, Mr Wilkins does not have the requisite technical system is appropriate and sufficient” (p v). However, with votes have been included and this evidence will be auditable. expertise and did not assess the code or the cryptographic Clear and unambiguous information about the internet voting service should be available to the public explain- page of his report, that “this is not a report that is going protocol directly. Indeed, Mr Wilkins indicated, in the first ing how to use the service and how the service operates. to be able to give detailed technological solutions. I do The service should be open for verification, assurance and not have that expertise.” scrutiny purposes. Observers, to the extent permitted by Mr Wilkins does mention the possibility of undetected law, shall be enabled to observe, comment on and scrutinise fraud: the internet voting component of an election, including the A more troubling premise might also be conceded as compilation of the results. well. That is the contention that any system could in theory This is well-intentioned, though light on detail. Unfor- be penetrated and manipulated without the penetration tunately, no Australian e-voting laws mandate any such and manipulation being detected (p 19). that the New South Wales Electoral Act requires exactly seems to identify the central risk of electronic voting—the transparency, privacy or verifiability. We have already seen the opposite. Australia’s only continuing Internet voting prospectWe remain of undetectable baffled by fraud—andthis section then,of the like report, Queen which Vic- system, the NSW iVote system, ignores this transparency toria’s apocryphal refusal to outlaw lesbianism (Jennings

60 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

2007), argue that fraud does not need to be prevented Respectfully, we wonder whether the author has ever because it could not possibly exist. experienced an undetected plane crash or would bank - online if there was no way to tell whether money had been stolen from his account. preconditionNote first that for theMr Wilkinssecond, whenrefers arguably(throughout the thisgreatest sec So this summarises perfectly the question for our elec- threattion) to (particularly “penetration for and undetectable manipulation” fraud) as if comesthe first from is a toral law: is it acceptable for electronic voting processes to insiders. Mr Wilkins then gives three reasons this threat permit undetectable electoral fraud, because we assume does not need to be defended against. We pull them apart

it does not exist, or is it necessary to enforce a verifiable (1) As indicated in this report, I consider that on the cur- here. The first is this: design, so that people can verify that fraud did not occur? rent scale of internet voting it is unlikely that people suggestions. will want to intervene to try to alter the election result. If we opt for the latter, we reiterate our specific In any event, this is a matter of intelligence and it is 10. SPECIFIC RECOMMENDATIONS an empirical question. The level of realistic risk is an We now provide a summary of the recommendations empirical matter, and a key recommendation of this given elsewhere in this essay. report is that electoral commissions should get very serious about integrating that intelligence into the • Repeal the secrecy laws and replace them with transparency laws. The Swiss laws are a good example way elections are run (Wilkins p 20). to follow. We do not understand how the likelihood of an undetected • it cannot be tested. This is simply an assertion, without published, there should be a rigorous statistical audit When the preference data files for Senate votes are anyevent evidence, could be that an empirical nobody will question, bother since attempting by definition fraud. - lots. This should be conducted in a way that allows (2) In theory, while penetration and manipulation of results Scrutineersto check that to theycheck accurately both the algorithms reflect the and paper the data. bal may not be detected, as a matter of fact it is highly likely that intervention that changed results would • Mandate openly-available source code for all impor- be detected. Psephologists, political parties, pollsters tant electoral processes. (We omit important details about exactly what “open” means for code: in this case, and other experts would most likely query and ques- we would say it must be readable and compilable so tion outcomes that are inconsistent with expectations. that a thorough examination can be made. See the - wording in the Swiss legislation, infra). mightAgain, be true respectfully, for large manipulations, Mr Wilkins gives but givenno specific the recent jus • Do not allow Internet voting, email voting, web-loading PDFs or any other form of remote paperless e-voting. tificationhistory of of very this close issue election of fact. Weresults believe at both his assumption a state and federal level in Australia, and the considerable disparities • Require verifiability for all (electronic) electoral between predicted and actual poll results (Cockburn and processes. Kontominas 2019), we do not see any adequate way of • - distinguishing manipulation of results from inaccuracy parency into laws. Design an electronically-assisted of predictions. votingDevelop solution the guidelines in a polling about place, verifiability for those and voters trans who need one, and ensure that it meets rigorous to understand that the key word is undetected. mandatory standards for transparency, security, The final argument demonstrates a complete failure (3) If the mere theoretical possibility of intrusion and These recommendations are also aligned and are com- manipulation were sufficient to stop doing things, then we privacy and verifiability. would not be flying in aeroplanes, using mobile phones, and patible with the AI4People recommendations on trans- engaging in electronic commerce and banking (Wilkins p 20). parency and explainability for AI governance (Pagallo et

Law in Context, Vol 37, Issue 1, 2020 61 ISSN: 1839-4183 al. 2019). All of the workable methods for verifying votes that we know of involve a human-readable paper record to give scrutineers evidence of a correct election result. of the vote. What changes to the law would achieve all of However, examination of the code is not sufficient deliberate manipulation of the data might mean that the resultsConfiguration are wrong errors, even malware, though nounauthorised errors were access detected or these objectives? 11. CONCLUSION in the code that was meant to be used. If the electoral We reiterate our observations that accuracy is of process is not designed to produce a Software Indepen- paramount importance, and indeed constitutional im- dent evidence trail, errors or fraud could be undetectable. portance, so legislation in Australian polities that fails Australian electoral laws should now be reformed to enable genuine auditing, monitoring and scrutiny by processes, and meaningful data for candidate-appointed to ensure verifiable designs, transparent software and accuracy should be repealed and replaced with laws that scrutineers to examine. aresuitably consistent qualified with professionals the design principles in order set to outdetermine above. This will mean abandoning secrecy provisions (such 12. REFERENCES as the ones in New South Wales and Western Australia) 1. Australian Capital Territory, Electronic voting and counting, and also abandoning the current systems and processes that cannot meet minimum standards for transparency, https://www.elections.act.gov.au/elections_and_voting/ - 2. Australian Electoral Commission. 2016. Managing the AEC: electronic_voting_and_counting, accessed 20 August 2020. tion of existing electoral law would imply that candidate- appointedsecurity, privacy scrutineers and verifiability. are allowed A purposive an opportunity interpreta for annualreport.aec.gov.au/2016/managing/scrutiny.html>., accessedExternal scrutiny.21 August Commonwealth 2020. of Australia , M. 2008. “Security Evaluation accessed of Es&S 24Voting August Machines 2020. Aviv, A., Černy, P., Clark, A., Cronin, E., Shah, G., Sherr, M. and into e-voting and e-counting systems. and Election Management System.” In Proceedings of the Meaningful verifiability must be specifically designed Conference on Electronic Voting Technology, 1–13. Analysis of an election system and analysis of election 5. Balinski, Michel and Laraki, Rida. 2010. Majority judgment: data are complementary and both contribute positively measuring, ranking, and electing. Cambridge, Massachusetts: to evidence of properly conducted electoral processes. MIT Press. Openly available source code and documentation is 6. Barrat i Esteve, J., Goldsmith, B., Turner, J. 2012. Interna- - tional Experience with E-Voting, International Foundation for Electoral Systems. critically important for finding errors. By carefully exam 7. Bowland, G. 2018. "Counting the Western Australian Senate processesining the software for both and the itsSwissPost specification, system observers and the iVotewere Election: How I ended up verifying the Australian Electoral system,able were which able allowed to show an critical attacker gaps to produce in the verificationapparently- Commission's count of the 2013 senate result" Oreamnos valid proofs of an accurate election outcome that had been (accessed 24 August 2020). manipulated. Thus the opportunity to examine the code 8. Bowland, G., Dividebatur: process single-transferable- vote elections as used for the Australian Senate under the exposed a failure of verifiability.

62 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

Commonwealth Electoral Act (1918) (8 May 2014) GitHub 21. Culnane, C., Ryan, P. Y. A., Schneider, S., Teague, V. 2015.

August 2020. Information and System Security (TISSEC) 18 (1): 1–30. 9. Breedon,. Kimberly and Bryant, Christopher A. 2019. accessed “Counting 21 22. Federal“vVote: ACouncil Verifiable of the Voting Swiss System.” Confederation. ACM Transactions 2019. "Federal on the Votes: Electronic Voting Irregularities, Election Integrity, and Public Corruption.” The University of Memphis Law gov/en/start/documentation/media-releases.msg-id-74508. Review 49(4): 979-1017. html>,Chancellery accessed to review 24 August e-voting." 2020. , accessed 24 August 2020. sage: How Australia Got Compulsory Voting. Text Publishing. 24. Ghale,of the DieboldM., Goré, Accuvote-TS R., Pattinson, Voting D., Tiwari, Machine.” M. 2018. Felten, E., Shacham, H. 2009. “Can DREs Provide Long-Lasting https://www.abc.net.au/news/elections/nsw/2019/guide/ legislative-council, accessed 24 August 2020 the AVC Advantage.” Proceedings of the USENIX workshop 27. - onSecurity? Electronic The Voting Case of (EVT/WOTE). Return-Oriented Programming and tronic Voting, Universitat Politècnica de Catalunya, 35-58. 15. Cockburn, P., Kontominas, B. 2019. "Election 2019: How 28. Gumbel,Guasch Castellò, Andrew. S., 2005. No date. Steal Individual This Vote: Verifiability Dirty Elections in Elec and the polls got it so wrong in predicting a Labor victory." ABC the Rotten History of Democracy in America. Nation Books. 29. Haines, T., Lewis, S. J., Pereira, O., Teague, V. 2020. “How not election-results-how-the-polls-got-it-so-wrong/11128176>, to prove your election outcome.” 41st IEEE Symposium on accessedNews , Buying.” California Law Review, 88(5): Report on all Aspects of the Conduct of the 2016 Federal 1323-1371. Election and Matters Related Thereto. 31. Hayne, J. and Bogle, A. 2018. "Elections ACT dismisses con- 17. Commonwealth of Australia. Parliament. Joint Standing cerns electronic ballots could be traced to voters." ABC News Committee on Electoral Matters. 1990. 1990 Federal Elec- tion: Report from the Joint Standing Committee on Electoral Matters. accessed, R. 2007. A Lesbian history of Britain: love and of New South Wales Electronic Vote Counting.” Proceedings sex between women since 1500. Oxford: Greenwood World of the Australasian Computer Science Week Multiconfer- publications/a-lesbian-history-of-britain-love-and-sex- 21 August 2020. between-women-since-150>,Publishing. Cited at , accessed 19. Cordover, Michael. 2004. Software by which Senate counts 33. Jones, D., Simons, B. 2012. Broken Ballots: Will Your Vote

34. Lewis, S. J., Pereira, O., Teague, V. 2019. “Trapdoor commit- 21are August conducted, 2020. Right to Know, , C., Eldridge, M., Essex, A., Teague, V. 2017. accessed “Trust Implications of DDoS Protection in Online Elections.” In- SwissVote>,ments in the accessedSwissPost 20 e-voting August shuffle 2020. proof.” University of ternational Joint Conference on Electronic Voting, 127–45. Melbourne.

Law in Context, Vol 37, Issue 1, 2020 63 ISSN: 1839-4183

35. Royal Society A: Mathematical, Physical and Engineering Voting in Switzerland and Estonia.” SAIS Review of Inter- Sciences, 366 (1881): 3759–67. nationalLust, Aleksander. Affairs, 38(1): 2018. 65-79. “I-Vote, Therefore I Am? Internet 47. Sawer, M. 2001. “Inventing the Nation Through the Ballot 36. Mannheim, M., Lowrey, T. 2020. "Most ACT election votes will Box.” Papers on Parliament, No 37, November 2001. likely be cast electronically in 2020. Here's how electronic 48. Serdült, U., Germann, M., Mendez, F., Portenier, A., Wellig, C. 2015. “Fifteen Years of Internet Voting in Switzerland: 02/electronic-voting-in-act-election-and-if-its-safe-and- History, Governance and Use.” Proceedings of the Second secure/12722912>,voting works."

News-media-releases/NSW-Electoral-Commission-iVote- government-rejects-senate-order-to-disclose-electoral- and-Swiss-Post-e-vo>,,Morning Herald ,2019 NSW State election" accessed , accessed 24 40. NSW Electoral Commission. 2019. Report on the Conduct August2020). "Count2020. Australian STV elections." ,decryption proof flaw identified in the accessed SwissPost 24 systemAugust affects 2020. 41. NSW Hansard. 2010. Parliamentary Electorates and Elec- 52. Tudeman,the iVote system N. 1995. too." “The , Further Amendment accessed Bill 24 2010. August (Mr 2020. Aquilina.) of Economic Perspectives, 9(1): 27-38. 42. O’Collins, M. 2002. An Uneasy Relationship: Norfolk Island 53. Victoria. Parliament. 2017. Inquiry into Electronic Voting. and the Commonwealth of Australia. ANU Press. Electoral Matters Committee. 43. Ottoboni, K., Stark, P. B. 2019. “Election Integrity and Elec- 54. Wilkins, Roger. 2018. Report on the Security of the iVote tronic Voting Machines in 2018 Georgia, USA.” In International Joint Conference on Electronic Voting, 166–82. Springer. NSWEC/Reports/iVote%20reports/Report-on-the-Security- 44. Pagallo, U., Aurucci, P., Casanovas, P., Chatila, R., Chazerand, P., of-the-iVote-System.PDF>,System. , accessed 21 https://www.eismd.eu/wp-content/uploads/2019/11/ AugustElections. 2020.

45. LEGISLATION relationAI4Peoples-Report-on-Good-AI-Governance_compressed.pdf to the online and telephone voting system (“iVote") for PwC. 2019. Post-Election Report – Specified Procedures in Australian legislation getmedia/b2280c43-a129-47ca-bd75-f9c98887736b/2019- 1997 (Cth), Schedule 2 State-Elections-iVote-review-(post-election-report)-June-17-2019 NSW State Election. , Ronald L. 2008. “On the Notion of ‘Software Independ accessed- Electoral Act 1992 (ACT), ss 122-3, 265 ence’ in Voting Systems.” Philosophical Transactions of the Electoral Act 2017 (NSW), Division 7, ss 156-9, 225 Electoral Act 2004 (NT), ss 46-7, 128 and Div 5 subdiv 2

64 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

Electoral Act 1992 (Qld), ss 104, 146 Voting (VEleS)] (Switzerland) 13 December 2013 SR Electoral Act 1985 (SA), ss 67, 106 and Part 10 161.116 articles 4.2, 5.4, 7b. Electoral Act 2004 (Tas), Parts 5 and 11, and ss 172 and 212 Case law Electoral Act 2002 (Vic), ss 76, 110, 111, 114, 116, 119 Abbotto v Australian Electoral Commission (1997) 144 and 126 ALR 352 Electoral Act 1907 (WA), ss 92, 99, 117, 134, 137, 144-6. CIC Insurance Ltd v Bankstown Football Club Ltd (1997) Norfolk Island Act 1979 (Cth) (repealed) 187 CLR 384 Norfolk Island Legislation Amendment Act 2015 (Cth) Cordover and Australian Electoral Commission (Freedom of Information) [2015] AATA 956 Public Finance and Audit Act 1983 (NSW), ss 27B(4) and 28 Fuduche v Minister for Immigration, Local Government California legislation and Ethnic Affairs (1993) 45 FCR 515 California Election Code § 15367. Lange v Australian Broadcasting Corporation (1997) 189 CLR 520. Swiss legislation Murphy v Electoral Commissioner (2016) 261 CLR 28.

Verordnung der BK über die elektronische Stimmabgabe [2019] HCA 14 (VEleS) [Federal Chancellery Ordinance on Electronic Palmer v Australian Electoral Commission Project Blue Sky Inc v Australian Broadcasting Authority (1998) 194 CLR 355 Rudolphy v Lightfoot (1999) 197 CLR 500

Law in Context, Vol 37, Issue 1, 2020 65 Received: June 5 May, Date of acceptance: September 25, 2020, Date of publication: November 30 2020, DOI: https://doi.org/10.26826/law-in-context.v37i1.127

Human Rights Issues in Constitutional Courts: Why Amici Curiae are Important in the U.S., and What Australia Can Learn from the U.S. Experience

By H. W. Perry Jr1, University Distinguished Teaching Professor, Associate Professor of Law, Associate Professor of Government at The University of Texas at Austin, Orcid: https://orcid.org/0000-0003-2947-8668 and Patrick Keyzer2, Research Professor of Law and Public Policy at La Trobe University, Orcid: https://orcid.org/0000-0003-0807-8366 1 The University of Texas at Austin, USA 2 La Trobe University

ABSTRACT Unlike thirty years ago, human rights issues are now routinely raised in Australian constitutional cases. In this article, the authors examine the role of the amicus curiae in the United States Supreme Court and consider how far and to what extent the amicus curiae device has been accepted in decisions of the High Court of Australia. The authors analyse the High Court’s treatment of applications for admissions as amici curiae, noting the divergent approaches - plicants who seek admission to make oral submissions. Human rights cases raise questions of minority rights that shouldtaken by not Chief be adjudicated Justice Brennan without and input Justice from Kirby, those and minorities. drawing attentionThe authors to therecommend practical thatdifficulties Australia faced adopt by apthe U.S. approach, to admit written submissions as a matter of course, and to allow applicants to make oral submissions

establishing legal policy norms for the entire nation, including for the identity groups that increasingly occupy the Court’swhen they attention. have a seriousThe focus and here arguable is on pointAustralia, to make. but theThis argument approach for is consistent the role of with amici the is Court’s more general significant and rolemight of well apply to high courts elsewhere.

Keywords – Amicus curiae, constitutional law, human rights, procedural law

Disclosure statement – No potential conflict of interest was reported by the authors. License – This work is under Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) https://creativecommons.org/licenses/by-nc-sa/4.0/ Suggested citation: Perry Jr, H.W. and Keyzer, P., “Human Rights Issues in Constitutional Courts: Why Amici Curiae are Important in the U.S., and What Australia Can Learn from the U.S. Experience” Law in Context, 37 (1): 66-98. DOI: https://doi.org/10.26826/law-in-context.v37i1.127 Summary 1. Introduction 2. Human rights and legal procedures 3. Amici curiae in constitutional cases in the Supreme Court of the United States 4. Amici curiae in constitutional cases in the High Court of Australia 4.1 Introduction 4.2 The perfectly general role of amici curiae at common law 4.3 The development of implied rights and freedoms in Australian constitutional law: developments that have made amici curiae needed 4.4 The enunciation of distinctive tests in Levy v Victoria 4.5 Justice Kirby’s broader approach 5. Conclusion

66 Law in Context, Vol 37, Issue 1; 2020 ISSN: 1839-4183

6. INTRODUCTION freedom to discuss political and governmental affairs. There is hardly any political question in the United These developments have been chronicled extensively by Australian constitutional scholars. Indeed, there is States that sooner or later does not turn into a judicial such a large volume of academic literature about this question. jurisprudence that it would be too time-consuming to Alexis de Tocqueville, Democracy in America, 1835 list all of the articles about these cases. But the jurisprudence, which stems from the High Court’s recognition What was considered a noteworthy observation in in Australian Capital Television Pty Ltd v Commonwealth the 19th Century seems trite today not only because it is (1992) 177 CLR 106 of the implied freedom to discuss so often quoted, but also because even the most casual political and governmental affairs, and its recognition in observer of the U.S. is aware of the important role played Kable v Director of Public Prosecutions (NSW) (1996) 189 by the Supreme Court in addressing social and political CLR 51 of the implication arising from Chapter III of the issues. The Court rules on a wide range of human rights Constitution of the autonomy and integrity of Australian courts, has spawned many, many cases, and indeed, has care, death with dignity, environmental protection, race transformed Australian constitutional law. The Australian andissues: other abortion, forms affirmative of discrimination action, gay to namemarriage, but healtha few. implied rights and freedoms jurisprudence has provided opportunities for the development and expansion of telling to see how constitutional interpretation is front human rights as an aspect of constitutional law (French andOnly center some havein the specific midst ofConstitutional a presidential guarantees. election, espe It is- 2019; O’Neill 1987). In addition, about forty years ago the cially as it relates to these social and human rights issues. In a recent interview with the New York Times, when power to use its power over “external affairs” to imple- mentHigh Court international confirmed treaty that obligations,the Commonwealth including has human wide Democratic presidential hopeful Vice-President Joe Biden rights obligations, into domestic law (Commonwealth v responded,asked “What “They are you have looking to have for an in expansive the Supreme view Court?” of the Tasmania (the Tasmanian Dam Case) (1983) 158 CLR 1). Constitution. Recognize the (implied) right, to privacy, These Commonwealth enactments can override State to unenumerated rights that exist in the Constitution" laws to the extent of their inconsistency, under s 109 of (Editorial Board of the New York Times, “Joe Biden”, 17 the Australian Constitution. This can mean that s 109 January 2020). On the other hand, President Donald cases raise questions about the ambit of Federal protec- Trump’s nominee to replace the late Justice Ruth Bader tions of human rights (Mabo v Queensland (1988) 166 Ginsburg, Judge Amy Coney Barrett, is a constitutional CLR 186). Finally, several Australian jurisdictions (the originalist who clerked for the late Justice Antonin Scalia Australian Capital Territory, Victoria and Queensland) and envisions a quite different perspective of what rights have implemented statutory charters of rights along the and whose rights are to be protected. However ask U.S. United Kingdom and New Zealand “dialogue” model (see students why we need the U.S. Supreme Court to have the e.g. Momcilovic v The Queen (2011) 245 CLR 1). These power of judicial review, and no matter their politics or developments have also increased the likelihood that the High Court will be called upon to consider human rights response is inevitably "to protect our rights." questions raised in disputes under these statutes. theoriesToday about an Australian Constitutional law student interpretation, might say their the same first The net result of all of these developments is that thing. But not thirty years ago. The Australian Constitution today, human rights issues are now commonly raised in does not contain a Bill of Rights, so historically, the types Australian constitutional cases. So Australian and U.S. of human rights issues that have featured prominently jurisprudence in Australia’s ultimate, appellate courts, is in U.S. constitutional jurisprudence have not featured more similar than it once was. But can Australian lawyers prominently in Australia. However in the last thirty learn anything from US lawyers about how litigation is years, the High Court has decided that the separation of judicial power effected by Chapter III of the Austral- ian Constitution requires that Australian courts must weutilized will focus in order on theto advance role of amici human curiae rights? in constitutional act and be seen to have independence, autonomy, and litigationWe think raising the answer human is rights yes. Specifically, issues. After in summarizthis article- integrity, and the creation of a system of representative ing the position in the United States, we will review the government by that Constitution gives rise to an implied position in Australia and then offer some thoughts about

Law in Context, Vol 37, Issue 1, 2020 67 ISSN: 1839-4183 what steps could be considered to enlarge opportunities what they can tell us about the Australian approach to for the involvement of amici curiae in Australian litigation. amici curiae. We also believe, however, that the role that amici can and should play as it relates to human rights ultimate courts play in establishing not only legal norms and other broad societal concerns is more general and butIn particular, individual we and argue collective that the values, significant requires role the that type our of goes beyond Australia. Focus on Australia, however, seems broader input that can be provided by amici curiae. As Martha Minow has argued, “[i]magining the range of peo- especially appropriate and timely. We are emboldened ple who share the future requires some actual knowledge in our choice of topic because the Australian High Court of people, and ideally, actual conversations with them” has recognised that U.S. precedents are directly relevant (Minow 2003, p.157). Or as Justice Kirby put it in Levy v Victoria, “I would have allowed them a voice” (Levy v Vic- of judicial power, judicial independence and judicial re- toria (1997) 189 CLR 579, p.652). In this article we argue viewand persuasive (French 2017; when Plaintiff it comes S157/2002 to the significant v Commonwealth topics that the High Court of Australia should hear the voices (2003) 211 CLR 476; Austin v Commonwealth (2003) 215 of people who have traditionally been denied standing, CLR 185, p.262). We believe our work might strengthen who were unable to access court due to the prohibitive costs of doing so, and, most importantly, have seriously arguments in the future about U.S. judicial procedures arguable submissions to make about how the law can being accepted in Australia. negatively impact the human rights of minority groups. A note about terminology. In the course of this article 7. HUMAN RIGHTS AND LEGAL PROCEDURES we use the expressions “minority groups” and “minority The role of procedure in the evolution and activity of rights”. We use these phrases to refer to people and the political institutions has been little heeded by political organisations who represent them to seek admission to scientists…the formalities and modes of doing business, court as amici curiae to ensure recognition of their hu- which we characterize as procedure, though lacking in man rights in circumstances where it has been argued or dramatic manifestations, may, like the subtle creeping is seriously arguable that the effect of legislation at issue of the tide, be a powerful force in dynamic process of in a constitutional case is to override those human rights government… (Keyzer 2010, pp.72-5). The story of…momentous political and economic Before we set off, though, we need to acknowledge issues lies concealed beneath the surface technicalities the usual disclaimers about comparative work. It is governing the jurisdiction of the Federal Courts. challenging. It always faces the issue of how we make Professor Felix Frankfurter (Frankfurter and Landis 1928) meaningful comparisons between different societies and regimes without either being too facile or suffering We agree with Professor (later Justice) Frankfurter’s from such attention to the particular that comparison is assessment. Legal rules and procedures play a major role stopped dead in its tracks. There is always a danger that in how courts come to decide human rights issues or not. a project of comparing access to courts and the role of Obviously the most basic way human rights issues get amici will be too simplistic if we do not take seriously the before the Supreme Court of the United States and the High Court of Australia, is if they are raised by litigants detailed understanding of legal rules, procedures, and with standing. (Much of what we say in this portion of significant differences in legal regimes, which requires technicalities. As such, we delve into some of the nitty the article applies to all federal courts in the U.S. and gritty, but we believe comparative work needs to be done Australia, but our focus in this article is on the ultimate because of the generally increasing importance of courts courts). Another way to get an issue before the courts is by worldwide in addressing human rights issues. intervention. A third way is by being admitted as an amicus This article focuses on Australia, and we hope that curiae. Before getting to the amicus device, the primary Australian lawyers and scholars will see some value in understanding how U.S. procedural rules operate, and focus of this article, we survey the other methods first.

68 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

Tocqueville may ultimately be right that in the U.S. Treaties,” constitutional limitations often exist because virtually every political issue will eventually become a judicial interpretations have said that they do. What is and judicial one, but it is not as easy for issues to get to the is not constitutionally required is the subject of extensive Supreme Court as one might assume. Procedural barriers debate among the justices and scholars. shape how access works, and, in turn, what outcomes are One of the most complex courses in any U.S. law school possible. In the United States, there are many procedural is “Fed Courts” which, at base, wrestles with jurisdiction, hurdles for litigants seeking to get a case into federal court justiciability, and procedure. The complexity prohibits (Perry 2010; Chemerinsky 2006). For those unfamiliar us from much elaboration on procedural hurdles in this with the U.S. system we go over some of the basics. - There are only 94 federal district trial courts in the U.S. and only 13 courts of appeal. State courts can also rule court,article. especially Suffice it tothe say Supreme that limitations Court. Certain such ascategories justicia on federal questions, but as a general matter, a case must ofbility cases, can particularlymake it difficult within to getthe issues political into and any societal federal make it all the way through the state’s highest appellate realm, are particularly vulnerable to being blocked from court and then hope to be one of the very few cases that judicial resolution because of jurisprudential barriers. make it into the U.S. Supreme Court—and only then would One notable category is cases dealing with the environ- the outcome serve as precedent for the nation. For a case ment (see, Juliana v United States 9th Circuit (2020) to to be decided in a federal court, which includes the U.S. be published; Sierra Club v Morton 405 US 727 (1972); Supreme Court, the court must have jurisdiction to hear Lujan v Defenders of Wildlife 504 US 555 (1992)), but the case. The basic contours for that jurisdiction are set even cases involving frequently adjudicated areas such out in Article III Sec. 2 of the Constitution. Essentially, the as racial justice get blocked by jurisprudential barriers case must present a “justiciable federal question.” Despite as well (Allen v Wright 466 US 737 (1984) or City of Los the general openness of American courts to considering Angeles v Lyons 431 US 95, 115 (1983)). human rights issues, justiciability issues often prevent One of the biggest justiciability hurdles is standing, important issues from being resolved by the federal and standing doctrine is a prime example of complexity courts. The main obstacles are doctrines fashioned by and dispute. The Court’s criteria for standing are often the Supreme Court. The doctrines include: prohibitions unclear and seem malleable when applied. (The same is against advisory opinions, standing, ripeness, mootness, true for other justiciability issues such as ripeness and the political question doctrine, and the constitutional mootness.) One need not take our word for it. The Court avoidance canon. (Similar doctrines operate in Australia itself has acknowledged the problem: – so, for example, Australia has an equivalent to the con- "We need not mince words when we say that the stitutional avoidance canon of the US (Attorney-General (NSW); Ex rel Tooth & Co v Brewery Employés Union (the with complete consistency in all of the various cases Union Label Case) (1908) 6 CLR 469). decidedconcept byof this‘Art. Court III standing’ which have has discussednot been definedit." The Supreme Court has distinguished between consti- Valley Forge Christian College v Americans United for tutional and prudential requirements for justiciability. If Separation of Church and State, 454 US 464, p. 475 (1982). the source of the limitation is the Constitution, neither Congress nor the Supreme can alter it; if it is a prudential Current doctrine holds that the Constitution requires doctrine, Congress can override it as can the Court. The that a plaintiff allege to have suffered or imminently will distinction between constitutional and prudential seems suffer an injury, that the injury is traceable to the defen- clear enough, but in reality, it is far more complicated. It dant’s conduct, and that a federal court decision is likely depends, of course, on the Court’s interpretation of Article to redress the injury. The prudential reasons enunciated III. Beyond some basic things such as the requirement of by the Court are related to the proper role and functioning an actual “case or controversy” or the case arising under of courts especially vis-a-vis other political institutions the “Constitution, the Laws of the United States, and

Law in Context, Vol 37, Issue 1, 2020 69 ISSN: 1839-4183 particularly in relation to appropriate separation of pow- to advance human rights issues in the High Court is worthy ers (Warth v. Seldin, 422 US 498 (1975)). of close study. Having reviewed the procedural context The rules governing standing in Australia are in some in each country, we can now consider the topic of amici ways similar to the U.S., but as a general matter, they are curiae. less formidable. In Australian constitutional cases, a person or organization will have standing to bring a case if 8. AMICI CURIAE IN CONSTITUTIONAL CASES IN they have a “special interest” (Croome v Tasmania (1997) THE SUPREME COURT OF THE UNITED STATES It is well-known that the expression “amicus curiae” is Latin for “friend of the court” (Keyzer 2010, pp 5, 19- or191 pecuniary CLR 119; interest, Evans 2011) or an or interest a “sufficient that is interest” more than in a 20, 65-7; Wiggins 1976). In the English common law merelymatter. intellectualThe test can or be emotional satisfied interest,by having such a proprietary as a desire preceding the independence of the United States, amici to see the law decided in a particular way (Australian curiae offered submissions on matters of fact or law to Conservation Foundation v Commonwealth (1980) 146 assist the courts to avoid errors of fact or law (Lilburne’s CLR 493, pp.508, 548). More recently, in Kuczborski v Case (1649) 4 State Tr 1270; The Prince’s Case (1606) 8 Queensland, Brennan, Kiefel, Gageler and Keane JJ said Coke 1 29a; Krislov 1963). Over time and particularly that a “person whose freedom of action is challenged can in the 20th century, the role of amici curiae expanded. always come to a court to have his rights and position They have ceased being merely independent sources of - knowledge and have increasingly asserted themselves as john in Pharmaceutical Society of Great Britain v Dickson advocates for a particular result (Krislov, p.708; Barker [1970]clarified” AC ((2014) 403, p.433). 254 CLR This 51, is 106liberal [175] standing, citing Lord but itUp is 1967; Caldeira and Wright 1988). not open standing. The more important thing, from the standpoint of in the U.S. Supreme Court (Kearney and Merrill 2000; the litigant, is the cost of a loss. Unlike the United States, OrrThe and amicus Devins curiae 2016). device Research plays aon highly the role significant of amici role is the loser in an Australian constitutional case ordinarily extensive in both political science and legal scholarship. pays the winner’s costs. This means that in some cases a Space restricts our ability even to begin to cite all who government will simply oppose the case on the ground of should be cited. The Orr and Devins article is a good place standing, placing the would-be constitutional litigant in to get a sense of the breadth and depth of that research. a position where they have to invest precious resources As with the early discussion, our focus here will be on that they do not have in contesting that issue without even the U.S. Supreme Court. But amici participate in lower getting to the substantive points. Then there is the chill- courts as well. Many of the same considerations would apply except where we discuss the role of amici in the cert. of a loss. This casts a mighty shadow over the would-be process. For a discussion of amici in the Circuit Courts of ing effect of the potentially catastrophic financial impact constitutional litigant (Keyzer 2010a). Appeal, see Collins and Martinek 2011. Indeed, it would To sum up to this point: in both jurisdictions, not be going too far to say that despite its peripheral role courts avoid addressing constitutional questions where within constitutional litigation per se, the amicus brief has there is a non-constitutional pathway to the resolution of become one of the “basics” of U.S. constitutional adjudica- the matter. Second, standing rules are problematic in both tion. They often play an important role in what cases are the U.S. and in Australia (though for different reasons). selected for review, and they affect case outcomes and Third, a key difference between the two jurisdictions is doctrine. In fact, the U.S. Solicitor General now appears more frequently in the U.S. Supreme Court as an amicus bringing a constitutional case in Australia. This means than as a litigant (Perry 2020; Cordray and Cordray 2010; thatcosts. any The procedural loser pays device rule is thata significant provides disincentive an Australian to Nicholson and Collins 2008). Amici have become such a person or interest group with a cost-effective opportunity common feature of U.S. constitutional litigation that it

70 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

The expectation is, however, that this Court will have that interest prominently-indeed, primarily-in is difficult to imagine constitutional law without them. The U.S. Supreme Court’s awareness of complexity of mind. Today we have failed that expectation, and Their influence is evident, and powerful (Collins 2004). human rights issues is in large measure due to the admis- that responsibility. sion of submissions from amici curiae. Justices, of course, Nevertheless, amici are generally welcomed. Unlike the are not monks removed from society, but the barrage of many barriers faced by litigants, the Court has essentially legal briefs asking the Court to be involved in these issues an open-door policy when it comes to amicus briefs. The plays a very important role. Amici not only bring societal formal rule, Rule 37 of the Supreme Court Rules, urges issues to their attention in ways that are not tied to the restraint and sets some conditions, but in practice, the justices recognize and address the human rights issues in amicus curiae. Of course, that does not mean that all briefs playspecific in constitutional facts of the case cases being that litigated, might not but be also so they obvious. help areCourt read accepts or are virtually taken seriously, every motion but there for leave is effectively to file as noan As mentioned earlier, the modern role of the amicus differs from its traditional function. Sam Krislov labeled - internal dam that stops the flow; and quite a flow there is. the phenomenon “from friendship to advocacy” (Krislov icus brief in the Supreme Court. The most well-known is 1963). Or as Caldeira and Wright said (1990): a briefTypically, on the there merits are after two a opportunities case has been to accepted file an am for …Contemporary amici are really friends of the a decision. Prior to that, however, when review is being parties, not necessarily friends of the Court, even sought, proponents and opponents can also submit an though the original intent of amicus curiae briefs amicus brief supporting or opposing a grant of certiorari. was, of course, to provide the Court with new in- Once a case is accepted for a decision on the merits, formation and to act in a neutral fashion. Political scientists often refer to amici as lobbyists. The the amici filings rush in. In October Term (hereafter OT) term is perhaps an unfortunate one—as the old saw goes, 2014, 781 briefs on the merits were filed. Over 98% of what parent wants his or her child to grow up to be a lob- one case had none (Franze and Anderson 2015). In OT cases had an amicus curiae brief filed on the merits; only linkage groups in democratic societies—two of the most cases that had been selected for decision. Neither term 2017 over 800 amicus briefs were filed on the merits for prominentbyist? Yet political being scientistspolitical parties often point and to interest the necessity groups. of was an aberration for modern times, but the number has These “linkage groups” provide a crucial link between the been growing rapidly over the last several years as the public and those who govern. In our opinion, amici serve number of argued cases has remained about the same. this function in the legal realm. However, this “lobbyist” role has, in some cases, created tensions (Wohl 1996; In OT 1995, 400 amicus briefs were filed (Collins and Susman 2006). In a dissent in Jaffe v. Redmond, in which the 1990s was about 5 briefs per case (Collins and Mc- Solowiej 2007). The average number of briefs filed in there were 14 briefs on one side and none on the other, Carthy 2017). The average number for OT’s 2012-2014 Justice Scalia complained (518 U.S. (1996) 1, pp.35-36): was 13 per case. In some blockbuster cases, the number can reach into the 100’s. In the “Obama Care” cases there 14 amicus briefs supporting respondents, most of were 136 amicus briefs (Franze and Anderson 2015) and which… the came Court from was such the organizations beneficiary of as no the fewer American than in the same sex marriage case, Nina Totenberg reported Psychiatric Association, the American Psychoana- on National Public Radio that there was an all-time high lytic Association, the American Association of State of 148 (Franze and Anderson 2015). Social Work Boards…. Not a single amicus brief was The numbers are impressive, but this begs the issue whether or not amicus briefs have much of an effect. One There is no self-interested organization out there way to document effectiveness is by citations to amicus devotedfiled in support to pursuit of of petitioner. the truth inThat the isfederal no surprise. courts. briefs. In OT 2010-2014, citations to amicus briefs ranged

Law in Context, Vol 37, Issue 1, 2020 71 ISSN: 1839-4183 from 53-63 (Franze and Anderson 2015). The individual are not theoretical but real, as major American justices vary in how much they cite to amici in their opin- businesses have made clear [in their amicus brief ions, but they all do it. They ranged from Justice Sotomayor, in support of the university] that the skills needed in today’s increasingly global marketplace can only averaged 22% (Franze and Anderson 2015). In one term, be developed through exposure to widely diverse Justicewho averaged Breyer cited45% overamici five in 63% terms, of histo Justice opinions Scalia (Franze who people, cultures, ideas, and viewpoints. What is and Anderson 2015). leaders of the United States military assert [in their Obviously, all amici are not created equal. Special amicusmore, high brief] ranking that, “[b]ased retired onofficers [their] and decades civilian of attention is always given to a brief by the U.S. Solicitor- brief even if it is not a right. Indeed, the Court often calls experience,” a “highly qualified, racially diverse forGeneral the views (S.G.) of The the S.G. Solicitor-General is always allowed even to filewhen an theamicus S.G. security.”officer corps [At] . present,. . is essential “the military to the military’s cannot achieve ability has not submitted a brief. We have chosen not to dwell to fulfill its principle mission to provide national on the role of the Solicitor-General because, in Australia, racially diverse unless the service academies and Solicitors-General from every jurisdiction have a statu- thean officer ROTC usedcorps limited that is race-consciousboth highly qualified recruiting and tory right to intervene, which means that an application and admissions policies.” for admission as amicus curiae is not needed (Judiciary Act 1903 (Commonwealth of Australia), ss 78A and 78AA. A phenomenon that is worthy of note is that of “repeat players” (McGuire 1995). Some organizations have become Another statistic to demonstrate the importance of amicus briefs is when justices mentioned them in oral briefs (Songer, Kuersten and Kaheny 2000). Their knowl- argument. In the 2009-2014 Terms, the total number of edgevery experiencedand experience at developingis valuable to and interest filing groupsamicus hoping curiae oral arguments in which a justice mentions an amicus brief to be admitted and have their briefs read. In some cases, was 67 (Larsen and Devins 2016). Even more important amicus briefs from repeat players have been remarkably than citations is evidence when a particular amicus brief is referenced prominently in an opinion. Justices often lift the Solicitor-General, the quint essential repeat player, passages directly from amicus briefs. There is a particularly hasinfluential always been(Smith very and important Terrell 1995). in Court As decisions. stated above, What good and oft-cited example of the potential importance of is relatively new is that an elite group of private lawyers an amicus brief. Grutter v. Bollinger 539 U.S. 306 (2003) was the very contentious and closely divided case about that garner special attention. One of us has written about the use of racial criteria in university admission policies. thisand (Perrylaw firms 2020). has Othersemerged have who documented routinely submitthis phenom briefs- enon as well (McGuire 1994; Lazarus 2008). Some clever names have emerged: “Amicus Wranglers” or the “Amicus surpriseMany predicted of many, that Justice the CourtO’Connor would wrote put thethe 5-4final major stake- Machine” (Larsen and Devins 2016). Once a case is ac- ityin theopinion heart upholding of affirmative the use action of race with in certain this case. instances. To the cepted in the Supreme Court, there is an extensive effort She quoted extensively from some amicus briefs in doing to recruit prominent attorneys to write amicus briefs and so. Rumor has it that a brief from an unexpected source- to recruit all sorts of organized groups to do so as well. -the military-- played an especially important role in her decision. In her opinion she said (Grutter v. Bollinger 539 Another group has also emerged as important amici U.S. 306 (2003)): players—state solicitors general. This is a relatively new phenomenon. U.S. states traditionally did not have very [Student] body diversity promotes learning out- good appellate lawyers, which was a frustration to the comes and “better prepares students for an increas- U.S. Supreme Court. Increasingly, states are adopting a ingly diverse workforce and society, and “better solicitor-general model, and many of those state SG’s

prepares them as professionals.” These benefits are top flight appellate lawyers (Perry 2020a). Unlike in

72 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

Australia, states in the U.S. do not have a general right of the periphery until relatively recently (Willheim 2010). intervention in constitutional cases, though they can be Things have changed. We now chronicle the history and admitted as intervenors at the discretion of the court. the change. Traditionally, state attorneys general banded together It was not until 1997 that members of the High Court provided clear reasons in support of a decision to about questions of state power. They also banded together admit amici curiae in constitutional cases. In Levy v Victoria, when they were fighting with the federal government when they shared a cause such as litigation with tobacco the High Court was invited to consider the constitutional companies (Nolette 2015; Lemos and Young 2018). More validity of regulations made by the Parliament of the State recently, however, they have been banding together to of Victoria that prohibited people from entering an area write amicus briefs over a wider range of social issues. This where duck shooting was taking place unless they had a permit ((1997) 189 CLR 579). Laurie Levy was an animal association, and now the Democratic A.G.s have done the rights protestor who sought a declaration in the High accelerated when, first, Republican state A.G.’s formed an Court that the Victorian regulations were constitutionally their interests as states, per se. They are often about is- invalid on the ground that they infringed his (implied) same. The amici that they file often have little to do with sues such as abortion, gay rights, or religion (Perry 2020). freedom to discuss political and governmental affairs. Much the same can be said with regard to amici in the Levy and other protestors had, in previous years, sought cert process. To a certain extent, it makes some sense to rescue ducks that had been shot by duck hunters, and to wait until a case has been selected before going into they courted the attention of local television stations to the expensive process of “amicus wrangling,” because so draw attention to the cruelty of hunting. Their objective very few cases are accepted. But as was said earlier, the was to draw attention to that cruelty, as part of a campaign agenda setting stage is crucial. Amicus wranglers are in- to ban duck hunting. The case, however, concerned sig- creasingly at work at the cert stage, be they from private

nificant questions about freedom to protest, in addition Is free speech impinged if the protestors cannot get Certfirms was or stategranted SGs. in Returning31 of the cases to OT that 2014, had 403 an amicus to the significant issue of animal cruelty. briefs were filed urging or opposing cert on 177 cases. court considering a case such as this not only would have a high percentage given the overall grant rate of fewer than richsufficiently history close of First to the Amendment situation they jurisprudence, are protesting? but alsoA U.S. a brief filed. That is almost an 18% grant rate, which is a 5% for paid cases (Feldman 2014). court would know that its ruling could have wide ranging Beginning where we started, it is hard to imagine the consequences. As we write this article, in the context of appellate process in the U.S. without amici. Historically, a U.S. national election and nationwide protests in sup- though, the contrast with Australia was stark. port of the Black Lives Matter movement, the case would obviously be important for the whole of society. Before 9. AMICUS CURIAE IN CONSTITUTIONAL CASES IN ruling, it can be conjectured that a number of Justices of THE HIGH COURT OF AUSTRALIA the U.S. Supreme Court might want to hear a variety of voices to help them think through the implications of their decision and any possibly unforeseen consequences of 4.1 INTRODUCTION resolving the matters consistently with the submissions The times they are a-changin’ of the parties only. As we have demonstrated above, amici Bob Dylan, 1964 curiae participating in U.S. constitutional cases raising Having quoted Tocqueville and Frankfurter earlier, we human rights issues would have clear procedures in place have just quoted another eminent observer of politics. In governing their involvement, and their involvement would contrast to the United States, the Australian High Court be extensive with widely varying points of view. throughout its history has admitted amicus curiae relatively rarely (Keyzer 2010). Amici curiae remained on

Law in Context, Vol 37, Issue 1, 2020 73 ISSN: 1839-4183

When Levy arose, however, the High Court of Australia be an array of interveners involved who are opposed to did not have a clear test governing whether amici curiae the further development of implied (human) rights and could be admitted in constitutional cases raising human freedoms, and a relative absence of voices of minorities. rights issues. Indeed, it was still grappling with the question This means that the Court is presented with a lopsided whether there was an implied freedom of communication set of arguments in these cases. We argue that this result is inappropriate, and that procedural rules governing the such important issues is typical in constitutional cases admission of amici curiae need to be changed in order to andin the seems Constitution to us to incry the out first for place.broader Yet input wrestling than withthat enable the input that the courts need to undertake the provided by parties. new role that implied rights and freedoms jurisprudence In the remainder of this article, we analyse the Austra- contemplates. lian law governing the admission of amici curiae. Firstly, In the 1960s, the Warren Court admitted the American we note that at common law, amici curiae were admitted Civil Liberties Union and the National Association for the in courts to perform many different roles. For example, Advancement of Colored People, among other organisa- one ground of appearance was to object to the jurisdiction tions, to make submissions as amici curiae in civil rights of the court (as in the Railways Servants case, considered cases (Barker 1967). Over time, as we set out above, the further below). This ground of appearance could provide “amicus brief” became a way for diverse associations of an open door for amici curiae in constitutional cases, people across the political spectrum to offer their per- who could argue that they offer a different or unique spectives on the proper resolution of cases before the perspective on whether an institution had exceeded its Supreme Court. Often these briefs and cases involved jurisdiction for constitutional reasons. human rights issues. From time to time, judges have la- In the High Court, an initially broad approach to the mented that parties are scarcely in control of litigation, admission of amici curiae appears to have been replaced by the rule, which was also part of the common law in feature of U.S. jurisprudence. but the advancing influence of amici curiae is a notable the early 1900s, that Attorneys-General were the proper We contend that once human rights are at stake, representative of the public in any public law case and amici curiae who have a serious, arguable submission should have a presumptive right of intervention. While this to make in a constitutional matter raising human rights right of intervention was narrowed in 1930, it has been issues should be admitted to do so. We do not argue that authorised by statute and without apparent limits for over a matter should be taken out of the hands of the parties, forty years. Throughout Australia’s federal constitutional but once it can be seriously contended that a case raises history the Attorneys, represented by their Solicitors- human rights issues, that legal matter is no longer just General, have occupied a large role in the development of about the parties. If it can be seriously argued that the constitutional law. However, we argue that in the last thirty decision in question will or may affect human rights, then years, since the High Court has recognised implied rights adjudicative procedures should enable the admission of and freedoms arising from the Constitution, the limitations amici curiae to provide the court with input relevant to of relying on the Attorneys-General to represent minority the human rights issues. interests has become apparent. Attorneys-General can We also contend that, for the reasons that follow, we represent majorities, and are well-funded to do so, and argue that neither the test for amicus curiae admission enjoy special privileges when it comes to costs (costs are developed by Chief Justice Brennan nor the one developed not awarded against Attorneys-General who intervene in by Justice Kirby in Levy v Victoria address the unique constitutional cases). But why would Attorneys-General role that an amicus curiae can perform in constitutional intervene in constitutional cases to support the develop- litigation raising human rights issues. ment of new constitutional principles that would limit their In our respectful opinion, the “Brennan test”, which has since been adopted by the Court, places too much weight means that in implied rights and freedoms cases there can legislative power? We demonstrate that they do not. This

74 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 on the role of the parties in constitutional cases raising “[p]articipants in the legal system need to refocus their efforts on responsibility and inclusion. From within a curiae negatively, by reference to submissions that might rhetorical community, we are entitled to ask, ‘What place behuman raised rights by amici issues. after The the test parties defines have the said role what of amici they want to say. Of course it is accepted that the jurisdiction of the court is derived from the judicial remedy that is is there for me, and for others, in the universe defined by this discourse, in the community created by this text? sought in a constitutional case (Kuczborski v Queensland We argue below that the Kirby test recognises these What world does it assume? What world does it create’.” (2014) 254 CLR 51, pp. 59-62). However once parties important functions of the High Court. Justice Kirby also with standing have brought a properly constituted case recognises the inconvenience that the current approach or controversy (what Australians call a “matter”) to the poses to would-be amici curiae. However, we have two High Court (Constitution, Chapter III; In re Judiciary and Navigation Acts (1921) 29 CLR 257), whether in its original Kirby. In his judgments on the admission of amici curiae, jurisdiction or by way of an appeal (section 73), the role of Justicesignificant Kirby quibbles is critical with of “busybodies” the approach and taken contemplates by Justice amici curiae is concerned with something different—not the use of adverse costs orders to prevent abuse of the the resolution of the matter, per se (although amici curiae device. We argue below that these methods for limiting are likely to prefer a particular outcome) — but, rather, possible abuse of the amicus curiae device are unhelpful, adequate consideration of the human rights issues that as they tend to root the device within a “winners and los- may be raised or implicated by that dispute. This different ers” paradigm of litigation which is inappropriate given role is less related to the resolution of the legal dispute, the different role of amici curiae. and more related to the Court’s role in making national For the reasons we develop below, we argue that the and ultimate decisions of public importance. It could be High Court should open its door to amici curiae in consti- argued that when human rights are concerned, everyone tutional cases in the High Court, just as the U.S. Supreme has an interest. However, for obvious reasons, the court Court has done, to ensure that its constitutional juris- cannot hear submissions from everyone. We argue below prudence is properly cognisant of the diverse interests at that Attorneys-General should not necessarily be seen play in these cases. What is needed is an approach that: as proxies for the public interest. But if, in constitutional recognises that human rights cases raise questions that cases raising human rights issues, the parties make may require wider input than that which can be provided submissions focusing on resolving the dispute, and the by the parties, and that the proper source of constraint Attorneys (who have a statutory right to intervene) are on use of the device should not be the parties, or inapt characterised as having the function of representing the concepts like “busybodies” or inappropriate rules govern- majority of the relevant public, then amici curiae can be ing costs, but instead, any constraint should be based on admitted to represent minority interests. It is trite to point the issues, and whether the contentions of the amici are out that the High Court’s role is broader than the role of seriously arguable. inferior courts in the Australian judicial system. We argue that as constitutional cases raising human rights issues 4.2 THE PERFECTLY GENERAL ROLE OF AMICI CURIAE AT involve broader value judgments and questions of legal COMMON LAW policy that travel beyond the mere resolution of a legal dispute (Mason 1986), there should be mechanisms that At common law, amici curiae played a variety of different enable the Court to receive input as to what those values roles. Traditionally, the courts admitted amici curiae as should be. If amici curiae have seriously arguable points volunteers or appointees to assist courts to avoid errors to make about the resolution of the human rights issues of fact or law. However in 1963, U.S. political scientist at stake, then their voices should be heard. Adopting the and Associate Professor Samuel Krislov published an words of Martha Minow and Elizabeth Spelman, and ap- important article in the Yale Law Journal that pointed out plying them in the present circumstances, we argue that: that in the common law there were also cases where an

Law in Context, Vol 37, Issue 1, 2020 75 ISSN: 1839-4183 amicus curiae drew the attention of a court to matters submissions advanced by the Federated Association in that neither party was willing to argue, for example, in aid of this argument did not raise constitutional issues. suits where the parties had fraudulently colluded (Krislov The New South Wales Government sought leave to intervene, over the objection of the Federated Amalgamated amicus curiae was no longer “a detached servant of the Government Railway and Tramway Service Association. 1963). The significance of this development was that the court” but, instead, the amicus curiae could “act directly They argued that the Commonwealth Court was limited the case” (Krislov pp 696-7). By the 1930s amicus briefs CJ, Barton and O’Connor JJ, the three original members and officially as counsel for one not formally a party to ofby thea constitutional Court, granted doctrine admission of “implied to New immunity”. South Wales Griffith and rather than maintaining the pretense of the independent accepted its arguments. The headnote states: were openly identified with their organizational sponsor, “lawyer-like” amicus curiae, which Krislov argues “realisti- The rule … that when a State attempts to give to its legislative or executive authority an operation, which, pattern of behavior and its new function. The amicus is cally embraces and ratifies the transformation of the actual if valid, would fetter, control, or interfere with the free no longer a neutral, amorphous embodiment of justice, exercise of the legislative or executive power of the Com- but an active participant in the interest group struggle” monwealth, the attempt, unless expressly authorized by (Krislov p.703). Krislov is not critical of this development, the Constitution, is to that extent invalid and inoperative, nor should he be. In U.S. constitutional law the device is reciprocal. It is equally true of attempted interference by the Commonwealth with State instrumentalities. In Australia the amicus curiae device came from the serves a significant role, as we said earlier in this article. The application of the rule is not limited to taxation. The New South Wales association argued that there was been recognised by the High Court as one that can provide no constitutional issue before the Court, that the matter same common law heritage. It is a flexible device that has additional input into the resolution of cases between par- could be determined on technical and non-constitutional ties. Early decisions of the High Court demonstrated its grounds, and that the Court should not determine a con- utility in constitutional cases. In 1906, only a few short stitutional question unless it was absolutely necessary years after the High Court started sitting, it held that an to do so. But the Court, which at this point in its history amicus curiae could appear to make an objection to juris- believed that the doctrine of implied immunity as sum- diction on constitutional grounds in the Railways Servants marized above should be part of Australian constitutional case (Federated Amalgamated Government Railway and law, allowed New South Wales to intervene to advance their Tramway Service Association v New South Wales Railway Traffic Employés Association (1906) 4 CLR 488). In this concurred and O’Connor J agreed, held that the interven- case the court had to determine whether New South Wales constitutional argument. Griffith CJ, with whom Barton J railway workers were subject to the jurisdiction of the that they objected to the jurisdiction of the court below, Commonwealth Court of Conciliation and Arbitration. The andtion suchof New an objectionSouth Wales “may could with bethe justified sanction on of the Courtbasis be made by any person, if only as amicus curiae” ((1906) registered their association with the Commonwealth Court New South Wales Railway Traffic Employés Association of Conciliation and Arbitration under federal legislation, and Barton J, the importance of an objection to jurisdic- 4tion CLR (on 488, constitutional p.495). Furthermore, grounds) occludedaccording any to Griffithobjection CJ Court upheld the registration. The Federated Amalgam- to the argument that constitutional points should only be and the President (the presiding judicial officer) of that ated Government Railway and Tramway Service Associa- determined when a constitutional point has been raised tion, a competing union, challenged the registration of by the parties themselves. the New South Wales association on the ground that the This decision has been considered in some detail be- Commonwealth Court of Conciliation and Arbitration cause, when the current position is reviewed later in this could not register a NSW association. Importantly, the article, it might seem remarkable today that the Court

76 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 had ever allowed an amicus curiae to raise a constitu- The ground of the Attorney-General's right to inter- tional point where the parties had not. It is all the more fere is a common injury to the public, and it appears remarkable when one considers the basis upon which the to me that, if a person claims to be authorized by law intervention as amicus curiae was made in the Railways to exercise some public function which is in fact not Servants case, viz., as an “objection to jurisdiction” on authorized by law and is injurious to the public of a State, the Attorney-General for the State may sue to ground for raising a constitutional argument. This objec- protect the rights of the public of the State. constitutional grounds. It is difficult to think of a wider tion could be deployed in any case in order to advance a The other members of the majority took a similar constitutional argument (where such an objection was approach (pp. 520, 533). Isaacs J, one of the judges who otherwise arguable, ie., not doomed to fail). dissented in the case, also agreed with the majority that Interestingly, the wide approach taken to amici curiae Attorneys-General had a special interest in federal cases in the Railway Servants case did not last very long. In the Union Label case in 1908 (Attorney-General (NSW); Ex rel standing rested on the notion that State Attorneys-General Tooth & Co v Brewery Employés Union (NSW) (the Union couldthat justified protect their their citizens standing from (pp.557-8). a form of Commonwealth However this Label case) (1908) 6 CLR 469), the Court held that a State legislative trespass (Australian Railways Union v Victo- Attorney-General could not only intervene in any case rian Railways Commissioner (1930) 44 CLR 319, p.330 where the jurisdiction of a Commonwealth body could be (Isaacs CJ)), if indeed the Commonwealth had exceeded challenged on constitutional grounds (here, whether the its jurisdiction. The use of the phrase “amicus curiae” in Federal Registrar of Trade Marks could register a trade the Railways Servants case was abandoned in the Union mark [or “union label”] on behalf of a State-registered Label case, and the new rule of standing for Attorneys- trade union), but could initiate such a constitutional mat- General appears to have closed down an argument that the “objection to jurisdiction on constitutional grounds” the argument put by the respondent trade union that the could be invoked by a future amicus curiae. ter on its own motion. Griffith CJ was plainly sensitive to Attorney-General had no real interest in the matter. Near Union Label case for the devel- the start of his judgment, the Chief Justice wrote (p.491): opment of the amicus curiae rule appears to be that after The first condition of any litigation in a Court of 1908,The Attorneys-General significance of the had their own special standing Justice is that there should be a competent plaintiff, in the High Court and a clear majority of the court would i.e., a person who has a direct material interest in the grant the Attorneys standing on the grounds set out in determination of the question sought to be decided. the majority judgments (Keyzer 2010, p.93, footnote 71). The Court will not decide abstract questions, nor will it Since Attorneys had their special rule, there was no need decide any question except when raised by some person for them to apply for admission as amici curiae; correla- entitled by reason of his interest to claim a decision. This tively, there was no need for them to advance arguments doctrine should certainly not be relaxed for the purpose that might have given the justices of the High Court an of bringing in question the validity of Statutes passed opportunity to develop its thinking about that device. either by the Commonwealth Parliament or by a State Notwithstanding this development, it was clear to the legislature. It is, therefore, material to consider and deal Court that amici curiae or interveners were needed in with this point before considering the substantial point some (constitutional) cases. While the rule was tightened of the validity … For, if the plaintiffs are not entitled in 1930 in Australian Railways Union v Victorian Railways to ask for a decision, any opinion of the Court given at Commissioner (1930) 44 CLR 319, it was rarely invoked to their instance would be extra-judicial. prevent a State from intervening. In 1976 it was replaced That said, the NSW Attorney-General had standing by a statutory right of intervention for the States, and in (pp.499-500): 1983, for the self-governing territories (Keyzer, 2010, pp. 91-5).

Law in Context, Vol 37, Issue 1, 2020 77 ISSN: 1839-4183

Perhaps the need was less apparent after it became DEVELOPMENTS THAT HAVE MADE AMICI CURIAE common practice for the Commonwealth and State NEEDED Solicitors-General (and later Territory Solicitors-General) It is well known that amici curiae emerged strongly to intervene in constitutional cases. Since the Australian in the Warren Court to advance human rights arguments Constitution is a federal Constitution with no Bill of Rights, in U.S. constitutional cases (Krislov, pp.711-3). When the many if not most constitutional cases raise federalism issues, Warren Court was active, some Australian constitutional not human rights issues. Obviously, the Commonwealth, lawyers were watching closely. One enthusiastic consumer States (and the Territories, once self-governing Territories of U.S. constitutional decisions was Justice Murphy, who emerged) have a direct interest in such disputes. sat on the High Court from 1975 to 1986. During this However once the High Court developed a jurisprudence period, His Honour delivered a number of judgments of implied rights and freedoms (interestingly, mirroring establishing the foundations of an implied freedom of their emergence in the Railways Servants case to offer the political communication. In a 1977 decision, Murphy J “implied immunity” argument that won the day then), it cited several decisions in the heyday of the Warren Court quickly became apparent that the human rights issues when he said (Ansett Transport Industries (Operations) Pty these cases raised would not necessarily be the subject of Ltd v Commonwealth (1977) 139 CLR 54, p.88): balanced submissions by the intervening Attorneys-General In my opinion the concept of the Commonwealth (via their Solicitors-General). At the outset, it needs to be and the freedom required for the proper operation borne in mind that their role is to represent the majorities of the legislative, executive and judicial branches in their polities (Keyzer 2010 Chapter 4; Keyzer 2014). This in the democratic society contemplated by the means that they cannot really be expected to represent Constitution necessitate the implication of such a the interests of minorities; in short, minority groups are guarantee (see Crandall v. State of Nevada (1867) not their clients. In earlier work (Keyzer 2010), one of us 6 Wall 35 (18 Law Ed 745); Slaughter-House Cases; (1872) 16 Wall 36 (21 Law Ed 394); R. v. Smithers; the Court was offered submissions by a Solicitor-General Ex parte Benson (1912) 16 CLR 99; Buck v. Bavone thatidentified favoured only thetwo expansion cases in a twenty of constitutional year period rights in which or (1976) 135 CLR, at p 136-137). freedoms (Kable v Director of Public Prosecutions (NSW) Elections of the federal Parliament provided for (1996) 189 CLR 51 and Kartinyeri v The Commonwealth in the Constitution require freedom of movement, (1998) 195 CLR 337); a period during which there were speech and other communication, not only be- many decisions in which submissions were made that did tween the States, but in and between every part not favour the expansion of those rights or freedoms. So of the Commonwealth. The proper operation of the system of representative government requires We will take a detour now to describe the emergence the same freedoms between elections. These are who can represent minority rights in these cases? of the High Court’s implied rights and freedoms jurispru- also necessary for the proper operation of the dence. This detour is vital for our international audience, Constitutions of the States (which now derive but also to underpin the argument we advance at the their authority from Ch. V of the Constitution. conclusion of this article that the public importance of From these provisions and from the concept of the High Court cases, particularly High Court cases that raise Commonwealth arises an implication of a consti- human rights issues, require the Court to take an open tutional guarantee of such freedoms, freedoms so approach to the admission of amici curiae. elementary that it was not necessary to mention them in the Constitution (see United States v. Guest (1965) 383 US 745, at pp 757-758 (16 Law Ed 239, 4.3 THE DEVELOPMENT OF IMPLIED RIGHTS AND at pp 248-249); Shapiro, Commissioner Welfare of FREEDOMS IN AUSTRALIAN CONSTITUTIONAL LAW: Connecticut v. Thompson (1968) 394 US 618, at pp 632-633 (22 Law Ed 2d 600, at p 614)). The

78 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

freedoms are not absolute, but nearly so. They to people who would produce commemorative souvenirs are subject to necessary regulation (for example, for sale. Davis, an Aboriginal political activist who was freedom of movement is subject to regulation for actively opposed to celebrating the Bicentenary, sought purposes of quarantine and criminal justice; free- and was refused permission by the ABA to print t-shirts dom of electronic media is subject to regulation that used these trademarked expressions within slogans to the extent made necessary by physical limits that protested the celebration of the beginning of the upon the number of stations which can operate invasion and conquest of Australia, and the destruction simultaneously). The freedoms may not be re- of Aboriginal people and culture that ensued. He chal- stricted by the Parliament or State Parliaments lenged the constitutional validity of the ABA, saying that except for such compelling reasons. there was no constitutional power that could be invoked We have set out the passages above in full to demonstrate to support its set-up, that therefore any appropriation of Commonwealth moneys to it was unconstitutional, and At the time, Justice Murphy’s judgments were regarded that the Commonwealth’s legislative power over trade- ashow radical influential and were U.S. criticizeddecisions bywere other on membersJustice Murphy. of the marks or indeed over any topic could “not authorize the court. His use of U.S. decisions to justify his conclusions prohibition of the use of common names”. was also regarded with disdain. In 1986, Sir Anthony Although the submissions advanced on behalf of Mr Mason, a senior member of the High Court at that time, Davis on this last point summarised in the Commonwealth regarded Murphy J’s judgment about the implied freedom Law Reports were remarkably brief, the Court unanimously of political communication to be tantamount to “implying struck down the provisions purporting to grant trademark a new s 92A into the Constitution” ((1986) 161 CLR 556, protection to the expression “200 years” in conjunction p. 579) (Section 92 protects freedom of trade, commerce with “1788” and “1988”, and on very wide grounds indeed. and intercourse (movement across State borders, subject Mason CJ, Deane and Gaudron JJ with whom Wilson and to reasonable limits) in Australia. There is no s 92A in the Dawson JJ, and Toohey J agreed on this point) said that Constitution). the regulation in question reached ((1988) 166 CLR 79, pp.100-1): the Commonwealth Law Reports later (and sadly, after far beyond the legitimate objects sought to be JusticeHowever, Murphy’s two death) short yearsthe High and Court, only ledfive by volumes now Chief of achieved and impinges on freedom of expression by Justice Mason, struck down a federal law on the basis that enabling the Authority to regulate the use of com- it infringed “freedom of expression”. The case was Davis v mon expressions and by making unauthorized use a Commonwealth (1988) 166 CLR 79). We cover it in a bit of criminal offence. Although the statutory regime may be related to a constitutionally legitimate end, the it is crucial to a genuine understanding of the new role provisions in question reach too far. This extraor- thatdetail the for High the benefit Court was of our starting international to build readers for itself because (a role dinary intrusion into freedom of expression is not that makes it more like the U.S. Supreme Court). reasonably and appropriately adapted to achieve the In 1980, the Commonwealth Parliament enacted ends that lie within the limits of constitutional power. the Australian Bicentennial Authority Act, creating an Brennan J agreed with the plurality that the relevant Australian Bicentennial Authority (ABA) which would provisions restricting use of the symbols and phrases prepare for the celebration of the 200th anniversary of were invalid, and also said (p. 116): the arrival of the “First Fleet” of British settlers in what The form of national commemorations of histori- later became Sydney on the 26th of January 1788. The ABA was granted exclusive use of particular trademarks, majority of people place upon the event. But there including the expression “200 years”. The idea was that the maycal events well be usually minority reflects views the which significance place a which different the ABA would license the use of this and other trademarks

significance on the same event, as the present case

Law in Context, Vol 37, Issue 1, 2020 79 ISSN: 1839-4183

illustrates. It is of the essence of a free and mature decisions in the late 1980s and early 1990s demonstrates nation that minorities are entitled to equality in that justices of the Court in this period listened carefully the enjoyment of human rights. Minorities are thus to arguments that the Constitution contains implied rights entitled to freedom in the peaceful expression of and freedoms. One visitor from the United States has even dissident views. called this period The Mason Court Revolution (Pierce, Now these opinions would be utterly unremarkable if 2006). For about eight years, roughly coinciding with the the Australian Constitution protected freedom of speech. period during which Sir Anthony Mason was Chief Justice But in 1988, it did not. The Commonwealth Government of Australia, the High Court took an activist turn, and the and the other States that intervened in this case (invok- implication of a bill of rights was entertained as a real ing the statutory provision that succeeded the special possibility. In a series of judgments stretching from 1988 standing rule upheld in the Union Label case) must have to 1996, High Court majorities held that the Constitution been very puzzled when they read this decision, and the protects “freedom of expression” (later backtracking to plaintiffs’ lawyers must have been very pleasantly sur- an implied freedom of political communication) (Lange prised that their apparently thin submission relating to v Australian Broadcasting Corporation (1997) 189 CLR the regulation of common names was able to achieve so 520; Levy v Victoria (1997) 189 CLR 579), that there much. We are not saying the case was wrongly decided; is an implication rising from the separation of judicial we are both very fond of freedom of expression and it is, power in the Constitution that a law cannot authorize after all, a human right. It just must have been surpris- a court to enforce a bill of attainder or a bill of pains or ing to the Commonwealth because it was clear that the penalties (Polyukhovich v Commonwealth (1991) 172 limitation on legislative power fashioned by the justices CLR 501, pp. 539, 612, 686, 706-7, 721), and that there in the Davis case asserted “freedom of expression” and is an implication arising from the separation of judicial “freedom in the peaceful expression of dissident views” power in the Constitution that prohibits legislation that as a ground for the striking down of federal regulations purports to authorise a court to re-imprison a person under (among other powers) the trademarks power. This who had already served their sentence without a fresh was a truly remarkable constitutional development. Constitutional lawyers are like bloodhounds, and a an implication that the Constitution prohibited a bill of criminal trial and finding of guilt (later backtracking to faint scent can be enough to get them to give chase. Within penalties that single out one person) (Kable v Director of Public Prosecutions (NSW) (1996) 189 CLR 51, pp. 96-7, lawyers, Sir Maurice Byers, argued in 1992 that there 106, 120, 125; Fardon v Attorney-General (Queensland) wasa few an short implied years freedom one of ofAustralia’s political communicationfinest constitutional that (2004) 223 CLR 575; Keyzer, 2008). Smaller pluralities invalidated Commonwealth legislation that criminal- held that there is an implied right to a fair trial (Dietrich v Nationwide The Queen (1992) 177 CLR 292, p. 326 (Deane and Toohey News Pty Ltd v Wills (1997) 177 CLR 1. Justice Murphy’s JJ), p. 362 (Gaudron J)), that there is an implied right to judgmentsized criticism in the of acases government from the official, 1970s andin 1980s did legal equality (Leeth v Commonwealth (1992) 174 CLR not need to be relied upon, because counsel had all the 455, p. 492 (Deane and Toohey JJ), p. 502 (Gaudron J)), encouragement in the world to advance an argument and that there is an implied prohibition on retrospective that the implied freedom exists after the Davis case was criminal legislation (Polyukhovich v Commonwealth (1991) decided. The 1992 High Court led by Sir Anthony Mason 172 CLR 501 (Deane J dissenting)). enthusiastically upheld the arguments, and the implied During the same period, amici curiae were applying freedom of political communication, foreshadowed by for admission without luck. In Brandy v Human Rights and Equal Opportunity Commission (1995) 183 CLR 245, This was but one of many cases of this variety. The a case about the power of the Australian Human Rights Murphy J, was finally born. most striking feature of the High Court’s constitutional Commission’s power to enforce its own orders, the Public

80 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

Interest Advocacy Centre made the following submission The plaintiffs had been removed from their families in support of its application to contribute as amicus curiae: when they were young children and detained in institu- In public interest litigation such as this it is de- tions, where many of them were physically and sexually sirable that the Court should have the assistance abused. They were separated from the land, culture and of submissions from community organisations traditions that are so central to Aboriginal identity. They such as the Public Interest Advocacy Centre. Over were regarded to be members of a “Stolen Generation”, the past twelve years the Centre’s work has de- and a federal inquiry was underway to tell their stories veloped in line with its charter of promoting the and lay the groundwork for reparations. public interest and enhancing the quality of public The plaintiffs launched an ambitious and wide rang- policy-making through analyzing and seeking the ing constitutional challenge to the Aboriginals Ordinance reform of laws, policies and practices which are on a number of grounds, including that the Ordinance interfered with the plaintiffs’ freedom of religion, which is protected by s 116 of the Constitution (this is one of the Counselunjust and for thedeficient. Public Interest Advocacy Centre, John Basten QC, had an illuminating exchange with one member very few express freedoms in the Australian Constitution), of the High Court, Justice McHugh: that the Ordinance infringed an implied right to equality before the law, that it infringed an implied freedom of movement and association, and that it usurped the sense that there would be nobody here to support judicial power of the Commonwealth by authorising the Basten QC: We are not here to fill a gap in the the validity (of the provisions). We are here, in a punishment of the plaintiffs without due process. Justice - Murphy had previously opined that some of these rights ested beside the Commonwealth of Australia in and freedoms could be drawn by implication from the maintainingsense, to fill thea gap validity in that of there the legislation are others and inter the Constitution, and, as noted above, there were threads legislative scheme. in some previous decisions justifying the advancement McHugh J: There might be six million who are. of some of these arguments. But by 1996, Mason CJ was Basten QC: That might be so, Your Honour. gone, and the High Court had been castigated by jurists, McHugh J: Why should you be given any special scholars and politicians for its “judicial activism”. The tide was turning. Basten QC: We are not seeking any special privi- It was in this context that the International Commis- privilege? lege, Your Honour. sion of Jurists, a group of concerned lawyers, lodged an application to be heard as amicus curiae to advance McHugh J: That is what it seems to me to be, Mr Basten. arguments that were supportive of the arguments of the plaintiffs. Brennan CJ and Dawson J repeatedly asked coun- Leave was refused. sel whether the Court had any reason to suppose that the Probably the most ambitious attempt to argue that issues raised in the case would not be fully and adequately there are a series of implied constitutional rights and argued by the parties (transcript of hearing, Kruger v freedoms in the Constitution was in Kruger v Common- Commonwealth, 12 February 1996, pp. 15-25). Dawson J wealth (1996) 190 CLR 1. This was a very sad case. Alec was conscious of the possibility that their holding might Kruger and several other Aboriginal people challenged create a precedent, remarking that “if leave were granted the constitutional validity of the Northern Territory Ab- in this case it would be a principle which was established originals Ordinance 1918, a Commonwealth regulation which would be brought up in later cases” (p.17). The justices of the Court expressed sympathy for the Aborigines” to undertake the “care, custody and control” plight of the plaintiffs but said that none of the constitutional thatof Aboriginal authorized people an official in the Northern called the Territory “Chief Protector of Australia. of arguments could hold and there was no constitutional

Law in Context, Vol 37, Issue 1, 2020 81 ISSN: 1839-4183 remedy for what had happened. Chief Justice Brennan issues in play in constitutional cases. Kruger appeared said the stories of the Stolen Generation had “profoundly to be a real setback. distressed the nation”, and: In retrospect, many would say that the risk of a 4.4 THE ENUNCIATION OF DISTINCTIVE TESTS IN LEVY child suffering mental harm by being kept away V VICTORIA from its mother or family was too great to permit One more contextual feature of the constitutional even a well-intentioned policy of separation to landscape needs description before we launch into a be implemented, but the existence of that risk consideration of the principles relating to amicus curiae did not deny the legislative power to make laws that were expounded in Levy v Victoria (1997) 189 CLR which permitted the implementation of that policy. 579. A little earlier in this article, we drew attention to the As for the application by the would-be amicus curiae, Brennan CJ, speaking on behalf of the Court, was not an implied freedom of communication in the Australian encouraging: Constitution.High Court decision In 1994, in a 1992 majority that of confirmed 4:3 went eventhat therefurther is As to his application to be heard as amicus curiae, in Theophanous v Herald and Weekly Times Ltd (1994) he fails to show that the parties whose cause he 182 CLR 104 and Stephens v West Australian Newsarticles would support are unable or unwilling adequately Ltd (1994) 182 CLR 211 and held that the implied free- to protect their own interests or to assist the Court dom of communication could be invoked as a defense in arriving at the correct determination of the to a defamation action against a publisher of material case. The Court must be cautious in considering that was critical of a member of parliament or political applications to be heard by persons who would candidate. A politician would only succeed in an action for defamation if they established that the defendant was Court be prejudiced. Where the Court has parties aware of the falsity of the material, published the mate- beforebe amicus it who curiae are lest willing the efficient and able operation to provide of adthe- rial recklessly, and the publication was reasonable in the equate assistance to the Court it is inappropriate circumstances. (To American ears, even the words sound to grant the application. very familiar using much of the same language that is in The Court appeared to have retreated from its willing- New York Times Co. v. Sullivan (1964), a landmark case ness to consider new arguments that additional implied that forever changed First Amendment jurisprudence. rights and freedoms might exist (and, indeed, some of these Not surprisingly, it was a case in which amici made the arguments were never raised again). But, importantly to Court keenly aware of the broad implications of the dan- our present focus, the Chief Justice’s procedural ruling on ger of current libel law on freedom of the press). After the admission of the amicus curiae also signaled a retreat changes to the composition of the bench (two members to litigation led by parties and for parties. of the majority in these cases left – Mason CJ retired and Deane J was appointed Governor-General), the stage was This was not welcome news to would-be amici curiae. set for a reversal of these decisions. As we observed earlier in this article, while standing in Australian constitutional cases is nowhere near as com- Arguments that Theophanous and Stephens should plicated, and considerably more open than U.S. standing, be rolled back were advanced by the applicant in Lange the rules governing costs in Australian constitutional v Australian Broadcasting Corporation and by a number cases make the advancement of constitutional cases as a of government interveners in both Lange and Levy. In a party potentially very expensive. Also, as we noted ear- remarkable act of judicial leadership, Brennan CJ was lier, amici curiae in the U.S. cases bring societal issues to able to wrangle a unanimous judgment in Lange, which the attention of justices in ways that are not tied to the required Toohey and Gaudron JJ to recant the wider positions they had taken in Theophanous and Stephens. help justices recognize and address the human rights Importantly for our purposes though, while the implied specific facts of the case being litigated, and can thereby

82 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 freedom of communication was wound back a little, the It is not possible to identify in advance the situations Court allowed the Media, Entertainment and Arts Alliance in which the Court will be assisted by submissions (the MEAA) to appear as amicus curiae. In their submis- that will not or may not be presented by one of sions the MEAA advanced submissions supporting the the parties nor to identify the requisite capacities retention of the wider approach to freedom of speech in of an amicus who is willing to offer assistance. All Theophanous and Stephens. The MEAA submission cited that can be said is that an amicus will be heard a number of U.S. cases in its text. The Australian Press when the Court is of the opinion that it will be Council, an organization that resolves complaints against media companies for breaches in journalistic standards, cost to the parties or any delay consequent on agreeingsignificantly to hear assisted the amicus thereby, is not provided disproportion that any- amicus curiae, and this submission was accepted. ate to the assistance that is expected (internal also filed an application to make written submissions as references omitted). So it is in this context that Levy v Victoria (the duck hunting case we mentioned earlier in the article) can be However Justice Kirby, in a separate judgment, favoured considered. We set out the facts earlier in this article. Ap- a much broader approach (1997) 189 CLR 579, p. 651): plications were made, as noted above, by the union that In the United States of America and Canada, the represents journalists (the Media, Entertainment and Arts practice of hearing submissions from interveners Alliance) and the body that resolves complaints against and amici curiae is well established. Such practice media organisations (the Australian Press Council). In is particularly common where matters of general Levy, the Chief Justice, Sir Gerard Brennan said ((1997) public interest are being heard in the higher ap- 189 CLR 579, p.604): pellate courts. … The hearing of an amicus curiae is entirely in the There is no need for undue concern about adopting Court's discretion. That discretion is exercised on a broader approach. The Court itself retains full a different basis from that which governs the al- control over its procedures. It will always protect lowance of intervention. The footing on which an and respect the primacy of the parties. Costs and amicus curiae is heard is that that person is willing other inhibitions and risks will, almost always, to offer the Court a submission on law or relevant - fact which will assist the Court in a way in which sist can usually be recognised and easily rebuffed. the Court would not otherwise have been assisted. Thediscourage submissions officious of interveners busybodies. and Those amici who curiae per In Kruger v The Commonwealth, speaking for the will typically be conveyed, for the most part, in Court, I said in refusing counsel's application to writing. But sometimes oral argument by them appear for a person as amicus curiae: will be useful to the Court. Such interests may "As to his application to be heard as amicus curiae, occasionally have perspectives which help the he fails to show that the parties whose cause he Court to see a problem in a context larger than would support are unable or unwilling adequately that which the parties are willing, or able, to offer. to protect their own interests or to assist the Court That wider context is particularly appropriate to in arriving at the correct determination of the an ultimate national appellate court. It is especially case. The Court must be cautious in considering relevant to a constitutional case. applications to be heard by persons who would Nothing in the Australian Constitution prevents such a procedural course. Conforming to the Con- Court be prejudiced. Where the Court has parties stitution, this Court should adapt its procedures, be amicus curiae lest the efficient operation of the before it who are willing and able to provide ad- particularly in constitutional cases or where equate assistance to the Court it is inappropriate large issues of legal principle and legal policy to grant the application." are at stake, to ensure that its eventual opinions on contested legal questions are informed by

Law in Context, Vol 37, Issue 1, 2020 83 ISSN: 1839-4183

relevant submissions and enlivened by appropri- Supreme Court and the Australian High Court that are ate materials. relevant to the roles played by amici. In the present matter, I would have allowed the The High Court of Australia has retained, in large Council for Civil Liberties and other relevant bodies, measure, the oral tradition of advocacy. While there is had they applied, to make brief submissions on always pressure on the High Court’s calendar, there does the constitutional controversy. Such submissions would have been subject to the same strict condi- Court’s practice in constitutional cases since the mid- tions as applied to other interveners and amici. 1980snot appear (when to havethe “special been any leave” significant procedure, shift in much the High like If necessary, the relevant bodies could have been the U.S. “cert” procedure, was introduced). In Australia, restricted to written submissions. But I would constitutional cases can take days. It has been more than have allowed them a voice (internal references a century since the U.S. Supreme Court allowed counsel omitted). to address that court for hours on end (Hayne 2004). In the U.S., oral argument for a typical case is only one Later, in Roadshow Films Pty Ltd v iiNet Limited [2011] hour. Oral argument remains important, and at times is upheld Chief Justice Brennan’s approach and made no determinative, but the justices rely much more on written referenceHCA 54 five to justices Justice Kirby’sof the Court, approach in a unanimous at all. Notwithstand judgment,- briefs of counsel and amici. In Australia, being allowed to ing that fact, there is always a possibility that, in time, the make oral arguments is of vastly greater importance. Or to Kirby test will be preferred to the Brennan test. (After all, borrow a phrase from the American musical “Hamilton” to we have just observed that Justice Murphy’s implied rights be in the courtroom presenting oral argument is “to be in and freedoms jurisprudence of the late 1970s and early the room where it happens.” Another difference is in the case selection process. In the U.S. the basic route to the and early 1990s). What we will call the “Kirby test” will Supreme Court is that losing parties petition the Supreme be1980s considered appears at to length, have been later influential in the article. in the Firstly late we1980s lay Court for a writ of certiorari, and it is either granted or out what we call “the Brennan test”. denied without any oral argument. In Australia the path is much more complicated with many more touch points Chief Justice Brennan said that the admission of amici between counsel and the High Court. curiae by the Court is: Returning to the Brennan test, this test has resulted in • A matter of discretion; the following procedure being adopted First, if a person • The amicus curiae must draw attention to a fact or law which will assist the Court in a way in which the only invoke the discretion of the court, draw attention to Court would not otherwise have been assisted; theor group issue wants of fact to or file law written they wish submissions to raise, only,point they out needhow • The amicus must show that the parties are unable their submissions travel beyond the submissions of the or unwilling adequately to protect their own interests or parties (that is, address the “unable or unwilling” crite- to assist the Court in arriving at the correct determination ria), and submit that the Court will be assisted. Since it is of the case; now common practice for parties to agree to allow their • An amicus will be heard when the Court is of the written submissions to be published on the High Court’s - website in advance of a plenary constitutional hearing, vided that any cost to the parties or any delay consequent an amicus curiae can make their submission directly after onopinion agreeing that to it hearwill be the significantly amicus is not assisted disproportionate thereby, pro to the publication of submissions. the assistance that is expected (emphasis added). Alternatively, if a person wants to make oral submis- Before we analyze the Brennan test and its aftermath, sions, then they would apply to the registry signaling their intention to do so, ordinarily accompanied by their point out some important differences between the U.S. written submissions. Since the Court will ordinarily given our comparative framework, we pause briefly to

84 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 convene a “directions hearing” inviting the parties to how to deploy amicus curiae briefs to advance human make submissions as to an appropriate timetable for the rights outcomes in High Court cases. (A number of the resolution of the matter, the Court can invite would-be post-Levy cases did not raise human rights issues so they amici curiae to make their argument for access at that will not be considered further: in chronological order point, or (more often) reserve the question whether the Garcia v National Australia Bank Ltd (1998) 194 CLR 395; Attorney-General (Cth) v Breckler (1999) 197 CLR 83; the substantive hearing. Stevens v Kabushiki Kaisha Sony Computer Entertainment amicus curiae should be admitted until the first day of To aid our international readers, we pause to pro- (2005) 224 CLR 193; Campbells Cash and Carry Pty Ltd vide a few basics of the process in Australia. Unlike the v Fostif Pty Limited (2006) 229 CLR 386, 404-7; Central process in the United States, in Australia, amici curiae Bayside General Practice Association Limited v Commis- do not seek admission at the time the Court decides sioner of State Revenue (2006) 228 CLR 168; Macedonian that a constitutional matter should be heard in the High Orthodox Community Church St Petka Incorporated v His Court. Seeking admission before the High Court comes Eminence Petar The Diocesan Bishop of The Macedonian in various ways: as an appeal (a matter of discretion of Orthodox Diocese of Australia and New Zealand (2008) 237 the court, as in the United States, and determined as part CLR 68; IceTV Pty Limited v Nine Network Australia Pty of a procedure referred to as “special leave”; as a “case Limited (2009) 239 CLR 458; Lehman Brothers Holdings stated” or “question reserved” (a procedure which allows Inc v City of Swan (2010) 240 CLR 509; Cumerlong Hold- people to make an application before a single justice of ings Pty Ltd v Dalcross Properties Pty Ltd (2011) 243 CLR the Court raising a constitutional question as a basis for 492, 495-6; Roadshow Films Pty Ltd v iiNet Ltd (2011) 248 commencing a dispute); as a “removal” (a case removed CLR 37; Williams v Commonwealth of Australia (2012) 248 from another court so the constitutional point can be re- CLR 156, 175; Westport Insurance Corporation v Gordian solved, at the discretion of the court); or in a case where Runoff Ltd (2011) 244 CLR 239; Commonwealth of Aus- a person invokes the original jurisdiction of the court to tralia v Director, Fair Work Building Industry Inspectorate; seek remedies (remedies that may have been removed Construction, Forestry, Mining and Energy Union v Director, by federal legislation from other lower courts, such as Fair Work Building Industry Inspectorate (2015) 258 CLR the Federal Circuit Court). 482, 487-8; Australian Competition and Consumer Commis- It is important to consider why the Court would raise sion v Flight Centre Travel Group Limited (2016) 261 CLR a higher bar for oral submissions. Two reasons can be 203, 210-1; Forrest & Forrest Pty Ltd v Wilson (2017) 262 posited. CLR 510, 513-4; Re Roberts [2017] HCA 39; Re Canavan Re Canavan; Re Ludlam; Re Waters; Re Roberts [No 2]; Re First, as we have already said, oral argument is par- Joyce; Re Nash; Re Xenophon (2017) 263 CLR 284; Re Nash ticularly important. The second reason is related to the [No 2] (2017) 263 CLR 443, 446-7; Northern Territory v Sangare [2019] HCA 25; Rinehart v Hancock Prospecting a premium for the court and the lawyers. There is a pos- Pty Ltd [2019] HCA 13). sibilityfirst. As that both the Brennan admission CJ and of aKirby great J manyindicate, amici time curiae is at to deliver oral submissions could place pressure on the An example of how the Brennan test operates is Maloney court and the parties. This should be avoided. For these v The Queen (2013) 252 CLR 168. Joan Maloney, an Indig- reasons, oral submissions can only be made when the enous woman and resident of Palm Island, Queensland, a Court expects significant assistance. predominantly Indigenous community, was charged with possession of more than a prescribed amount of alcohol How has the Brennan test operated in the cases that contrary to the Queensland Liquor Act. Palm Island had been declared by State law to be a “dry” community; al- cases, before considering the Kirby test, we then conclude cohol restrictions were in place. Maloney argued that the thefollowed? article byIn consideringthe next section what weAustralian analyse constitutional the relevant Queensland provisions were invalid to the extent of their lawyers might learn from US constitutional lawyers about inconsistency with the federal Racial Discrimination Act,

Law in Context, Vol 37, Issue 1, 2020 85 ISSN: 1839-4183 which prohibits laws that are racially discriminatory (and, - relevantly, implemented the Convention for the Elimina- gravated offence attracted a mandatory minimum sentence tion of All Forms of Racial Discrimination). Applications (people smuggling involving at least five people). The ag were made by the Australian Human Rights Commission of three years. The constitutional question was whether (which has a statutory authority to intervene), and also by theprovision: mandatory five yearsminimum with sentence a minimum was non-parole incompatible period with the National Congress of Australia’s First Peoples Limited, the independence, impartiality and integrity of the Ch III (instructed by the Human Rights Law Resource Centre Ltd) court called on to enforce it. The Court granted leave to seeking leave to appear as amicus (Maloney v The Queen the AHRC to participate as amicus curiae, but French CJ [2012] HCA Trans 342 (11 December 2012)). However advised counsel for the AHRC that it would be limited to written submissions only (presumably the submissions saying: “I should say that in the event that you seek to makeChief Justiceany oral French submission specifically to supplement applied the Brennanyour written test, assistance” threshold for oral submissions). passed the “assistance” threshold, but not the “significant As noted, the Court did not explain why it made this point you would seek to make that differs from those of ruling ([2013] HCA Trans 200). Although the Court is al- submissions we will want you to first state succinctly the the appellant” ((2013) 252 CLR 168). The case report ways very busy, failing to provide reasons why an amicus indicates that brief oral submissions by the National Con- curiae’s application to make oral or written submissions gress of Australia’s First Peoples were ultimately allowed (p. 172). The Australian Human Rights Commission also work out how they should mount their applications. If made brief oral submissions (pp.171-2). Kiefel J (as Her theis rejected Court does makes not it explaindifficult why for other it allows would-be amici amiciin some to Honour then was) (p.230) and Bell J (pp. 249-250) referred extensively to the submissions of the Australian Human Rights Commission in their decisions (which concerned actuallycases and requires. not in others, (For example,it is difficult in Westernto understand Australia what v Article 7 of the Universal Declaration of Human Right Brownthe “significant (2014) 253assistance” CLR 507 or theeven Solicitor-General the “assistance” testfor and Article 26 of the International Covenant on Civil and the State of South Australia was granted leave to appear Political Rights). Justice Gageler’s judgment extensively as amicus curiae in this Western Australian native title case. It is not clear why. Notably, Australian Lawyers for accepting the submissions of Queensland (thereby reject- Human Rights were denied leave to appear as amicus canvasses relevant human rights principles, specifically ing the submission of the National Congress) (p. 291). curiae: [2014] HCA Trans 14. No reasons were provided). It is not apparent from the judgments of the Court that This uncertainty may well deter applications and could the submissions made by the National Congress had any increase the expense of developing applications. Though impact on the Court (that does not mean the submissions we praise the U.S. Supreme Court for its openness to am- were not listened to, or read and considered, only that the icus participation, it is often criticized for not explaining why it denies certiorari. The rules of what makes a case In Maloney certworthy are similarly vague. Even hidden is the vote for judgments do not reflect any impact). Brennan Test in determining whether oral submissions who did and did not agree to take the case unless there is should be allowed. the ButChief the Justice Court doesspecifically not always applied provide the a public dissent from the denial of certiorari (Perry 1991). reasons why it rejects or restricts amici. For example, in - Magaming v The Queen (2013) 252 CLR 381, 385 the Aus- cedure adopted when amici curiae make applications to tralian Human Rights Commission (AHRC) sought leave makeAs oralnoted submissions. above, there Counsel is a significant representing defect a in would-be the pro to participate as amicus curiae in a case about whether amicus need to physically appear in court to do so. This mandatory minimum sentences under federal law usurped can be very expensive for the client whether it takes place judicial power contrary to Chapter III of the Constitution. at a directions hearing or in the substantive hearing itself. These hearings are almost invariably heard in Court No 1 convicted of an “aggravated offence” of people smuggling The case involved an Indonesian fisherman who had been in Canberra. Sometimes the Court will empanel only five

86 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 justices to hear a constitutional appeal, but this is very In addition, it should be noted that while U.S. amici rare—the usual rule is that all available justices (seven, curiae are active in cert petitions, they are non-existent assuming none are ill or on sabbatical or leave) shall sit. in the High Court’s equivalent special leave hearings. Very few if any legal counsel who appear in the High Court As far as we are aware, an application for amicus curiae status has never been made at a special leave hearing, to Canberra, typically from Sydney or Melbourne, to appear. during a removal, or in any of the preliminary hearings Canberrareside in Canberra, Airport isso often that means plagued that by they fog, must particularly fly or drive in that precede a directions hearing or a substantive hear- the morning, and so the High Court requires counsel to ing in a constitutional case. This contrasts sharply to U.S. travel to Canberra the day before any High Court hearing practice, where amicus briefs provide important signals to ensure that counsel are in attendance and to avoid em- about the importance of cases. As “public importance” is barrassment to the Court when proceedings commence. one of the criteria for allowing an application for special Counsel therefore typically set aside a day and a half to leave to the High Court, it makes sense to enable amici to do so: the travel back and forth to their home State and participate at this point. We are not sure why this does not for the work. If admitted, they would stay for the hearing take place, but believe it is worthy of further investigation, day in case the Court required their services further, or if given the prominence of amici involvement in the United they sought leave to respond to an oral submission from States “cert” process. one or both or multiple parties. Sometimes cases last for more than one day. Counsel can ask to be excused, but even after considerable expense, the trip to Canberra in light of the travel and accommodation requirements, mayAnother be futile. significant So, for example, defect inof Ministerthe Brennan for Immigration test is that a charge to the client for one and a half days would not and Multicultural and Indigenous Affairs v QAAH of 2004 be unreasonable. There are also costs for preparation. ((2006) 231 CLR 1, pp. 4-5) the Gleeson Court rejected Experienced senior counsel – and senior counsel ordinar- an application by the United Nations High Commissioner ily make these applications – reputedly charge between for Refugees (UNHCR) seeking leave to make oral sub- $10,000 to $15,000 per day for their services. Now it is missions. QAAH was a man from Afghanistan who had very likely that counsel representing amici in cases raising received a temporary protection visa and then sought human rights issues would only charge a nominal fee, or permanent residence. The Refugee Review Tribunal had perhaps only a fee to cover their travel and accommoda- determined that there had been a change in circumstances tion expenses, but this could still amount to several thou- in Afghanistan since he had made his application, and this sands of dollars. This is plainly a drain on resources that meant that it was no longer unsafe for him to return to might be used for other varieties of legal mobilization in that country. The Court declined to hear counsel for the support of human rights. It would be preferable if amici Commissioner. Instead, the UNHCR was granted leave to curiae were advised before the hearing, and preferably make written submissions as amicus curiae, and argued well before the hearing, if their application will be granted that there “must be positive information demonstrating or not. It is well understood that the purpose of an oral a settled and durable situation incompatible with a real hearing is to give parties every opportunity to advance chance of persecution arising from the circumstances in their client’s case, and that means that a party or parties connection with which the person has been recognised as might object to an application by a would-be amicus on a refugee”. The majority of the Court allowed the appeal the day of the hearing. However, in our age of electronic but did not refer to the submissions of the amicus curiae. communications, interactive audio-video conferencing, This may be interpreted as an example of an applicant and bearing in mind the Court’s recently-adopted practice of publishing virtually all submissions made in High by the Brennan test. failing to exceed the “significant assistance” threshold set Court cases, surely there must be less expensive ways of Justice Kirby dissented from the procedural ruling. Kirby J said: making this decision? Could the decision be made by the Court in advance of the hearing? We think so.

Law in Context, Vol 37, Issue 1, 2020 87 ISSN: 1839-4183

“this appeal, UNHCR exceptionally, and so far as I a split decision, that this constitutional guarantee did not am aware, uniquely, sought to be heard as an amicus apply in Commonwealth territories. The court was invited curiae. Counsel were retained for this purpose and to reconsider this decision in the following circumstances: travelled to the hearing to make the application. I would unhesitatingly have granted leave for UN- of legislation, including the Northern Territory National HCR to be heard in these proceedings. However, Emergencyin August 2007, Response the Commonwealth Act, which was said enacted to address five pieces issues unrestricted leave for oral argument was withheld raised in a report of the Northern Territory Government by the Court. The UNHCR's participation was con- called Little Children Are Sacred. That report had said that there was widespread sexual abuse of children in remote of UNHCR is recorded in important proceedings communities in the Northern Territory. The Emergency infined national to written courts submissions. overseas. In … my The view, intervention it should be welcomed, not resisted. Decisions of national in favour of the Commonwealth over land that was oth- courts play an important role in expressing the erwiseResponse subject Act authorized to native title, the provision ostensibly of to five-year provide leases legal meaning of the Convention and deciding the ap- certainty to business owners that may be willing to invest plication of such treaty law. In effect, in deciding in activities in these communities. Wurridjal challenged cases such as the present, national courts are the constitutional validity of the law on the basis that it exercising a species of international jurisdiction. The more assistance courts can receive from the failed to provide just terms for the property acquisition. relevant international agencies, in discharging Two academics from Australian National University such international functions, the better (internal references omitted)”. in the case. The application was dismissed and the submis- sionstogether were filed subjected an application to criticism to be by heard a majority as amicus of the curiae Court submissions in the later decision of CPCF v Minister for the decision to would-be amici, the relevant passages of ImmigrationPerhaps chastened, and Border the Protection UNHCR chose ((2015) to file 255 written CLR the([260]-[263] judgments (contrast are set out 312-3)). at length: Given the significance of 514). In this case, a number of Sri Lankan Tamils arrived FRENCH CJ. A majority of the Court is of the opinion that this is not a case in which the submissions and material offered by those who would inter- in Australian waters in a boat carrying an Indian flag.- vene as friends of the Court are likely to be of any missionerThe Australian for Refugees Human wasRights also Commission granted leave filed as writtenamicus submissions. The Office of the United Nations High Com assistance. The Court may be assisted where a prospective amicus curiae can present arguments other than on issues of statutory construction and the on aspects of a matter before the Court which are Constitution.curiae to file written Unfortunately, submissions the case on reportbehalf ofprovides the Office no otherwise unlikely to receive full or adequate indication of the content of the submissions). treatment by the parties because (a) it is not in Dissent from a procedural ruling of the Chief Justice is the interests of the parties to present argument an extraordinary step for a justice of the Court to take, so on those aspects or (b) one or other of the par- it was very clear that Justice Kirby was making a point. In ties lacks the resources to present full argument some cases, the Court appears to have been not merely to the Court on them. In some cases it may be in negative about an application by an amicus curiae, but the interests of the administration of justice that positively hostile. Wurridjal v The Commonwealth (2009) 237 CLR 309 concerned the scope of s 51(xxxi) of the Con- the matter before it than the parties are able or stitution, which authorises the Commonwealth Parliament willingthe Court to offer. have In the the benefit present of case, a larger the Court view has of to make laws for the compulsory acquisition of property for received a large volume of material said to sup- a purpose of the Commonwealth. These acquisitions must port the proposition that the rights claimed by the be made on “just terms”. In 1997, the Court had decided, in plaintiffs constitute property for the purposes of s

88 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

51(xxxi) of the Constitution. The material consists and the practice and decisions of foreign and of what are said to be relevant international law international courts and bodies relevant to the instruments and international jurisprudence. The treatment of indigenous peoples. Such materials may be relevant to this Court’s deliberations as that contention and some general statements the arguments develop. Therefore, I would be aboutsubmissions the wide do meaning not travel to significantlybe given to the beyond word inclined at this stage to receive the amici’s written “property” in s 51(xxxi). They do not show how, submissions and simply to use those materials, having regard to the particular statutory framework with discretion, as they prove to be relevant as in which the plaintiff’s property rights arise and the argument advances. the operation of the impugned laws, the material Alternatively, I would reserve the question of is of any relevance. Before the Court will accept whether the amici should be heard or should be the offer of assistance of an amicus curiae it must permitted to place their written materials before the Court for decision later in the proceedings. I a large amount of material, supported by what would note that most of the written materials that isbe little satisfied more that than it willan assertion be assisted. about The itstender utility, of were tendered with, and to support, the written submission are publicly available in any event. We these proceedings. The summons will therefore are now on notice of them. Most of them could be beis not dismissed sufficient (emphasis to give to added). the tenderer a voice in used by the Court as background or contextual KIRBY J. The practice of this Court in recent years materials in any case. has moved in the direction of widening the cir- The formal order that I would propose is that the cumstances in which amici curiae will be heard, or reception of the submissions of the amici should at least permitted to tender written submissions be reserved by the Court until later in the hearing. and materials (1). In taking this course, the Court I note that the parties to the proceedings have has simply, if somewhat belatedly, followed the no objection to the Court’s receiving the written submissions of the amici. Nor, at this stage, should law countries. It has done so out of recognition this Court. ofpractice the special of other role final played national by such courts courts, in common including this Court, in expressing the law, especially CRENNAN J. I agree with Kirby J. in constitutional cases in a way that necessarily This decision sounds a cautionary note for would-be goes beyond the interests and submissions of the amici curiae. It will be necessary for an applicant to ap- particular parties to litigation. The present is a case involving such issues. Whether the Court would ply the law to the facts and reach conclusions, and not be assisted by the submissions of the proposed merely refer the Court to materials that may be relevant. An argument must be advanced. stage before the Court has heard any argument. Justice Kirby draws attention to a further problem with amici is difficult, or impossible, to decide at this the contemporary approach to amici curiae: it may be that I agree with the other members of the Court that the utility of their submissions may not become apparent some of the materials proffered by the proposed until the argument has advanced within the proceedings. amici appear somewhat undigested and lacking If the court makes a decision to deny access at the com- in demonstrated application to the issues in the mencement of the hearing then there is apparently no proceedings. Nevertheless, the actual submission way to reconsider this decision later in the proceedings. of the proposed amici is quite brief, being but twenty pages. It refers to new materials that are This point plainly commended itself to Justice Crennan. not referred to in the submissions of the parties The experience of the unsuccessful applicants in Wur- and, in particular, materials on international law ridjal can be contrasted to the experience of the Combined Community Legal Centres Group, who were granted

Law in Context, Vol 37, Issue 1, 2020 89 ISSN: 1839-4183 admission to make oral and written submissions in APLA (written submissions)) and LibertyWorks Inc appeared as Limited v Legal Services Commissioner (NSW) (2005) 224 amicus curiae in Preston v Avery (written submissions)). CLR 322 (pp.328-330). In this case, the Combined Com- The submissions made extensive reference to human munity Legal Centres Group argued that a State regulation rights issues ([2019] HCA 11. An application by Access prohibiting the advertising of legal services for personal Zone Action Group, an anti-choice group, for amicus status injury matters materially affected their freedom of po- was rejected by Gordon J (see Clubb v Edwards & Anor [2018] HCATrans 181). information to enable them to bring matters to court. Mind you, the record number of appearances in these litical communication, specifically, a freedom to obtain They argued that their interest was different to APLA, cases were for written submissions, which only need cross a group of plaintiff lawyers that relied on advertising to the threshold of providing “assistance” to the Court, not increase their business. John Basten QC, a gifted lawyer the higher threshold requirement for oral submissions. (and former academic) with substantial experience in Still, via written submissions, amici curiae have advanced the High Court who has since been elevated to the New human rights principles by drawing the attention of the South Wales Court of Appeal (and who represented Mr High Court to international human rights conventions Davis in Davis v Commonwealth), was even granted an that explain the meaning of particular words and phrases opportunity to provide an oral reply to other submissions used in statutes, assisting the court to reach conclusions made (p.340). Their submissions were largely accepted that are consistent with those human rights conventions. by the dissentients, McHugh and Kirby JJ. Justice Kirby J Amnesty International Australia addressed the question remarked that the submissions underscored “the need, whether denial of freedom to express sexual preference in large and complex legal (and especially constitutional) could constitute persecution for the purposes of the concerns, for this Court to be ready to receive submissions refugee test (2003) (in Appellant S395/2002 v Minister from non-parties that have substantive arguments to the for Immigration and Multicultural Affairs (2003) 216 CLR issues that fall for decision” (p.309). Justice Gummow did 473, two homosexual men from Bangladesh had sought not review the submissions of the amicus curiae exten- protection in Australia as refugees, citing a well-founded sively (pp.379-381). fear of persecution in Bangladesh on the basis of their Likewise, the amici curiae in Clubb v Edwards and Pres- sexuality. The Refugee Tribunal decided that the men ton v Avery [2019] HCA 11 had a very positive reception could return to Bangladesh and live safely, commenting from the Court. The case sets a new record for numbers that they might need to be “discreet”. Although the appeal of amici represented in a case (four). It concerned efforts was disposed of on other grounds, Amnesty International by Victoria and Tasmania to decriminalise abortion. Each State had sought to provide that those seeking access to, the question whether denial of freedom to express sexual or working in, premises where terminations are available, preferenceAustralia was could granted constitute leave persecution).to file written Insubmissions another case, on were protected from the hindrance of protests or other made submissions that the Family Court’s jurisdiction to communications about abortions being made within 150 protect the welfare of the child was, in part, implementing metres of such premises. The appellants challenged the the United Nations Convention on the Rights of the Child, constitutional validity of the statutory provisions reinforcing the safe access zones. They argued that the laws children (Minister for Immigration and Multicultural and were not necessary, or rationally connected to a legitimate Indigenouswhich significantly Affairs v limited B (2004) the 219immigration CLR 365, detention pp. 373-4). of purpose, but instead infringed on the implied freedom People with Disabilities (NSW) Inc drew the Court’s atten- of political communication, which includes a right to - protest. The High Court rejected these arguments. Four tional human rights law in a case about federal disability organisations, the Castan Centre for Human Rights Law, discriminationtion to the definition legislation, of the whichword “disability” in turn implemented in interna international human rights principles (2003) (Purvis v Law Centre (appeared as amici curiae in Clubb v Edwards New South Wales (2003) 217 CLR 92, p. 96). the Fertility Control Clinic (A firm), the Human Rights

90 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

The position of the Australian Human Rights Commis- Art 22 would also be an infringement of the freedom of sion (AHRC) as an amicus curiae is a little bit different. political communication, there will be invalidity”. While the High Court is under no obligation to accept In North Australian Aboriginal Justice Agency Limited its submissions, it does have the advantage of being a v Northern Territory, the Australian Human Rights Com-

amicus curiae. The submissions were not considered in publicly-funded agency, with its role confirmed in statute amicus briefs in courts includes Racial Discrimination Act anymission of the was judgments. granted leave to file written submissions as (the statutory provisions authorizing the AHRC to file 1975 (Cth), s 20(1)(e); Sex Discrimination Act 1984 (Cth), In M47/2018, the plaintiff has been in immigration de- s 48(1)(gb); Disability Discrimination Act 1992 (Cth), s tention since 2010. The plaintiff said that he was stateless 67(1)(l); Age Discrimination Act 2004 (Cth), s 53(1)(g) and that continuing his detention was unconstitutional and under the Australian Human Rights Commission Act and breached his human rights. The respondent said that 1986 (Cth) s 11(1)(o) and s. 31(j)). Its policy regarding his state of origin could not be established because he had failed to cooperate with authorities. The Australian Hu- website: “When a relevant human rights or discrimination man Rights Commission made written submissions as an making submissions as amicus curiae is reflected on its issue arises in a case and the Commission could provide amicus curiae on the construction of relevant provisions expert assistance that would otherwise not be available of the Migration Act 1958 (Cth). to the Court, the Commission may seek leave of the Court To sum up to this point, notwithstanding the apparent to intervene in the proceedings. The Commission will narrowness of the Brennan approach, since Levy v Victoria, then make submissions on the issues that relate to the and at the time of writing, there has been a steady stream of decisions, over forty, in which the High Court heard ap- submissions in Tajjour v New South Wales (2014) 254 CLR 508,Commission's North Australian powers”. Aboriginal The AHRC Justice has made Agency significant Limited rise: indeed, based on calculations in previous work, we v Northern Territory (2015) 256 CLR 569 (p. 578) and estimateplications that for thereleave fromhave beenamici twice curiae. as This many is applicationsa significant M47/2018 v Minister for Home Affairs [2019] HCA 17. for amicus status in the last 20 years than there were in In Tajjour, Tajjour and others challenged Queensland the preceding century (Keyzer 2010, pp 102-18). Now legislation that prohibited consorting with criminals. It this is a far cry from the over one hundred applications in was argued this was contrary to the implied freedom of the Obama Care case. But given that Australia has about political communication, and an implied freedom of asso- the same population as Texas, four applications in an ciation arising therefrom. Tajjour also argued that Art 22 of the ICCPR restricted the scope of the legislative power United States (after all, the population of the US is over 13 of NSW. The Court rejected this argument (French CJ [48], timesAustralian the population case might of be Australia). comparable However, to, say, importantly, fifty in the Hayne J [58],[98], Gageler J [136] and Keane J [249]). The there was no indication from the justices deciding Clubb Australian Human Rights Commission was given leave v Edwards/Preston v Avery that they felt inundated by the to appear as amicus curiae, and made submissions that additional reading, or any indication from the parties, and were less ambitious than the submissions made by the the many Attorneys-General intervening, that they were applicants: “Habitual consorting … is a subspecies of a concerned about this new record being set. larger genus of association within the meaning of Art 22 It is interesting to note that Amnesty International of the International Covenant on Civil and Political Rights. Australia and the Human Rights Law Centre have become There is no human rights concern if habitual consorting “repeat players”, building knowledge about how to prepare submissions in future cases that are more likely to be exercise of the freedom of association in the interests of accepted, which is of course the predicate to their consid- thefits ordrewithin public Art 22(2),. … Domestic which legislationpermits restrictions is not required on the to eration. The Human Rights Law Centre has managed to conform to international agreements … if infringement of persuade the High Court to enable it to participate in six

Law in Context, Vol 37, Issue 1, 2020 91 ISSN: 1839-4183 of the cases on the second list above: the Human Rights Dialogue Model case (Momcilovic v The Queen), the Street amicus curiae. Centre was granted leave to file written submissions as Church Freedom of Speech case (Attorney-General (SA) v In Commonwealth v Australian Capital Territory (the Corporation of the City of Adelaide), the Marriage Equality “Marriage Equality Case”) (2013) 250 CLR 441, 449-52, case (Australian Capital Territory v Commonwealth), the Australian Marriage Equality Inc (solicitors, the Human Right to No Gender case (NSW Registrar of Births, Deaths Rights Law Centre) delivered extensive submissions on the and Marriages v Norrie), the Freedom of Assembly case (Brown v Tasmania) and the Pro-Choice Safe Zones cases the Australian Capital Territory that if the Commonwealth (analysed elsewhere in this article). Marriagedefinition Act of “marriage”,was restricted and to supporting heterosexual the couples, argument then of Momcilovic v The Queen (2011) 245 CLR 1 was a test the ACT could pass valid legislation enabling same sex case on the “dialogue model” of human rights protec- civil unions. The Court accepted their submission that tion implemented by the Victorian Charter of Human the marriage power in s 51(xxi) of the Constitution could Rights and Responsibilities Act 2006. Section 36 of that enable the Commonwealth to pass legislation authorizing Act empowers the Supreme Court of Victoria to make a same sex marriage (p. 452 [2]). declaration that a statutory provision or provisions of In NSW Registrar of Births, Deaths and Marriages v Norrie the Victorian Parliament is incompatible with a human (2014) 250 CLR 490, Norrie had applied for registration of right. The declaration has no effect on the legality of the statutory provision; instead the declaration is provided Gender Agenda Inc, instructed by the Human Rights Law to the Attorney-General and the Minister administering Centre,her sex aswas “not granted stated” admission after a sex as affirmation amicus curiae procedure. to tender A the statute in question, who must then prepare a written written submissions that relied on analysis of State and response for consideration by the Victorian Parliament. federal human rights bodies (p. 492). A constitutional question was raised whether such a The Freedom of Assembly case (Brown v Tasmania) declaration is incompatible with the Supreme Court’s (2017) 261 CLR 328 concerned the constitutional validity constitutional position as a Chapter III court under the of provisions of the Workplaces (Protection from Protestors) Commonwealth Constitution, as Chapter III courts can Act 2014 (Tasmania) which provided police with move only make binding declarations of law. A majority of on and arrest powers that limited the scope of protest the Court declared that s 36 did not adversely affect the activity, on the facts, in an area of forest being logged. The impartiality and independence of the court. The Human Rights Law Centre made submissions in support of s 36 amicus curiae, including the submission that the Court and noted the similarity between the process it adopts “shouldHuman Rightsrecognise Law a limitedCentre filedform writtenof freedom submissions of assembly as and the legal position in Canada and South Africa. as indispensable to the system of representative govern- In Attorney-General (SA) v Corporation of the City of ment, not a wider free-standing principle of freedom of Adelaide (2013) 249 CLR 1, the “Street Church case”, Caleb association or movement”. This submission appears to and Samuel Corneloup were members of an unincorpo- have been ignored. rated association in Adelaide called Street Church. They wanted to preach and distribute religious pamphlets in 4.5 JUSTICE KIRBY’S BROADER APPROACH the Rundle Mall in Adelaide. Council by-laws prohibited The contemporary approach, what we have called the any person from preaching, canvassing, haranguing, etc. Brennan Test, has resulted in a steeply increasing number without permission. The Corneloups challenged the by-law. of amici curiae participating in constitutional cases to raise The Attorney-General for South Australia intervened. On human rights issues. However it does create practical and appeal to the High Court, a question was raised whether costly obstacles for would-be applicants, and, as Justice the by-law infringed the implied freedom to discuss po- Kirby pointed out in Wurridjal, denial of admission might litical and governmental affairs. The Human Rights Law deprive justices of access to materials that may become

92 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 relevant as an argument progresses, or which may be amici may not be clear until an argument has advanced. available to the justices at any rate. The High Court’s tradition is one of oral argument, and As noted in our review of the post-Levy decisions these arguments can and do evolve during the course of above, Kirby J has consistently maintained a more open proceedings. That being so, reserving any decision on approach than the majority of the Court. A summary of admission to the end of proceedings makes sense. his position, mosaiced from the public reasons he has The second aspect of the Kirby test is that applicants delivered for admitting amici curiae in a number of cases “may occasionally have perspectives which help the Court above, is as follows: to see a problem in a context larger than that which the par- 1. Amici should be admitted “when matters of general ties are willing, or able, to offer”. This echoes the language public interest are being heard in the higher appellate of the Brennan Test but emphasizes the context within courts” because “decisions of national courts play an which the Court makes decision. This is an important word important role” – that “wider context is particularly because it invites consideration of “policy” (another word appropriate to an ultimate national appellate court. used by Justice Kirby in his honour’s judgments about the It is especially relevant to a constitutional case” and admission of amici curiae, a word that is absent from the also “important” in cases about “the meaning of” in- Brennan Test). ternational treaties. As we have argued in this article, we think that amici 2. Amici “may occasionally have perspectives which help curiae have an important role to play in drawing the court’s the Court to see a problem in a context larger than that attention to the impact of its decisions on the human which the parties are willing, or able, to offer”. rights of minorities. However we contend that the input 3. The court can use costs and other inhibitions to dis- of amici curiae should not be yoked to any consideration of what the parties “are willing, or able, to offer”. Instead, we think the High Court, when it is making (constitutional) 4. One approach the Court could take would be to “receive courage amici curiae who are “officious busybodies”. decisions about human rights is doing important work the amici’s written submissions and simply to use those that needs to be properly-informed and well-informed. materials, with discretion, as they prove to be relevant It cannot be assumed that parties and solicitors-general as the argument advances” or, alternatively, to “reserve will be able to represent the diversity of important views the question of whether the amici should be heard or that should be taken into account. Australian rules govern- should be permitted to place their written materials ing the admission of amici curiae need to give interested before the Court for decision later in the proceedings”. people and interest groups a reasonable opportunity to - register their views about the way that the law should be sition. After all, “public importance” is one of the criteria developed. Unfortunately, the current approach neces- appliedIt is difficult by the Highto disagree Court whenwith Justice granting Kirby’s special first leave propo to sarily limits their preparation time and may narrow the parties to appear in the High Court. That said, although submissions they could make. Again, to repeat a point individual constitutional cases may vary in their degree made above, this is a far cry from the approach taken by of importance, they are all important. The same can be said for High Court cases that raise issues of statutory narrow than the U.S. approach, which operates more like interpretation requiring the consideration of the mean- aChief presumption Justice Griffith of access. in 1906. It is also considerably more ing of words or phrases in international human rights conventions. One could reasonably argue that every High Court case is important. Is this aspect of the “Kirby test” languageOur final is an quibble unfortunate with what legacy we from call theearlier Kirby common Test is a sound basis for limiting access by amici curiae to raise lawthe usedecisions of the onexpression standing “officiousthat abjured busybodies”. the ideological This plaintiff in public law cases. We are, hopefully, a long arguments in the High Court? Probably not. point. It may well be that the utility of the submissions of It is also difficult to argue with Justice Kirby’s final

Law in Context, Vol 37, Issue 1, 2020 93 ISSN: 1839-4183 way past the point where serious amici curiae might be curiae to the party-versus-party model of constitutional dismissed as mere meddlers. litigation. The amicus curiae must demonstrate to the court We believe that a person or organisation that goes to that they will 1. assist the court; 2. make a submission that the time and trouble of making an application to appear the parties have not or will not make; and 3. any cost to as amicus curiae will usually have a serious argument to the parties or any delay consequent on agreeing to hear make. The Court can certainly lay down expectations about the amicus must not be disproportionate to the assistance the quality of the submissions, as it did in Wurridjal. Any that is expected. Practically speaking, this means that the test of access should focus on the issues, rather than the people bringing the application to the court. Unlike the the parties’ arguments. Then, if the amicus curiae wants amicus curiae must have significant foreknowledge of early High Court in the Railway Servants case, we do not the opportunity to make oral submissions in a case, they endorse the principle that an amicus curiae can raise a may have to make those submissions at the commence- constitutional argument in a case where the parties have ment of the hearing, carrying the risk of incurring costs not done so. However it seems to us that once the parties for counsel and having no assurance that they will have join issue in a constitutional case, the High Court should that opportunity. The Kirby test is not really different on have input from people who can offer serious arguments this point: Justice Kirby would still require that a would- about the proper resolution of the matter (the “seriously be amicus “have perspectives that can help the Court” arguable” criterion provides a basis for rejecting an ap- (although it is unclear whether this is a threshold test of plication that is not seriously advanced). admission, per se) Our review demonstrates that while the High Court has 5. CONCLUSION been prepared to listen to more oral submissions and to The relatively restrictive posture of the High Court is read many more written submissions from amici curiae - since Levy, the test of admission remains problematic. The icus will be heard when the Court is of the opinion that it test creates uncertainty and unnecessary cost and incon- often justified by Justice Brennan’s position that “an am venience for the parties. Given the current test emphasises we often don’t know what we don’t know. Starting from the need for amici curiae to add value in their submissions, thatwill beself-awareness significantly assisted should bethereby.” the default The problemposition iswhen that we wonder whether litigants might be required to publish it relates to issues of human rights—especially those that their submissions in advance of not only plenary hearings burden minorities. Australian constitutional cases raise but special leave hearings, and then amici could be given a broad range of human rights concerns—access to safe submissions, with oral submissions on the application to termination services, support for marriage equality, the some time, perhaps a week, to prepare and file written right to protest, and other identity-based rights—and be admitted only being required should the Court decide demonstrate that amici curiae have an important role in about its workload, applications for amicus curiae status keeping the High Court informed. Since High Court cases to reject the written application? If the Court is concerned raising human rights concern all Australians, access to could be resolved by a single judge, or as in special leave court for amici curiae to make submissions on behalf of hearings, by two or three justices. minority interests is vitally important (Keyzer 2010, Ch 4). We also wonder whether the only really relevant criteria The question is of course how and when. The Brennan that should be applied to would-be amici curiae is that the test, and also the Kirby test, ties the fortunes of amici curiae submission be serious and arguable. Submissions that do to the parties and the arguments that they advance. This is and be determined on the papers. Page limits could be not meet this standard could forfeit a nominal filing fee CJ in 1906, who could raise a constitutional point in a non- set to discourage the prolix. Submissions that meet these constitutionala far cry from thecase amicus as part curiae of an “objection contemplated to jurisdiction”. by Griffith threshold tests could then be available for consideration In a very real way, these tests yoke the would-be amicus by the justices if they wish (as Justice Kirby observed in Wurridjal). The justices of the Court would have the

94 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

curiae to be admitted to enlarge the normative horizon too much additional reading. benefit of opening up their normative horizons, without Ultimately, the treatment of submissions is a matter that value choices about the terms of our collective life of the court. This is only fitting, since it is in these cases for the judges. Since justices can decide a matter having are made (Cheatle v The Queen (1993) 177 CLR 541). regard to their own opinion, there seems to be no harm, as Kirby and Crennan JJ noted in Wurridjal, having the 6. REFERENCES attention of the justices drawn to relevant materials, 1. Barker, Lucius J., 1967 “Third Parties in Litigation: A Sys- and then leaving it up to the individual justices to decide temic View of the Judicial Function”, The Journal of Politics whether they have regard to them. If our recommendations 291 (1): 41-69. are not satisfactory, perhaps relevant peak organisations 2. Caldeira, G. and Wright, J., 1988. “Organized Interests and such as the Law Council of Australia or the Australian Agenda Setting in the U.S. Supreme Court” American Political Bar Association could write to the Rules Committee of Science Review 82(4): 1109-1127. the High Court to invite it to consider taking steps that 3. Chemerinsky, Erwin, 2006. Constitutional Law: Principles would make the process more transparent and less costly and Policies (Aspen Publishers). for would-be applicants along these lines. 4. Collins Jr, Paul M., 2004. “Friends of the Court: Examining As we observed earlier in the article, there are Court Litigation”, Law and Society Review 38(4): 807-832. many problems with the American amici system. For 5. Collinsthe Influence Jr, Paul of M., Amicus and McCarthy, Curiae Participation Lauren A., in2017. U.S. Supreme“Friends example, the growth of “Amicus Wranglers” has added and Interveners: Interest Group Litigation in a Comparative vastly to the cost of litigation. Many a tree has died in the Context,” Journal of Law and Courts 5, 1: 55–80, https://doi. writing of amicus briefs that may never be taken very org/10.1086/690275. seriously by the Court. Perhaps more worrisome is that 6. Collins Jr., Paul M. and Martinek, Wendy L., 2011. “Who some of the Amicus Wranglers are very elite high-priced Participates as Amici Curiae in the U.S. Courts of Appeals,” lawyers in the private bar who can orchestrate amici in Judicature, 3: 128–36. a way that drowns out the voices of less powerful amici 7. Collins Jr., Paul M. and Solowiej, Lisa A., 2007. “Interest Group (Perry 2020). But we believe as a general matter, when it comes to constitutional decisions that involve human Court,” Law and Social Inquiry, 4: 955–84. Participation, Competition, and Conflict in the U.S. Supreme rights, erring on the side of bringing to the Court’s at- 8. Cordray, M.M. and Cordray, R. 2010. “The Solicitor General’s tention a wide range a views and understandings about Changing Role in Supreme Court Litigation”, Boston College problems that go beyond the parties is well worth any Law Review 51: 1323–82. relatively modest inconvenience to the court. 9. de Tocqueville, Alexis, Democracy in America, 1835. We hope that our survey of the amicus cases decided 10. Editorial Board, New York Times, “Joe Biden”, 17 January by the High Court in recent decades will assist practi- 2020, https://www.nytimes.com/interactive/2020/01/17/ opinion/joe-biden-nytimes-interview.html, accessed 24 tioners and researchers in devising strategies for their January 2020. continuing and increased involvement in these cases. 11. Evans, S., 2011. “Standing to Raise Constitutional Issues”, Our argument, however is about more than strategy. The Bond Law Review 22: 38-59. U.S. practice belies the notion that allowing broad-based 12. Feldman, A., 2016. “Successful Cert Amici 2014,” Empirical amici participation cannot work, and our two systems SCOTUS (blog), March 15, 2016, https://empiricalscotus. are no longer so different that drawing the comparison com/2016/03/15/certamici-2014/. is inappropriate. The choice of the Australian High Court 13. Frankfurter, Felix and Landis, James M. 1928. The Business to limit the role of amici is just that, a choice. Since the of the Supreme Court: A Study in the Federal Judicial System High Court has regard to community standards when it (New York: Macmillan). determines constitutional cases raising human rights issues (Nedelsky 2000), it is perfectly appropriate for amici

Law in Context, Vol 37, Issue 1, 2020 95 ISSN: 1839-4183

14. Franze, Anthony J and Anderson, R Reeves, 2015. “Record 30. Nicholson, C. and Collins, Paul, M. 2008. “The Solicitor General’s Amicus Curiae Strategies in the Supreme Court, 36 New Norm,” The National Law Journal August 19, 2015. American Politics Research 382–415. 15. French,Breaking Robert, Term for 2019. Amicus “The Curiae Constitution in Supreme and theCourt Protection Reflects 31. O’Neill, Nick. 1987. “Constitutional Human Rights in Aus- of Human Rights”, Edith Cowan University Vice-Chancellor’s tralia”, 17 Federal Law Review 85. Oration, 20 November 2019. 32. Perry, Jr., HW, 2010. “Access to Justice: Procedure, Polity and 16. - Politics,” Bond Law Review 22, no. 3, 192. tralian Legal System”, US Asia Centre, University of Western 33. Perry, Jr., H.W., 1991. Deciding to Decide: Agenda Setting in French, Robert, 2017. “United States Influence on the Aus Australia, 10 April 2017. the United States Supreme Court, Harvard University Press: 17. Cambridge, Massachusetts. Amicus Curiae Briefs on the Supreme Court”, University of 34. Kearney, J.D., and Merrill, T.W., 2000. “The Influence of Pennsylvania Law Review 148: 743-855. Court and Appellate Lawyering" 72 South Carolina Law Review. 18. Keyzer, P., 2010a. “A Battle and a Gamble: The Spectre of an 35. Perry, Jr., H.W.H.W., 2020a. 2020. “The“The ElitificationRise of State of Solicitors the U.S. SupremeGeneral”. Adverse Costs Order in Constitutional Litigation” Bond Law Article presented at the Annual Meetings of the Southern Review 22(3): 82. Political Science Association, San Juan, Puerto Rico. 19. Keyzer, P., 2010. Open Constitutional Courts, Federation Press. 36. Pierce, J., 2006. The Mason Court Revolution: The High Court 20. Keyzer, P., 2008. “Preserving Due Process or Warehousing of Australia Transformed, Carolina Academic Press. the Undesirables: To What End the Separation of Judicial 37. Smith, G.F. and Terrell, B.E., 1995. “The Amicus Curiae: A Power of the Commonwealth”, Sydney Law Review 30: 101. Powerful Friend for Poverty Law Advocates”, Clearinghouse 21. Keyzer, P., 2014. “Solicitors-General and the Public Interest”, Review 29(2): 772-792. in Appleby, G., Keyzer, P., and Williams, J., eds., Public Sentinels: 38. Songer, Donald, Kuersten, Ashlyn and Kaheny, Erin, 2000. A Comparative Study of Australian Solicitors-General, Ashgate “Why The Haves Don’t Always Come Out Ahead: Repeat 22. Krislov, S., 1963. “The Amicus Curiae Brief: From Friendship Players Meet Amici Curiae for the Disadvantaged”, Political to Advocacy,” Yale Law Journal, no. 4: 694–722. Research Quarterly 53(3): 537-556. 23. Larsen, Allison Orr and Devins, Neal, 2016. “The Amicus 39. Susman, Thomas, 2006. “Lobbying in the 21st Century - Machine” (2016) 102(8) Virginia Law Review 1901-1968. Reciprocity and the Need For Reform”, Administrative Law 24. Lazarus, Richard J., 2008. “Advocacy Matters before and Review 58: 737-751. within the Supreme Court: Transforming the Court by 40. Walker, K., 2002. “The Bishops, The Doctor, His Patient and Transforming the Bar,” Georgetown Law Journal 96: 1487. the Attorney-General” Federal Law Review 30: 507. 25. Lemos, Margaret, Young, Ernest. (2018) “State Public-Law 41. Wiggins, S., 1976. “Quasi-Party in the Guise of Amicus Curiae” Litigation in an Age of Polarization,” 97 Texas Law Review, 43 (1976) Cumberland Law Review 7: 293-305. 26. McGuire, Kevin T., 1994. “Amici Curiae and Strategies for 42. Willheim, E., 2011. “Amici Curiae and Access to Constitutional Gaining Access to the Supreme Court,” Political Research Justice in Australia”, Bond Law Review 22(3): 126-147. Quarterly 4: 821–37. 43. Wohl, A., 1996. “Friends with Agendas” American Bar As- 27. McGuire, Kevin T., 1995. “Repeat Players in the Supreme sociation Journal 82: 46-50. Court: The Role of Experienced Lawyers in Litigation Suc- cess”, The Journal of Politics 57: 187-196. LEGISLATION 28. Martha Minow, “Lawyering At The Margins: Lawyering for Human Dignity”, (2003) American University Journal of Australian Constitution, ss 92, 92A, 109 Gender, Society, Policy and Law 143-170. Judiciary Act 1903 (Cth), ss 78A and 78AA 29. Nedelsky, Jennifer, 2000. “Communities of Judgment and Age Discrimination Act 2004 (Cth), s 53(1)(g) Human Rights”, Theoretical Inquiries in Law 1: 2-28. Australian Human Rights Commission Act 1986 (Cth) s 11(1)(o) and s. 31(j). Disability Discrimination Act 1992 (Cth), s 67(1)(l) Racial Discrimination Act 1975 (Cth), s 20(1)(e)

96 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

Sex Discrimination Act 1984 (Cth), s 48(1)(gb) Commonwealth v Tasmania (the Tasmanian Dam Case) Workplaces (Protection from Protestors) Act 2014 (1983) 158 CLR 1 CPCF v Minister for Immigration and Border Protection CASE LAW (2015) 255 CLR 514 Croome v Tasmania (1997) 191 CLR 119 Cumerlong Holdings Pty Ltd v Dalcross Properties Pty Ltd AUSTRALIAN CASES (2011) 243 CLR 492 Ansett Transport Industries (Operations) Pty Ltd v Com- Davis v Commonwealth (1988) 166 CLR 79 monwealth (1977) 139 CLR 54 Dietrich v The Queen (1992) 177 CLR 292 APLA Limited v Legal Services Commissioner (NSW) (2005) Fardon v Attorney-General (Queensland) (2004) 223 CLR 575 224 CLR 322 Federated Amalgamated Government Railway and Tram- Appellant S395/2002 v Minister for Immigration and Mul- way Service Association v New South Wales Railway Traffic ticultural Affairs (2003) 216 CLR 473 Employés Association (the Railway Servants case) (1906) Attorney-General (Cth) v Breckler (1999) 197 CLR 83 4 CLR 488 Attorney-General (NSW); Ex rel Tooth & Co v Brewery Forrest & Forrest Pty Ltd v Wilson (2017) 262 CLR 510 Employés Union (the Union Label Case) (1908) 6 CLR 469 Garcia v National Australia Bank Ltd (1998) 194 CLR 395 Attorney-General (SA) v Corporation of the City of Adelaide IceTV Pty Limited v Nine Network Australia Pty Limited (2013) 249 CLR 1 (2009) 239 CLR 458 Austin v Commonwealth (2003) 215 CLR 185 Kable v Director of Public Prosecutions (NSW) (1996) 189 Australian Capital Television Pty Ltd v Commonwealth CLR 51 (1992) 177 CLR 106 Kruger v Commonwealth (1996) 190 CLR 1 Australian Competition and Consumer Commission v Flight Kuczborski v Queensland (2014) 254 CLR 51 Centre Travel Group Limited (2016) 261 CLR 203 Lange v Australian Broadcasting Corporation (1997) 189 Australian Conservation Foundation v Commonwealth CLR 520 (1980) 146 CLR 493 Leeth v Commonwealth (1992) 174 CLR 455 Brandy v Human Rights and Equal Opportunity Commission Lehman Brothers Holdings Inc v City of Swan (2010) 240 (1995) 183 CLR 245 CLR 509 Buck v Bavone (1976) 135 CLR 110 Levy v Victoria (1997) 189 CLR 579 Campbells Cash and Carry Pty Ltd v Fostif Pty Limited M47/2018 v Minister for Home Affairs [2019] HCA 17 (2006) 229 CLR 386 Mabo v Queensland (1988) 166 CLR 186 Re Canavan; Re Ludlam; Re Waters; Re Roberts [No 2]; Re Macedonian Orthodox Community Church St Petka In- Joyce; Re Nash; Re Xenophon (2017) 263 CLR 284 corporated v His Eminence Petar The Diocesan Bishop of Central Bayside General Practice Association Limited v The Macedonian Orthodox Diocese of Australia and New Commissioner of State Revenue (2006) 228 CLR 168 Zealand (2008) 237 CLR 68 Cheatle v The Queen (1993) 177 CLR 541 Magaming v The Queen (2013) 252 CLR 381 Clubb v Edwards [2019] HCA 11 Maloney v The Queen (2013) 252 CLR 168 Commonwealth v Australian Capital Territory (Marriage Re McBain; Ex parte Australian Catholic Bishops Conference Equality Case”) (2013) 250 CLR 441 (2002) 209 CLR 372 Commonwealth v Director, Fair Work Building Industry Miller v TCN Channel Nine Pty Ltd (1986) 161 CLR 556 Inspectorate; Construction, Forestry, Mining and Energy Minister for Immigration and Multicultural and Indigenous Union v Director, Fair Work Building Industry Inspectorate Affairs v B (2004) 219 CLR 365 (2015) 258 CLR 482 Minister for Immigration and Multicultural and Indigenous Affairs v QAAH of 2004 (2006) 231 CLR 1

Law in Context, Vol 37, Issue 1, 2020 97 ISSN: 1839-4183

Momcilovic v The Queen (2011) 245 CLR 1 UNITED KINGDOM CASES Re Nash [No 2] (2017) 263 CLR 443 Lilburne’s Case (1649) 4 State Tr 1270 Nationwide News Pty Ltd v Wills (1992) 177 CLR 1 Pharmaceutical Society of Great Britain v Dickson [1970] North Australian Aboriginal Justice Agency Limited v AC 403 Northern Territory (2015) 256 CLR 569 The Prince’s Case (1606) 8 Coke 1 Northern Territory v Sangare [2019] HCA 25 NSW Registrar of Births, Deaths and Marriages v Norrie UNITED STATES CASES (2014) 250 CLR 490 Plaintiff S157/2002 v Commonwealth (2003) 211 CLR 476 Allen v Wright 466 US 737 (1984) Polyukhovich v Commonwealth (1991) 172 CLR 501 City of Los Angeles v Lyons 431 US 95 (1983) Preston v Avery [2019] HCA 11 Crandall v State of Nevada 6 Wall 35 (1867) Purvis v New South Wales (2003) 217 CLR 92 Grutter v Bollinger 539 US 306 (2003) R v Smithers; Ex parte Benson (1912) 16 CLR 99 Jaffe v Redmond 518 US 1 (1996) Republic of Nauru v WET040 [2018] HCA 58 Juliana v United States 9th Circuit (2020) to be published Republic of Nauru v WET040 (No 2) [2018] HCA 60 Lujan v Defenders of Wildlife 504 US 555 (1992) Rinehart v Hancock Prospecting Pty Ltd [2019] HCA 13 New York Times Co. v. Sullivan 376. U.S. 254 (1964) Roadshow Films Pty Ltd v iiNet Limited [2011] HCA 54 Shapiro, Commissioner Welfare of Connecticut v Thompson Roadshow Films Pty Ltd v iiNet Ltd (2011) 248 CLR 37 394 US 618 (1968) Re Roberts [2017] HCA 39 Sierra Club v Morton 405 US 727 (1972) Stevens v Kabushiki Kaisha Sony Computer Entertainment Slaughter-House Cases 16 Wall 36 (1872) (2005) 224 CLR 193 United States v Guest 383 US 745 (1965) Tajjour v New South Wales (2014) 254 CLR 508 Valley Forge Christian College v Americans United for Sepa- Western Australia v Brown (2014) 253 CLR 507 ration of Church and State 454 US 464 (1982) Westport Insurance Corporation v Gordian Runoff Ltd (2011) 244 CLR 239 OTHER MATERIALS Williams v Commonwealth of Australia (2012) 248 CLR 156 Clubb v Edwards & Anor [2018] HCATrans 181 Wurridjal v The Commonwealth (2009) 237 CLR 309 Transcript, Kruger v Commonwealth, 12 February 1996 Maloney v The Queen [2012] HCA Trans 342 United States Supreme Court Rules, r 37

98 Law in Context, Vol 37, Issue 1, 2020 Received: June 29 2020, Date of acceptance: September 28 2020, Date of publication: November 24 2020, DOI: https://doi.org/10.26826/law-in-context.v37i1.118

The Adultification of the Youth Justice System: The Victorian Experience

By Natalia Antolak-Saper, Orcid: https://orcid.org/0000-0002-2680-546X, [email protected] Faculty of Law, Monash University, Melbourne, Australia

ABSTRACT In early 2018, an Inquiry into Youth Justice Centres in Victoria (Inquiry) found that a combination of a punitive approach to youth justice, inadequate crime strategies, and a lack of appropriately trained and experienced staff at youth justice centres, greatly contributed to the hindrance of the rehabilitation of young persons in detention in Victoria, Australia. In addition to identifying these challenges, the Inquiry also determined that the way in which

young offenders have been described by politicians and portrayed in the media in recent times, has had a significant onimpact the welfare on shaping model youth and justice the justice policies model. and practices.This is followed This article by a discussionspecifically of examines the perception the role and of the reality media of youthin the offendingadultification in Victoria. of the Victorian Here, it is youth demonstrated justice system. that through It begins framing, with a historicalthe media examination represents heightened of youth justice, levels drawingof youth offending and suggests that only a ‘tough on crime’ approach can curb such offending; an approach that has been adopted by the Victorian State Government in recent years. Finally, the article considers how recent youth justice

these reforms are inconsistent with the best interests of the child. reforms are examples of adultification, and by not adequately distinguishing between a child and adult offender, Keywords – youth justice, criminal law and procedure, welfare model, justice model

Acknowledgments – I am grateful to the anonymous reviewers’ for their helpful comments. License – This work is under Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) https://creativecommons.org/licenses/by-nc-sa/4.0/ Suggested citation: Law in Context, 37 (1): 99-113. DOI: https:// doi.org/10.26826/law-in-context.v37i1.118 Antolak-Saper, N “The Adultification of the Youth Justice System: The Victorian Experience.” Summary 1. Introduction 2. The Framework of the Youth Justice System 2.1 The Victorian Youth Justice System 2.1.1 The Children, Youth and Families Act 2005 (Vic) 3. The More Things Change …Youth Offending in Victoria: Perception vs Reality 3.1 Youth Offending: Perception 3.2 Youth Offending: Reality 4. The Adultification of the Victorian Youth Justice System 5. Conclusion

Law in Context, Vol 37, Issue 1, 2020 99 ISSN: 1839-4183

6. INTRODUCTION discusses the two dominant theoretical models that underpin the youth justice system – the welfare model and In May 2017, the Victorian State Government intro- the justice model. This is then followed by an examination of the perception and reality of youth offending in Victoria, to, inter alia, ensure that ‘serious young offenders are with a particular emphasis on the years of 2016-2018. heldduced accountable significant for reforms their actions to the andyouth punished justice for system their The purpose of this, is to demonstrate that youth offend- crimes’(White 2017, p.12). These reforms were introduced ing does not necessarily warrant the ‘tough on crime’ predominantly through the enactment of the Children and approach that has been adopted by the Victorian State Justice Legislation (Youth Justice Reform) Act 2017 (Vic) Government in recent years. Further, this article suggests (YJR) and included the following changes to the youth that the laws and policies that were enacted in response justice system: the introduction of youth control orders to media reports and community concerns about youth which would facilitate the imposition of curfews and restrict a youth offender’s use of social media;1 increase the the tough on crime wave and the due process wave, which maximum penalties for certain offences from three to four resultsoffending in arethe inintroduction line with the of two a number waves of of adultification; due process years;2 ensure that young offenders who commit serious protections. Although the introduction of due process crimes be heard in adult court rather than the Children’s protections may appear to be prima facie positive, the Court of Victoria;3 and disclose information pertaining to effect of providing youth offenders with due process the identity of youth offenders in certain circumstances.4 alters the nature of the youth justice system. Both of Heralding a shift towards the increasing adultification of these ‘waves’ and examples of them are discussed further the youth justice system (Alida and Benekos 2010, p.5) below. Finally, the article suggests that the introduction of such reforms do not adequately distinguish between youth justice system that recognises that children who a child and adult offender and thereby, are inconsistent offend,these changes ought topose be atreated significant differently challenge to adultsto an effective (Taylor with the best interests of the child.6 2016, [6]-[9]). A key contributing factor in shaping this in which young offenders were described by politicians 7. THE FRAMEWORK OF THE YOUTH JUSTICE andadulitified how they youth were justice portrayed system in thewas media, said to in be particular the way SYSTEM between the years of 2016-2018, a period of heightened The idea of distinguishing children from adults is a reporting on youth justice matters.5 relatively modern concept. According to Aries, ‘in me- This article examines the role of the media in this dieval society the idea of childhood did not exist’ (Ar- context. It begins with a brief historical discussion of the ies 1979, p.128). For a large period of history, children Victorian youth justice system. In particular, this section were simply viewed as ‘mini adults’. In the context of

1 Children and Justice Legislation (Youth Justice Reform) Act 2017 (Vic), Part 3. the criminal justice system specifically, this meant that 2 Children and Justice Legislation (Youth Justice Reform) Act 2017 (Vic), Part 8, Division 2. 3 Children and Justice Legislation (Youth Justice Reform) Act 2017 (Vic), Part 2, Division 2. 4 Children and Justice Legislation (Youth Justice Reform) Act 2017 (Vic), Part 7. 5 This trend of media reporting appears to have continued in the years since. For example, news reporting on the Malmsbury Youth Justice Centre lockdown in 2019 emphasised the violent and ‘wild’ nature of the young offenders detained at the Centre: see, Rohan Smith, ‘Inside Melbourne’s wild youth prison’ news.com.au (9 Sep- tember 2019). However, a detailed analysis of these incidents is beyond the scope of this article. 6 Although the concept of the ‘best interest of the child’ is readily adopted in the context of child welfare decisions, in the youth justice area it is relatively uncommon. Nonetheless, the concept is contextualised by art 3 of the United Nations Convention on the Rights of the Child which requires that, inter alia, ‘in all actions concerning children, whether undertaken by …courts of law…the best interests of the child shall be a primary consideration.’ Because of the various factors that can contribute to determining whether the best interest of the child is satisfied, assessing what the concept is intended to mean includes an evaluation of ‘all the elements necessary to make a decision in a specific situation for a specific individual child or group of children.’ (Committee on the Rights of the Child, General Comment No 14 (2013) par. 47). In the context of this article, this includes factors such as situations of vulnerability; the development of the child and their gradual transition into adulthood life; and factors that are governed by section 23 of the Charter of Human Rights and Responsibilities 2006 (Vic): that a child offender who is detained must be segregated from all detained adults; that an accused child must be brought to trial quickly and that a child who has been convicted of an offence must be treated in a way that is appropriate for their age.

100 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 children were convicted and sanctioned as adults with no young offenders are ‘rational, responsible and account- separate processes or principles governing their trial or able’ (Sentencing Advisory Council 2012, p.44). Any of- sentencing (Feld 2013, p.1-21). The social construction of fence committed is a result of free choice. Consequently, ‘childhood’ and ‘children’ as it is now understood in most the appropriate criminal justice response is to impose western countries – innocent, less mature and in need of protection – developed slowly. The institutionalisation This model does promote due process in that the rules of ‘childhood’ did not occur until the Industrial Revolu- governinga sanction adult that criminalreflects thetrial seriousness processes are of provided the crime. to tion (Dolgin 1997, p.348), and the additional distinction young offenders. The justice model is reactive, whereas of adolescence was not recognised until the early 20th the welfare model is future-oriented (Freiberg, Fox and century (Monahan and Young 2008, p.32). Hogan 1989, p.283). This emerging need to recognise children as distinct The evolution of the two models brought about a philosophical debate as to which of the two – or a combination that predominantly underpin youth justice today; the of both – is most appropriate as a guiding principle for ‘welfarefrom adults model’ was and reflected the ‘justice in the model’. two theoretical The welfare models model the youth justice system. In this context, Freiberg sug- focusses on the rehabilitation needs of the young offender, gests that the debate cannot simply be a ‘needs vs deeds’ whereas the justice model is concerned with providing a debate, but rather, that an adequate youth justice system young offender with appropriate formalities, such as due ought to address the various underlying problems (Frei- process (O’Connor 1997, p.1). Although the two models berg 2002, p.2-3) such as ‘drugs and alcohol use, poverty, mental illness and unemployment’ (Sentencing Advisory ‘[y]oung people are seen as being in need of guidance and Council 2012, p.45). Travers on the other hand, believes assistancedeveloped separately(the welfare they aspect), have, over whilst time, at beenthe same conflated; time that the ‘needs vs deeds’ debate ‘remains central to policy offending is said to be the result of calculated decisions debates on youth justice’ (Travers 2010, p.99). However, by rational actors’ (Palmer and Walters 1995, p.161). Travers limits the application of this debate to only those The welfare model originated in the United States at young offenders who repeatedly commit serious offences the beginning of the 20th century and was the prevalent (Travers 2010, p.111). There may be therefore a large number of young offenders to whom this debate proves model in the mid-20th century (Sentencing Advisory to be less relevant. Counciltheory of 2012, influence p.44). until The focusthe development of the welfare of model the justice is on Finally, the Australian Law Reform Commission (ALRC) the needs of an offender and emphasises the treatment suggests that a third model has emerged in relatively and rehabilitation of an offender, rather than punishing recent years; a model that is based on restorative justice them (Sentencing Advisory Council 2012, p.44). For this principles (Australian Law Reform Commission 1997, reason, the welfare model is not largely concerned with [18.34]). According to the ALRC, this model is a ‘contex- the imposition of criminal justice procedures. According tual [one] that acknowledges the desirability of balancing to this model, the commission of a crime is a by-product young offenders’ rights against their responsibilities to the community’ (Australian Law Reform Commission 1997, moral choice. Typically, pursuant to this model, a young [18.34]). The purpose behind this model is for a young offenderof external is dealt influences, with informally, rather than privately the resultand distinctly of free offender to take individual responsibility, but for the re- differently from adult offenders (Freiberg, Fox and Hogan sponse to the commission of the crime be rehabilitative 1989, p.283). in nature with a particular focus on reparation for harm Partly as a response to the disillusionment with re- done, rather than the imposition of punitive sanctions. habilitation the justice model on the other hand, places Although Australia never embraced the welfare model as robustly as the United States did, youth justice has (Cuneen and White 2011, p.107). According to this model, evolved to be distinct from the adult criminal justice significant emphasis on ‘retribution and deterrence’

Law in Context, Vol 37, Issue 1, 2020 101 ISSN: 1839-4183 system (Cunneen and White 2011, p.111). The youth justice age of criminal responsibility ‘is 12 years in Canada and the Netherlands; 13 years in France; 14 years in Austria, reforms over the last thirty years. What should be noted Germany, Italy and many Eastern European countries; 15 islandscape that these in Australiachanges are broadly not simply has undergone dichotomous significant – it is years in Denmark, Finland, Iceland, Norway and Sweden; not the case that a reform is clearly in the ambit of either 16 years in Portugal, and 18 years in Belgium and Lux- the progressive or welfare model. Notwithstanding, the embourg’ (Youth Justice Fact Sheet 2011-2012, p.2). In overall trend of youth justice reforms in Australia has an international study of over 90 countries it appeared been towards punitive policies, in particular, in the area of sentencing. This next section considers the Victorian criminal responsibility of 12 years of age or higher (Youth Justicethat a significant Fact Sheet majority 2011-2012, (68%) p.2). had a minimum age of for the 2017 reforms discussed further below. Many key stakeholders have therefore advocated for rais- youth justice system specifically, to provide greater context ing the minimum age of criminal responsibility in Victoria, 2.1 THE VICTORIAN YOUTH JUSTICE SYSTEM and more broadly, Australia (Crofts 2019, p.26-40). For Currently, the Victorian youth justice system adopts example, as recently as July 2020 media advocates across elements of each of the models discussed above. Both the Australia called for states and territories to enact legislation Children, Youth and Families Act 2005 (Vic) (CYFA) – the that would raise the age of criminal responsibility from key legislative instrument governing youth justice in Vic- the age of 10, to the age of at least 14 years through the 7 toria - and youth justice policies, place strong emphasis ‘#RaiseTheAge’ campaign. This campaign coincided with on diversion and restorative justice, impose what may the Council of Attorneys-General meeting on 27 July 2020 be described as ‘punitive’ sanctions, and ensure that a where the issue of the age of criminal responsibility was on the agenda. The campaign consisted of advocates using these models appears to be the ideal, as they are not to be both mainstream and social media to urge the Attorneys- regardedyoung offender as ‘opposite is provided ends of with a continuum’ due process. but Conflatingrather can General to raise the age of criminal responsibility to 14 operate as a compromise to each other (Fox and Freiberg as a matter of priority. Between January and August 2020 8 1985, p.828). there were 106 print media articles referencing the issue of raising the age of criminal responsibility with 60 of those articles published in July. The slogan ‘#RaiseTheAge’ 2.1.1 THE CHILDREN, YOUTH AND FAMILIES ACT 2005 appeared in hundreds upon hundreds of tweets on the (VIC) The CYFA begins by distinguishing the capacity of a of the phrase in July and August 2020. Unfortunately, the child by reference to his or her biological age. Pursuant Councilsocial media of Attorneys-General platform Twitter with held a thatsignificant further grouping review to section 3 of the CYFA was required before replacing ‘the current system should at the time of the alleged commission of the offence was the age be lifted’ and therefore no change to the age of under the age of 18 years child but is of defined or above as ‘a the person age of who 10 criminal responsibility was made Australia wide (SBS years’. This age threshold has been criticised by the United News 2020). Only the Australian Capital Territory has Nations Committee on the Rights of the Child, who argue announced its commitment to raise the age of criminal that the criminal age of responsibility should be 12 years responsibility from 10 to 14 in accordance with United of age. Anything below that is considered as internationally Nations standards (Allam 2020). unacceptable. Raising the age of criminal responsibility to 12 years of age, would be consistent with existing brain Therefore, pursuant to the current framework in Vic- development research, and with the experience in other toria, a child under the age of 10 years of age cannot be comparative jurisdictions. For example, the minimum charged with a criminal offence. For those children aged

7 See Raise The Age, ‘Keep Kids in the Community’ https://www.raisetheage.org.au. 8 This analysis was limited to mainstream publications in Australia and did not include an analysis of alternative mediums.

102 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 between 10 and 14 years of age there is a rebuttable than an adult would receive. Should the court for example, presumption – referred to at common law as doli incapax ‘determine that the young person appearing for a criminal – which deems the young person as being incapable of matter was in need of wardship, and thus declare them a committing a criminal act.9 The prosecution may rebut ward of state until their eighteenth birthday’ (Cunneen this presumption by demonstrating that the young accused and White 2011, p.108). 10 is capable of distinguishing between right and wrong. Further, the incorporation of the youth justice system For the prosecution to do so, several elements need to provisions into the CYFA also indicates that the legislation was not designed to address the justice element of youth that the young accused had enough appreciation of the justice (Ogloff and Armytage 2017, [6.1.1]). The lack of a be established. The first is that there must be evidence wrongness of the act in general terms (Crofts 2018). The clear and comprehensive youth justice framework means second requirement is that such evidence may not simply that there is no express statement of the principles and be the same that will demonstrate that the offence was objectives that are intended to govern youth criminal justice. Where the CYFA does refer to principles – for ex- the evidence must demonstrate that the young accused ample, where a decision maker has to consider the best committed (Crofts 1998). The final requirement is that understood that the act in question was ‘seriously wrong, interests of the child – these do not appear to apply to those as opposed to something merely naughty or mischievous’ parts that deal with young offenders and the criminal law. (Crofts 1998, p.186). In Australia, this is the typical formulation of the doli incapax presumption, however some adopted by relevant parties in the youth justice system. jurisdictions focus on the young accused’s capacity to This split approach influences the policies and practices 11 know right from wrong, rather than actual knowledge. overhaul the current framework in Victoria – and other The CYFA further contains provisions that stipulate jurisdictionsFor an efficient that system do not however, separate there their is childa crucial protection need to the relevant trial procedures and standard of proof12 legislation from their youth criminal justice legislation16- sentencing orders13 and governs the appeals process,14 CYFA and parole.15 Embedding the legislative framework for and secondly, appropriately articulates the objectives youth justice into the CYFA does raise a number of chal- and principlesdevise a system of youth that justice is first, by separatedrawing onfrom the the welfare, lenges. The CYFA is predominantly concerned with the justice and restorative justice models as needed. Enacting separate legislation dedicated to the Victorian Act is dedicated to the welfare needs of young persons youth justice system would also bring Victoria in line with welfare interests of children and a significant part of the through provisions that govern child and family services, other Australian jurisdictions, and other cognate interna- child protection and out of home care. This means that the tional jurisdictions, such as Canada. The majority of states welfare tone of the CYFA permeates into the discussion and territories in Australia have independent youth justice of youth offending. Although this may seem innocuous, legislation enacted.17 For example, in Western Australia, the integrating issues of welfare and offending may result in Young Offenders Act 1994 a far longer ‘sentence’ being imposed on a young offender (WA) is specifically concerned 9 CC v DPP [1995 2 All ER 43 at 48; R (a child) v Whitty (1993) 66 A Crim R 462 (Harper J). 10 C v DPP [1995 2 All ER 43 at 48; R (a child) v Whitty (1993) 66 A Crim R 462 (Harper J). 11 Criminal Code Act 1899 (Qld), s 29(2). 12 Children, Youth and Families Act 2005 (Vic), Part 5.2. 13 Children, Youth and Families Act 2005 (Vic), Part 5.3. 14 Children, Youth and Families Act 2005 (Vic), Part 5.4. 15 Children, Youth and Families Act 2005 (Vic), Part 5.5. 16 See, Children and Young People Act 1999 (ACT). 17 Young Offenders Act 1997 (NSW); Juvenile Justice Act 1992 (Qld); Young Offenders Act 1994 (WA); Young Offenders Act 1993 (SA); Youth Justice Act 1997 (Tas) and Juvenile Justice Act 2005 (NT). The Australian Capital Territory does not have a separate juvenile justice from the welfare framework, see Children and Young People Act 1999 (ACT).

Law in Context, Vol 37, Issue 1, 2020 103 ISSN: 1839-4183 with ‘the administration of juvenile justice’. Section 6 of both of which are designed to facilitate the interpretation the Young Offenders Act 1994 (WA) states that the Act: of legislation in line with a number of relevant values ‘embodies the general principles of juvenile justice, for and statements that Parliament wishes to prioritise. The dealing with young persons who have, or are alleged to have, Preamble includes the following values and statements: committed offences; ensures that the legal rights of young per- ‘Society has a responsibility to address the developmental sons involved with the criminal justice system are observed… challenges and needs of young persons …[and that] the minimises the incidence of juvenile crime; punish and manage young persons who have committed offences; rehabilitates youth justice system should take into account the interests young persons who have committed offences; integrates young of victims and ensure accountability through meaningful persons who have committed offences into the community and consequences, rehabilitation and reintegration.’ The Dec- ensures that young persons are dealt with in a manner that laration of Principles includes the following principles: ‘[t] is culturally appropriate and which recognises their cultural he youth justice system must be separate from the adult 18 identity’. system and must be based on the principle of diminished Other examples of youth justice legislation also in- moral blameworthiness or culpability…[and] within the clude references to restorative justice. For example, the limits of fair and proportionate accountability, interventions objectives of the Youth Justice Act 1997 (Tas) are, inter should reinforce respect for societal values; encourage the alia, designed to ‘ensure that, wherever practicable, a repair of harm done; be meaningful to the young person; youth who has committed an offence is provided with respect gender, ethnic, cultural and linguistic differences; appropriate opportunities to repair any harm caused by and respond to the needs of Aboriginal young persons the commission of the offence to the victim of the offence and young persons with special requirements.’ 21 What and the community and to reintegrate himself or herself is notable about these principles and values is that they 19 in the community.’ draw on research that prioritises the best interests of the In Canada, youth justice is governed by the Youth child and, that they are periodically reviewed to ensure Criminal Justice Act (S.C. 2002, c.1) (YCJA). Pursuant to section 3 of the YCJA, the principles that are to govern the The existing provisions under the Victorian CYFA on that they are sufficiently contemporary. operation of the Act are that: ‘the youth criminal justice the other hand, have not been modernised (Ogloff and system is intended to protect the public; the criminal jus- Armytage 2017, [6.1.1]). For example, the Act does not tice system for young persons must be separate from that expressly incorporate international covenants, such as of adults, must be based on the principle of diminished the UNCRC to which Australia is a signatory. The UN- moral blameworthiness or culpability; must be within the CRC includes guidelines for dealing with children in the limits of fair and proportionate accountability and special criminal justice system. Instruments such as these could considerations apply in respect of proceedings against 20 young persons. The YCJA for the large part, adopts the responses, and in line with the justice model in terms of processesbe described governing as fitting youth the justice. welfare model in terms of possess a reduced level of maturity and therefore ac- Some principles are made express in section 23 of the countabilitywelfare model, mechanisms reflecting forthe suchfact thatoffenders young need offenders to be Charter of Human Rights and Responsibilities 2006 (Vic). response to that level. Further, the Canadian youth justice These include that an accused child who is detained must framework emphasises rehabilitation and reintegration, be: segregated from adults; brought to trial as quickly as and timely intervention. What is notable about the YCJA is possible; and treated in a way that is appropriate for his that it contains a Preamble and a Declaration of Principle, or her age.22 However, it is notable that the CYFA omits

18 Young Offenders Act 1994 (WA), s 6. 19 Youth Justice Act 1997 (Tas), s 4(i). 20 Youth Criminal Justice Act (S.C. 2002, c.1), s 3. 21 Youth Criminal Justice Act (S.C. 2002, c.1), s 3. 22 Charter of Human Rights and Responsibilities 2006 (Vic), s 23.

104 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 any reference to these or comparable rights. Equally, the 1989, p.283). Thirty years later, one would be forgiven CYFA is silent on incorporating Australia’s broader human for thinking that the comments had been uttered recently rights obligations under the International Convention on to describe the Victorian experience of the youth justice the Elimination of All Forms of Racial Discrimination and system, media reporting of it and reforms being enacted the International Covenant on Civil and Political Rights by the State Government. which can also play a role in shaping the youth justice Media reporting on issues relating to youth justice and system (Victorian Law Reform Commission 2009, p.116). youth crime is not a new phenomenon, but the frequency Although the CYFA in which such reporting occurs has arguably increased in since its enactment, these reforms have been focussed recent years in Victoria. This may be explained in part by narrowly, enacting changes has undergone to the child significant protection reform and notable characteristics of youth offending, such as visibil- out of home care provisions (Ogloff and Armytage 2017, ity. Young persons are easily noticed and may therefore [6.1.1]). In respect of its criminal law framework, the easily gain attention by media outlets (Clifford and White CYFA has largely been ignored. Since its introduction the 2017, p.128). only change in respect of the criminal law jurisdiction Further, in a news environment saturated with violence, has been the introduction of group referencing (Ogloff acts of violence involving children – and particularly where and Armytage 2017, [6.1.1]). This is partly because the the children are offenders – are deemed to be newswor- consensus is that the current system operates well (Ogloff thy. Historically, discussions of children and crime in the and Armytage 2017, [6.1.1]). media emphasised the children’s inherent innocence. As discussed above, once ‘childhood’ was socially constructed, - the notion of children as ‘evil’ was not prominent (Jewkes tiveHowever, of best practice. in addition For to example, the challenges according identified to the above, 2016 2015, p.67). However, with time, this perception of youth Lancetit should Commission be noted that on Adolescent the existing Health system and is notWellbeing reflec, shifted, and the mainstream media developed stereotypes criminal offending by youth requires a scaled and gradu- of young people. These ranged from the ‘ideal young per- ated response in recognition of the developmental needs son’ – someone who is ‘healthy, wealthy and fun-loving’ of adolescence (Patton et al., 2016, p.242). The CYFA does to ‘young people as victims’(Clifford and White 2017, not attempt to provide graduated responses to offending, p.128). Of particular interest, is the media stereotype of and thereby fails to recognise the ‘understanding of brain ‘young people as threats’ which includes narratives of development’ (Ogloff and Armytage 2017, [6.1.1]). young persons ‘challenging convention: lack of respect for authority, drugs and alcohol, alternative dress/hair/ 8. THE MORE THINGS CHANGE …YOUTH music, juvenile offenders: vandals, hoons, larrikins, youth OFFENDING IN VICTORIA: PERCEPTION VS gangs’ (Clifford and White 2017, p.128). These labels are REALITY not new and have been adopted by mainstream media for several decades. Interestingly, they tend to be based 3.1 YOUTH OFFENDING: PERCEPTION on ‘myths that go back several centuries’ (Clifford and White 2017, p.139). In 1988, Freiberg, Fox and Hogan wrote that ‘notions of a “juvenile crime wave” about to engulf the community In Victoria - as the 2018 State election drew nearer have wide popular currency. It seems to be commonly - reports of a youth crime epidemic and out of control believed that youth commit a disproportionately large young offenders reached a heightened level. This exten- number of serious personal and property offences, or that new legislation and programs lead to an increase in youth the community’s perception of youth offenders and of the sive media coverage resulted in a negative influence on crime, or that society is getting soft on its delinquents, youth justice system more broadly (Legislative Council and that tougher institutions and harsher penalties Parliament of Victoria 2018, [1.4.2]). Further, as reported would help curb youth crime’ (Freiberg, Fox and Hogan in the 2018 ‘Inquiry into youth justice centres in Victoria’

Law in Context, Vol 37, Issue 1, 2020 105 ISSN: 1839-4183 erroneous media narratives can ‘perpetuate negative article entitled ‘Andrews slammed for soft treatment of stereotypes that case young people as something to be thugs’, Opposition spokesperson Georgie Crozier is quoted feared and youth offending as an overwhelming problem’ as often criticising the current Premier of Victoria Daniel (Legislative Council Parliament of Victoria 2018, [1.4.2]). Andrews: ‘Once again youth offenders are calling the shots, In particular, between the years of 2016- 2018 there and Daniel Andrews goes weak at the knees’(Minear 2016, was an emphasised focus by the Herald Sun and The Age p.1). Policy solutions are offered – longer terms of impris- - the two largest mainstream forms of print media in Mel- onment; sending young offenders to adult prison; doing bourne, Victoria’s capital - on young offenders and the away with policies that are too focused on the welfare of youth justice system, with consistent calls for a ‘tougher’ the young accused (Willingham, Hall and Lee 2016, p.1). on crime approach to be adopted by both State Govern- Interestingly, the Victorian experience of youth offending Herald Sun is a even attracted attention from federal politicians who have tabloid newspaper whereas The Age is in ‘compact form’; limited responsibilities for youth offending. For example, ament term and used criminal to describe justice the printingofficials. of The broadsheet quality Minister for Home Affairs Peter Dutton, has stated that newspapers in a tabloid format (Rowe 2011, p.454). Table ‘Andrews should consider resigning if he does not admit 1 demonstrates the total number of stories published by Victoria has a problem with African gang violence and TABLE 1. Total number of reports on issues relating to youth commit to solutions’ and that his ‘left-wing approach to justice the law and order system’ has resulted in youth violence in 267 (81% of which presented young offenders in Victoria (Preiss, Hunter and Koziol 2011, online). Dutton’s Herald Sun a negative light (n=217)) comments followed criticisms made by Prime Minister Malcolm Turnbull about the way in which the Andrews’ 162 (73% of which presented young offenders in The Age a negative light (n=119)) government was addressing gang and youth violence in Victoria (Preiss, Hunter and Koziol 2011, online). the Herald Sun and The Age from January 2016 to January This style of reporting on youth offending is consistent 23 2018 on issues relating to youth justice: with Sasson’s Faulty System frame; a frame commonly The articles published on issues concerning youth adopted by newspapers when reporting on crime. Ac- justice overwhelmingly presented young offenders as cording to this frame, offenders perpetrate crime because ‘juvenile thugs’ (Quill 2016, p.23), ‘out of control’ (Hud- they ‘know they can get away with it’ (Sasson 1995, p.14). son and Hosking 2016, p.1), ‘young crims’ (White 2016, Those responsible for enforcing the law are typically por- p.21), ‘hard core youths’ (Butler 2016, p.1), and ‘jailed trayed as being restrained by ‘liberal judges’ and prisons teens’ (Willingham, Hall and Lee 2016, p.1). Further, these as merely ‘revolving doors for serious offenders’ (Sasson newspaper reports typically present youth offending as 1995,p.14). Further, the criminal justice system is described a ‘sausage machine…arresting the same, again and again’ as one that is ‘riddled with loopholes and technicalities (Hudson and Hosking 2016, p.1) and that the ‘extreme that render punishment neither swift nor certain’ (Sasson nature of [youth] offending’ has increased (Buttler 2016, 1995,p.14). Public safety can therefore only be attained p.1). through the increase of swiftness, certainty and severity In this context, it is interesting to note that such of punishment. Equally, loopholes and technicalities that newspaper reports attribute youth offending to a broken prevent the imprisonment of an offender ought to be criminal justice system, suggesting that the only appro- abolished (Sasson 1995,p.14). Table 2: Faulty System and priate solution is for the State Government to respond Youth Justice, demonstrates the frequency of the Faulty immediately by enacting reforms that are tougher. This System frame adopted by either the Herald Sun or The

23in This turn information reflects was the obtained political through discourse. a search of the For Factiva example, database. in Each an value represents individual news reports on the issue of youth justice and therefore the data excludes duplications. A date range of 2016 to 2018 was used to ensure that no newspaper article was excluded. The limitation of these research results is that they allow for only a textual analysis, therefore no analysis of images or placing of headlines was omitted.

106 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

TABLE 2. Faulty System and Youth Justice draws on, inter alia, statistical data from a number of dif- Number of Uses as a Percentage of ferent sources in recognition that such data can be biased. Total Number of Articles Term advocating for a Herald Sun The Age 3.2 YOUTH OFFENDING: REALITY Faulty System (Total n = 267) (Total n = 162) Despite the media’s labelling of crimes involving young Lenient 27% (n = 72) 26% (n = 42) people as a ‘crime wave’ (Chalkley-Rhoden 2017, online), Soft 42% (n =112) 17% (n = 26) actual youth offending is minimal. In fact, according to the Weak 67% (n = 178) 46% (n = 75) Sentencing Advisory Council, ‘the vast majority of children and young people do not commit crime’(Sentencing Inadequate 31% (n = 83) 17% (n = 27) Advisory Council 2016, p.2). In 2015, only 1.4% (7,507) of persons aged between 10 and 17 years of age were Number of Newspaper Articles processed by police (Sentencing Advisory Council 2016, Terms for ‘tougher’ Herald Sun The Age p.2). Only 0.6% of these matters, proceeded to the Chil- reform (Total n = 267) (Total n = 162) dren’s Court and were sentenced, of which 0.2% received Tougher 26% (n =72) 20% (n = 34) a sentence of imprisonment (Sentencing Advisory Council Accountable 5% (n = 16) 5% (n = 9) 2016, p.2). In 2015, the total number of young persons sentenced by the Children’s Court were 3,341. This number represents a 47% decline since 2008-2009 when 5,385 Number of Newspaper Articles young persons were sentenced by the Children’s Court. Terms advocating for Herald Sun The Age moderate reform (Total n = 267) (Total n = 162) In the year ending June 2017, the Crime Statistics Agency reported that there were 7,497 offenders under the age of Smarter 0 (0%) 11% (n = 18)

Strengthen 0.4% (n = 13) 2% (n = 4) the same period the year before (Crime Statistics Agency, Evidence-based 18 which signified an increase of just 1.9% compared to 0 (0%) 3% (n = 5) https://www.crimestatistics.vic.gov.au). However, it may analysis be interesting to note that offenders under the age of 25 Age and the frequency of media advocating for moderate more broadly as a group had been ‘decreasing steadily reforms in the context of youth justice. The culmination of this mainstream media reporting Agency, https://www.crimestatistics.vic.gov.au). Further, as a proportion over the past five years’(Crime Statistics on youth offending leaves the audience believing that the proportion of alleged offenders aged 10-19 has also youth crime is rising exponentially which can have an effect on legislation, policies and police practices. ‘[T]he been steadily decreasing over the last five years: intersection between media reporting and the political use of community fears over law and order constructs a certain reality that inhibits our understanding of the nature of youth offending and constrains our conceptions of possible responses. Within this climate, mandatory sentencing, increased police powers, and restrictions on community-based sentences appear popular, while community development and social justice strategies are portrayed as intellectual and “soft”’ (Cunneen and White 2011, p.82). FIGURE 1. Unique alleged offenders, proportion of unique al- As discussed in the next section, the media reporting leged offenders by age group, July 2012 to June 2017 (Crime of youth crime does not necessary accord with reality. It Statistics Agency, https://www.crimestatistics.vic.gov.au)

Law in Context, Vol 37, Issue 1, 2020 107 ISSN: 1839-4183

It should be noted that youth offending can differ from holistic, integrated thinking to address health, mental adult offending in ways that may increase young persons’ health, disability, education and employment needs in representation in crime statistics. For example, Cunnen, order to reduce reoffending’ (Ogloff and Armytage 2017, White and Richards suggest that youth offending tends to p.11). What ought to be avoided is the implementation be committed in groups in public areas, and typically in of ‘law and order’ reform; focusing on police powers, close proximity to their place of residence (Cuneen, White tougher penalties, and the adultification of the youth and Richards 2015, p.55). Further, the offences tend to justice system, which does little to address the underly- be ‘public, gregarious, and attention-seeking…episodic, ing causes of youth offending. The next section considers unplanned, opportunistic’ (Cuneen, White and Richards the Victorian State Government’s 2017 reforms through 2015, p.55). Consequently, such visible offending is likely to result in arrests. Further, several young persons may are not in the best interests of the young person and do be arrested for the one same offence, which may give the notan ‘adultification’ attempt to address lens. theIt is complicated argued that socioeconomicthese reforms appearance that there are more offenders or offences framework within which youth offending occurs. being committed than there are in practice. For example, a group of young offenders could all be arrested with af- 9. THE ADULTIFICATION OF THE VICTORIAN YOUTH fray, giving the appearance of a greater level of offending, JUSTICE SYSTEM despite the fact that actual participation in the offending In 2017 the Victorian State Government introduced could vary considerably. reforms to the youth justice system. The primary piece of legislation, the YJR, was designed to implement mecha- form of youth offending are property offences such as nisms for addressing ‘community concerns about crimes theft,Official vandalism, police statisticsand trespassing, indicate asthat well the as most fare common evasion committed by children and young people’.24 Although (Richards 2011, p.3). Property offences are also more Attorney-General Pakula acknowledged that there was likely to have higher reporting rates, in part because of an overall reduction in the youth crime rate, he explained the need to support insurance claims. In contrast, young that the focus of the reforms was a small number of youth persons are under-represented in the more serious of- offenders who were ‘entering the criminal justice system fence categories such as homicide and rape (Victoria early and reoffending more often and in an alarmingly Police 2013/2014, p.14). serious manner’.25 The YJR reforms included: (a) the cre- Youth offending is a complex problem which is very ation of a presumption that certain youth offences would much connected with the environment in which children not be heard in the Children’s Court; (b) that offenders are raised. In a review conducted into youth offending in aged between 18-21 years of age who commit serious - offences would be sent to adult prison in place of youth tics and circumstances including ‘poverty/socioeconomic detention; and (c) increasing the maximum penalty or disadvantage,2017, Armytage trauma/grief, and Ogloff identified childhood relevant abuse, characteris disrupted period of detention that a Children’s Court and other education, family circumstances, including abusive or of- higher courts could impose.26 The YJR Act also introduced fending parents, and the involvement of Child Protection reforms that would facilitate diversion for young offend- involvement and exposure to family violence’ (Ogloff and ers and introduced a new sentencing option that would Armytage 2017, p.6-9). These factors are almost entirely allow for youth offenders to be subjected to an intense supervision order requiring young offenders to participate for the youth justice system. ‘It necessitates the need for in education, training or work.27 outside the child’s control and pose significant challenges 24 Attorney-General Martin Pakula, Children and Justice Legislation Amendment (Youth Justice Reform) Bill 2017, 8 June 2017, Assembly, Second Reading Speech, 1818. 25 Attorney-General Martin Pakula, Children and Justice Legislation Amendment (Youth Justice Reform) Bill 2017, 8 June 2017, Assembly, Second Reading Speech, 1818. 26 Children and Justice Legislation (Youth Justice Reform) Act 2017 (Vic), s 1. 27 Children and Justice Legislation (Youth Justice Reform) Act 2017 (Vic), s 13.

108 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

Supreme Court decisions granted youth offenders due of the youth justice system that has been particularly process rights.28 However, a result of this was a more experiencedThese reforms in Victoria exemplify in recent the progressive years, and adultification in similar ju- formalised youth justice system that now closely mirrored the adult criminal justice system. The implicit consequence that is commonly utilised in a number of areas concern- of this shift was that youth offenders were simply viewed ingrisdictions children over including recent decades.social work, Adultification child psychology is a concept and as different criminal defendants. psychiatry, child welfare and youth justice to describe In the decision of In re Winship, the United States Su- the ‘processes that act to impart adult responsibilities, preme Court further provided a number of due process behaviours and treatment upon children’ (Arnett 2018, protections because youth offenders were no longer ‘re- p.410). This adult treatment of children can occur by a ceiving the treatment that was supposed to result from number of different individuals and institutions, including the informality of the juvenile court’ (Bolin 2014, p.16). communities, media, law enforcement, legal institutions, This sentiment was echoed in the decision of Kent v United government policies and even schools (Arnett 2018, States ‘there is evidence, in fact, there may be cause for concern that the child receives the worst of both worlds; is a framework that helps understand the application that he gets neither the protections accorded to adults nor p.410). In the context of youth justice, ‘adultification’ of adult criminal justice policies and practices to child the solicitous care and regenerative treatment postulated offenders, and is commonly made up of two ‘waves’. The for children.’ 29 The Supreme Court attempted to address this concern increased penalties for youth offenders, in particular, for by granting due process guarantees to youth offenders. first, the ‘tough on crime’ wave, sees the introduction of serious young offenders (Bernard and Kurlychek 2010, In this way, the youth justice system underwent a trans- p.71-94). This wave either follows, or is complementary formation from an informal institution with substantial discretion and focusing on the best interests of the child, is typically manifested by the progressive provision of due to a system that was not too dissimilar from the adult to, the second or ‘due process’ wave of adultification. This process protections to young offenders. criminal justice system and its focus on due process and Although enacting due process protections may appear adversarial proceedings. This instigated the second wave that doing so alters the nature of the youth justice system The two ‘waves’ are typically interrelated. The ‘tough (Bolinto be beneficial 2014, p.16). to young The youth offenders, justice it system has been was observed initially of adultification, the ‘tough on crime’ wave. a construct of the parens patriae ideology, indicating a in response to a group of youth offenders that have com- level of accountability by the system itself (Ward and mittedon crime’ serious wave ofoffences. adultification Media tendsreports to beand implemented community Kupchik 2014, p.85). Its focus was largely on facilitating concerns about these group of offenders, alongside with rehabilitation and reintegration concepts that uphold the political agendas, culminate in the introduction of new principle of ‘the best interest of the child’. For this reason, laws that either create new offences or result in more the system developed more as a civil proceeding rather punitive sanctions. These reforms align the youth justice than a criminal one, and due process protections were system more closely with the adult criminal justice system considered largely unnecessary. Providing such protec- and include ‘changes to the purpose of the youth justice tions signals a departure from the traditional objectives system, changes to the youth court process, changes to the of the youth justice system, and increasingly erodes the available dispositional outcomes, and changes to jurisdic- distinction between it and the adult system. tion’ (Bolin 2014, p.21). These are then accompanied or This trend has been observed in other cognate juris- followed by due process protections that are a reaction dictions. For example, in the United States, a number of

28 See In Re Gault 387 U.S 1 (1967); In Re Winship 397 U.S 358 (1970); Breed v Jones 421 U.S. 519 (1975). 29 Kent v United States (1966) 274 F.2d 556, 556-559.

Law in Context, Vol 37, Issue 1, 2020 109 ISSN: 1839-4183 to the increasing adversarial nature of the process, and policies were partially enacted in response to heightened the severity of its consequences. media reporting, the effect of which is a disproportion- Although the Victorian youth justice system attempts to strike a balance between a philosophy that prioritises ally punitive youth justice system. This is specifically a child’s best interests and one that adequately punishes, were enacted by the Victorian State Government. This exemplified by the ‘tough on crime’ 2017 reforms which laws such as the YJR Act may encourage a prioritisation may result in a system that does not clearly differentiate of the tough on crime approach over others. For example, between child and adult offenders, ignoring best prac- by increasing penalties that could be imposed on youth tice and undermining the best interests of the child. Far offenders, the YJR Act diminishes the distinction between from addressing these concerns, the enactment of due adult and youth offenders. In creating a presumption process protections may serve to entrench rather than against hearing certain matters in the Children’s Court, ameliorate the adversarial nature of the proceedings. the YJR Act exposes young offenders to potentially longer It is hoped that this article serves as a caution for those periods of incarceration. Although well-grounded in deter- wishing to develop and enact youth justice policies and rence and incapacitation theory, such presumptions have negative repercussions because they do not address the underlyinglegislation that complexities adultification of youth of the offending. youth justice Rather, system, the while superficially appealing, may do little to address the example, young offenders whose matters are referred focus should be on enacting youth justice policies that tospecific adult complexitiescourt instead and of children’scause of youth court offending. are typically For draw on both the rehabilitative and restorative justice sentenced for longer periods of time and demonstrate models, and crucially, encourage a multi-faceted approach higher recidivism rates than their counterparts (Bishop to ‘address health, mental health, disability, education and 2006, p.653). Research in cognate jurisdictions indicates employment needs’ (Ogloff and Armytage 2017, p.11). that young offenders sentenced by an adult court are more likely to be imprisoned, and for longer periods than 11. REFERENCES if sentenced in the childrens’ court (see, Kupchik 2006). 1. Allam, L ‘Austraian Capital Territory votes to raise the age of They also face the potential of being transferred to adult criminal responsibility from 10 to 14’ The Guardian, Sydney, prison, which in turn, can expose the young offender to an 20 August 2020 increased risk of physical and sexual victimisation (Forst, 2. Aries, P. 1979. Centuries of Childhood: A Social History of Fagan and Vivona 1989, p.9). Family Life. Harmondsworth: 3. Penguin 10. CONCLUSION 4. Attorney-General Martin Pakula. Children and Justice Legis- lation Amendment (Youth Justice Reform) Bill 2017. 8 June With an increased focus on punitiveness and account- 2017. Assembly, Second Reading Speech. - 5. Australian Law Reform Commission. 1997. Seen and Heard: tem to punish the young offender rather than to facilitate (1997) Final Report ability, the tendency is for an adultified youth justice sys Priority for Children in the Legal Process his or her rehabilitation. In this way, amendments to the 84. Sydney: Australian Law Reform Commission youth justice system such as the YJR Act, can result in a 6. Bernard, T. J. and Kurlychek, M. C. 2010. The Cycle of Justice. (2nd ed). New York: Oxford University Press. system, and undermine the best interest of the child. 7. Bishop, D.M. 2006. Public opinion and juvenile justice policy: modification of the original objectives of the youth justice This article has mapped the Victorian youth justice sys- myths and misconceptions. Criminal Public Policy, 5(4), 653 tem and, in particular, considered how the two dominant 8. theoretical models – the welfare model and justice model - - co-exist in an attempt to provide a holistic framework. sertation),Bolin, R. M. retrieved (2014) Adultificationfrom https://scholarcommons.sc.edu/ in Juvenile Corrections: A Comparison of Juvenile and Adult Officers (Doctoral dis An examination of the perception and reality of youth of- etd/2791 fending in Victoria demonstrates that recent youth justice 9. Butler, M. ‘Hard Core Youths in Crime Surge’ Herald Sun, Melbourne, 15 December 2016.

110 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

10. Carrington, K. and Pereira, M. 2009. Offending Youth: Sex, 27. Freiberg, A., Fox R., and Hogan, M., Sentencing Young Offenders, Crime and Justice. Sydney: Federation Press Sentencing Research Paper no. 11 (Australian Law Reform 11. Chalkley-Rhoden, S. ‘Violent Youth Crime in Victoria “Un- Commission and Commonwealth Youth Bureau, 1988) precedented”, Police Commissioner Graham Ashton Says’ 28. Fullerton, R.S. and Patterson, M.J. ‘Crime News and Privacy: ABC Online, Sydney, 29 January 2017. Comparing Crime Reporting in Sweden, the Netherlands, and 12. Clifford, K. and White, R. 2017. Media and Crime: Content, the United Kingdom’ in Julian Petley (ed) Media and Public Context and Consequence. South Melbourne: Oxford Uni- Shaming: Drawing the Boundaries of Disclosure, (Reuters versity Press. Institute for the Study of Journalism, 2013) 13. Coppins, V., Casey, S. and Campbell, A. 2001. The Child’s Best 29. Hakkert, A. 1998. ‘No More Excuses: A New Approach to Interest: A Review of Australian Juvenile Justice Legislation. Tackling Youth Crime in England and Wales: A Summary’ The Open Criminology Journal 4, 23 European Journal on Criminal Policy and Research 6(2), 279 14. Crime Statistics Agency, https://www.crimestatistics.vic. 30. Hoge, R. 2001. ‘A Case Management Instrument for Use in gov.au Juvenile Justice Systems’ Juvenile and Family Court Journal 15. Crofts, T. 2015. A Brighter Tomorrow: Raise the Age of Criminal 52(2), 25 Responsibility. Current Issues in Criminal Justice 27(1), 123 31. Hudson, F., and Hosking, W. ’15 and Out of Control’ Herald 16. Crofts, T. 2018. Prosecuting Child Offenders: Factors Relevant Sun, Melbourne, 17 July 2016. to Rebutting the Presumption of Doli Incapax’. Sydney Law 32. Jewkes, Y. 2010. ‘The Construction of Crime News’, in Chris Review 40(3), 339 Greer (ed) Crime and Media: A Reader (London, Routledge) 17. Crofts, T. ‘Doli Incapax: Why Children Deserve its Protection’ 33. Jewkes, Y. 2015. Media and Crime, (SAGE Publications, 3rd ed,) (2003) 10(3) Murdoch University Electronic Journal of Law 34. Kupchik, A. 2006. Judging juveniles: Prosecuting adolescents in adult and juvenile courts. (New York University Press) html 35. Mathews, B. 2000. ‘Children’ Criminal Responsibility In 18. Crofts,

Law in Context, Vol 37, Issue 1, 2020 111 ISSN: 1839-4183

43. O'Brien, W. and Fitz-Gibbon, K, ‘The minimum age of criminal 59. Taylor, C, (2016) Review of the Youth Justice System in England responsibility in Victoria (Australia): examining stakehold- and Wales, London: Ministry of Justice ers’ view and the need for principled reform’ (2017) 17(2) 60. - Youth Justice 134 ing Minor Offences Committed by Young People in Tasmania 44. O’Connor, I, (1997) Models of juvenile justice in Alder C (ed.). andTravers, Victoria’ M, ‘Welfare, (2010) 22(1)Punishment Current or Issues Something in Criminal Else? Justice Sentenc 99 Juvenile crime and juvenile justice. Canberra: Australian 61. UN Committee on the Rights of the Child (CRC) 2007 CRC/C/ Institute of Criminology, Research and Public Policy Series GC/10 No. 14 62. Victoria Police, Crime Statistics 2013/2014 45. The 63. Victorian Law Reform Commission, Protection Applications Role of Statistics and Public Opinion in the Implementation in the Children’s Court’ (Final Report, 2009) ofOffice International of the High Juvenile Commissioner Justice Standards for Human, Expert Rights, Workshop 64. on the Administration of Juvenile Justice, (September 2002) Orientations Towards Accountability-Based Juvenile Justice’ 46. Ogloff, J, and Armytage, P, (2017) Youth Justice Review and (2014)Ward, and 11(1) Kupchik, Punishment A, ‘Accountable and Society to 85 What?: Professional Strategy: Meeting Needs and Reducing Offending, Report, 65. White, A, ‘Teens Run into Strife’ Melbourne, 24 Victorian Government Herald Sun, May 2017 47. Palmer, D, and Walters, R, 'Crime prevention camps for 66. White, A; ‘Most Young Crims End up in Big House’ youth 'at risk': Blurring the boundaries of care and control' Herald Sun, Melbourne, 16 December 2016 in Cheryl Simpson and Richard Hil (eds) Ways of Resistance: Social Control and Young People in Australia (1995, Hale & 67. Willingham, R, Hall, B, and Lee, J ‘Back Down on Jailed Teens’ Iremonger) The Age, Melbourne, 30 November 2016 48. Patton et al, ‘Our Future: A Lancet Commission on Adolescent Health and Wellbeing’ (2016) 387 Lancet 2423 LEGISLATION 49. Petley, J, Film and Video Censorship in Modern Britain, (Ed- Charter of Human Rights and Responsibilities 2006 (Vic) inburgh University Press, 2011) Children and Justice Legislation (Youth Justice Reform) Act 2017 50. Preiss, B, Hunter, F and Koziol, M, ‘Andrews and Dutton trade (Vic) blows over youth violence’ 11 January 2011 The Age (Online) Children and Young People Act 1999 (ACT) 51. 2016 Herald Sun, 23 Children, Youth and Families Act 2005 (Vic) 52. Richard,Quill, J, ‘Some K, ‘What Juvenile Makes Thugs Juvenile Should Offender be Identified’ Different 17 October from Crime and Disorder Act 1998, c 37, (UK) Trends and Is- Criminal Code Act 1899 (Qld) sues in Crime and Criminal Justice, Australian Institute of Juvenile Justice Act 1992 (Qld) CriminologyAdult Offenders?’ (February 2011) No.409 53. Juvenile Justice Act 2005 (NT) (2011) 12(4) Journalism 449 Young Offenders Act 1993 (SA) Rowe, D, ‘Obituary for the Newspaper? Tracking the Tabloid’ 54. Sasson, T, Crime Talk: How Citizens Construct a Social Problem, Young Offenders Act 1994 (WA) (Aline de Gruyter, 1995) Young Offenders Act 1997 (NSW) 55. Scottish Home and Health Department. (1964). Children Youth Criminal Justice Act (S.C. 2002, c.1) and Young Persons: Scotland. [Kilbrandon Report]. Edin- burgh: HMSO Youth Justice Act 1997 (Tas) 56. Sentencing Advisory Council, Reoffending by Children and Young People in Victoria (December 2016) Sentencing Ad- CASES visory Council Breed v Jones 421 U.S. 519 (1975) 57. Sentencing Advisory Council, Sentencing Children and Young C v DPP [1995 2 All ER 43 People in Victoria (April 2012) Sentencing Advisory Council Crown Prosecution Service v P (2007) 171 JP 349 58. Smith, R. ‘Inside Melbourne’s wild youth prison’ news.com. au, Sydney, 9 September 2019 In Re Gault 387 U.S 1 (1967)

112 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

In Re Winship 397 U.S 358 (1970) Kent v United States (1966) 274 F.2d 556 R (a child) v Whitty (1993) 66 A Crim R 462 R v JTB (2009) UK HL 20

Law in Context, Vol 37, Issue 1, 2020 113 Received: June 15 2020, Date of acceptance: September 28 2020, Date of publication: December 13 2020, DOI: https://doi.org/10.26826/law-in-context.v37i1.116

At Play in the Field of Dreams: Theorising Attitudes, Perceptions and Practices of Law Students in conjunction with the Reflections of Early Career Commercial Lawyers

By Barry Yau, Senior Lecturer at ANU College of Law, Australian National University, Orcid: https://orcid.org/0000-0002-9511-0311 and David Catanzariti, Honorary Lecturer at ANU College of Law, Australian National University, Orcid: https://orcid.org/0000-0001-9828-2111 ANU College of Law, Australian National University

ABSTRACT Australian law schools are tasked with forming students in their knowledge and understanding of the law, with many

- dominantlystudents aiming linked to to fulfil an “elite” their dreamstier of law of pursuingpractice. Thea legal attitudes career. and Utilising perceptions Bourdieu’s of law conceptual students can tools, also this positively article orconsiders adversely whether shape theiraspirations career ofpath being amidst “real the lawyers” information are significantly at play in the influenced law school by space. motifs Drawing of career on success qualitative pre data, we have applied Bourdieu’s tools to understand undergraduate and practical legal training students’ responses

their law school experiences. With comparisons to contemporary surveys and research on student services for law students,to notions along of career with accomplishment. their wellbeing, Thisthe article is contrasted reasons with that the the reflections assorted ambitionsof early career of law commercial students requireslawyers abouta law

schoolKeywords environment – Bourdieu, promoting practical a legal more training, diversified legal perspective careers, real of law, “real real law” lawyering and “real lawyering”.

Acknowledgements: We thank Vivien Holmes, Alexandra Johnstone, Helen Rodriguez, Courtney Macintosh, Stephen Tang and Pam Foo for their valuable contributions to the research project. We thank Kristoffer Greaves for his insights on Bourdieu’s theory. The research was partly funded by the ANU College of Law Small Grant Scheme. Disclosure statement – No potential conflict of interest was reported by the authors. License – This work is under Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) https://creativecommons.org/licenses/by-nc-sa/4.0/ Suggested citation: Yau, B and Catanzariti, D, “At Play in the Field of Dreams: Theorising Attitudes, Perceptions and Practices of Law Students in conjunction Law in Context, 37 (1): 114-133 , DOI: https://doi.org/10.26826/law-in-context.v37i1.116 Summarywith the Reflections of Early Career Commercial Lawyers”, 1. Part I: Introduction 4.2 Cultural capital and habitus: Bodies in the crowd 2. Part II: Literature review and Bourdieu’s tools 4.3 Far from the madding crowd: Complicity and resistance 2.1 Cultural field 5. Part V Impact of the information gap 2.2 Capital 5.1 Symbolic violence: Taking the next drop 2.3 Habitus 5.2 Language and symbolic power: Upstairs, Downstairs 2.4 Bourdieusian concepts in North America and the UK 6. Part VI Broadening career information to law students 3. Part III: Methodology and limitations 6.1 “I wish that I knew what I know now, when I was younger.” 3.1 Recruitment 6.2 “Always be a first-rate version of yourself, instead of a 3.2 Data collection second-rate version of somebody else.” 3.3 Limitations 4. Part IV: Data analysis 7. Part VII: Conclusion 4.1 Dominance and disposition: The law school space and 8. References student headspace 114 Law in Context, Vol 37, Issue 1; 2020 ISSN: 1839-4183

5. PART I: INTRODUCTION unduly constrain some students’ appreciation of the wider range of career opportunities which are available to them “Now is the winter of our discontent.” than those routinely touted at law school. In law student circles the information gap can also have the intended or Shakespeare, Richard III unintended consequence of reinforcing constricted ideals of career success that gain credence from one cohort to The original object of the research project encompassed the next. Some of the early career commercial lawyers, researching the attitudes of undergraduate law students themselves representing an earlier generation of law (LLB or JD students), practical legal training (PLT) stu- students, stressed the importance of addressing the gap dents),1 to give law students a more realistic and broader picture years or less of commercial law practice) towards the of career paths. Several of these lawyers highlighted their study and and practice early careerof commercial commercial law.2 lawyers The objective (with alsofive struggles with the lingering effect of the information gap included researching the role of legal ethics in the study on their sense of professional competence and worth. and practice of commercial law. The commercial law focus stems from our commercial legal practice background and seemingly entrenched and continually active in the law the fact that commercial/corporate law3 is the most com- schoolThis points environment to the powerful and beyond. influence These of interests interests that deploy are mon legal practice area for entry-level lawyers in Australia various forms of capital (tangible and intangible) promot- (Evers 2011, p. 27). The project was also prompted by ing (as is their right) an elite perspective of lawyering, the principal researcher’s desire to comprehend student void of powerful countering perspectives. The effect of attitudes in the course of designing and convening Com- mercial Practice in the PLT program at the Australian that some students, proudly attesting resistance to elite National University (ANU). We also have an interest in these influences are so dominant that the research found understanding what motivates law students to pursue the narrow ideal of career competency promoted in law careers in commercial law. Nevertheless, we contend the school.influences, still accepted (and even sought to aspire to) research has currency across the legal profession. The phrase “with great power comes great responsi- The central focus of this article is the “information gap”4 bility” underlies our decision to analyse and present the in the law school environment and its impact on students research for this article by utilising the conceptual tools of and lawyers in their perceptions of “real law” and “real French sociologist Pierre Bourdieu. Bourdieu’s theory of the lawyers”. The information gap emerged organically5 as a dominant and recurrent theme arising from the focus to analyse aspects of this research. Bourdieu’s positioning groups and semi-structured interviews in this research. ofcultural actors field in this (Bourdieu social space 1984, according p. 24) is toan their apt framework economic, The information gap symbolises the paucity of informa- cultural and social characteristics helps, to an extent, make tion within the law school environment concerning the some sense of the paradigm where large and powerful law true scope of job opportunities in the legal career market and beyond. The information gap is also a manifestation legitimise a certain notion of professional competence of the hierarchy of career paths promoted (covertly or acrossfirms exert the general dominance student in the body. law Bourdieu’s school environment tools will as to- overtly) in law school. At the outset of the research project, sist in understanding the relationship between students the information gap research theme was not apparent to us. However, it became a consistent theme of discontent promoting and preserving their corporate interests. This amongst some of the research volunteers during the data articleand these will dominant consider the law responsibility firms, who are some responsible other actors for collection phase, which commenced on a winter’s night (for example, academics) have in the law school space to in Canberra in 2014 and ended over a year later. We will raise students’ awareness and impact of the “power-play” contend in this article that the information gap appears to that attempts to shape notions of real law.

1 Admission as an Australian lawyer requires completion of a law degree (LLB or JD) and an approved PLT course, see Hewitt (2015, p. 94). 2 Ethics approval granted 24 December 2013 (ANU HREC Protocol: 2013/717). 3 Corporate law is a subset of commercial law. 4 A term coined by an LLB volunteer in one of the focus groups. 5 Without prompting from the moderator/interviewer.

Law in Context, Vol 37, Issue 1, 2020 115 ISSN: 1839-4183

Part II will explain Bourdieu’s tools and its application (Fisher 2015). In Australia, competition has increased to the research project, and compare it to the current research and insights into student support services and – all specialising in corporate law 6– remaining largely wellbeing in law. It will also provide a review of how constantsignificantly over due many to the decades number (Pinnington of big (or and top-tier) Gray 2007,firms international legal education scholars have utilized the p. 154). In contrast, schools have increased sixfold from French theorist’s tools in their research on the power 1960 to 2018 (Kritzer 2013, p. 455).7 The law student’s school space in Canada, the UK and the USA. Part III will (or their global or medium-tier counterparts) usually com- that large law firms wield amongst students in the law path to securing a graduate position with a top-tier firm research project. Part IV will examine and analyse the (akin to an internship) which students take up during datapresent utilising the methodology Bourdieu’s andtools, identified and present limitations the research of our theirmences undergraduate by being selected degree. in the The firm’s clerkship clerkship application program outcomes regarding the information gap. We will seek to process is very spirited.8 It is largely self-evident that, demonstrate how the information gap narrows the no- even taking into account students who have no interest tion of real law and real lawyers. Part V will highlight the 9 an impact of the information gap on law students and early ever-growing number of future law students will fail in career commercial lawyers and the associated symbolism in pursuing career opportunities with a top tier firm, that Bourdieu articulated. Part VI will consider the real- We contend the competitive environment for law students ism of expanding the law school environment to promote hastheir not quest changed to become since graduateour data collectionlawyers in ended top-tier in firms.2015. a wider range of opportunities, giving students a higher Law school continues to be perceived as an institution level of awareness about their career paths relating to “structured to pump out high powered graduates for their ambitions and dreams. Part VII will conclude with the importance of keeping the dreams real for students “breeds an environment hostile to emotional vulnerability.” aspiring to be real lawyers. (Lanetop-tier 2020, firms” p. amidst43).10 a “culture of competitiveness” that The issue of student autonomy in determining career 6. PART II: LITERATURE REVIEW AND BOURDIEU’S choices, which underlies the information gap, is prevalent TOOLS in research and insights into the wellbeing of Australian law students since 2015. Research by O’Loughlin et al into

law students at Monash University sought to understand Shakespeare, The Taming of the Shrew “O this learning, what a thing it is!” students’the psychological perception distress of their of studiesfirst-year and undergraduate its impact on their future career paths (O’Loughlin et al 2020, p. 30). During the data collection phase, it became apparent Whilst their research did not elicit issues about the direc- that each student, commencing their undergraduate tion students can take in pursuing a career beyond law law studies, discovers before too long (if they were not school, Roberts provides relevant insight as a third year already cognisant of this upon enrolment), that they are JD student at the University of Melbourne (Roberts 2020, p. 40). We support Roberts’ assertion that autonomous There is competition amongst students to obtain high learning is an important aspect of wellbeing, where stu- grades,in a very with competitive failure to field do so with meaning their missingcourse colleagues. out on the dents feel empowered to direct the things that they can control and be actively engaged in (Roberts 2020, p. 40).11 opportunity6 As top-tier firms to have even large apply corporate for clients a job and with undertake a major big corporate law firm transactions, they are sometimes referred to as “corporate firms” in this article. This also aligns with some of the volunteers using “corporate” in connection with top-tier firms 7 As at June 2019, Australia has 38 university law schools, and two non-university law schools offering accredited law qualifications. 8 Selection for clerkship does not guarantee a graduate position. Top-tier firms usually take around four times the number of clerks compared to anticipated graduate positions, perpetuating a further level of competition: “Legal clerkships: what are they, and why are they important?” 22 June 2015 9 Students who choose not to pursue a clerkship with a top-tier/global firm, can still pursue a job with other law firms/organisations, through their graduate programs. Occasionally, a top-tier/global firm will offer a graduate position to someone who has not undertaken their clerkship program. 10 Christian Lane, at the time of writing his article, was a fifth year Law andArts student at Monash University. 11 The University of Melbourne Law School was ranked the top law school in Australia in 2020: https://www.topuniversities.com/university-rankings/university-subject- rankings/2020/law-legal-studies

116 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

We contend further that the autonomy of law students University. Undergraduate law students at the University includes making informed choices about career paths, of Adelaide and the University of Queensland provided a which can be adversely impacted by the competitive ac- positive rating of 79.6% and 71.8%, respectively.13 The ratings given by undergraduate law students at ANU, the We support Holland and Taylor’s arguments that a law University of Sydney, the University of New South Wales student’stivity driven ability by lawto see firms “multiple in the pathwayslaw school for environment. employment national average of 72.1% (the University of Melbourne achievedand Monash a positive University rating were of 72.9% significantly for support less thanservices the (Hollandand career and development” Taylor 2020, post-graduation pp. 113, 117). isWe significant agree with in - theiraddressing assertion the informationthat students gap be to encouraged the benefit ofto students adopt a ate degree, as they no longer offer a LLB – this compares “positive professional identity that links with their inher- tofor the their national JD program, average officially of 68.9% categorised from students as a postgradu in law ent values with a chosen career pathway.” We will contend postgraduate programs). We cannot say to what extent a in this article that a student’s identity can be shaped by perceived lack of support from careers advisors comprised the information available in law school. the relatively low ratings from undergraduate law students The Student Experience Survey (SES) is a suite of Aus- at ANU, the University of Sydney, the University of New tralian Government endorsed surveys for higher education South Wales and Monash University, as the SES data did that “covers the student life cycle from commencement to employment” (Quality Indicators for Learning and Teach- even if tenuously, of these statistics at the very least ing, 2020). Whilst the SES does not provide data about resonatenot drill down with our to that research subset. data, However, which maythe significance, be relatable law students’ career paths, it does provide quantitative to law schools located in Australia’s two largest metro- data about the level of law students’ satisfaction with “support services” which bundles support from “career presence (the University of Sydney and the University of 14 advisors”, along with enrolment systems, learning advice Newpolitan South cities Wales where are the in Sydney, top tier firms and Monash have their University biggest and health services. We acknowledge that the SES survey is in Melbourne). of law students’ satisfaction of support services (especially In light of this, we are drawn to applying Bourdieu’s tools careers advice) is not quite analogous to their satisfaction to examine the law school environment where distinctive of the career information made available to them in the law school space. However, the most recent survey, which activity, giving problematic rise to the information gap draws upon data collected in 2018-2019, provides (even whichpowers impacts (which oninclude law school top tier students firms) driveand beyond. competitive tenuously), a contemporary insight into the attitudes of law students towards the value of career information. 2.1. CULTURAL FIELD All but one of the law students who participated in our research were enrolled at ANU, which also happens to be a member of the Group of Eight, a coalition of Bourdieu’s theory of a “cultural field” is akin to the research-intensive universities in Australia.12 For com- oflaw objective school environment. relations between Bourdieu positions”. defined, It in is analytical a “space parative purposes, and to highlight the relevance of our terms, the cultural field as a “network or a configuration,

(Bourdieuof conflict andand competition”Wacquant 1992, conserving p. 17). Theor transforming existence of undergraduateown research findings law student to law satisfaction schools in largely at Group populated of Eight the structure of relational forces constituting the field universitiescities in Australia, towards we supportcompare services. our research ANU findingslaw students with gave a 61.2% positive rating of support services. Their –positions based on are the objectively structure defined of the distribution by what they of impose power counterparts at these universities provided positive rat- (oron the capital) field’s whose actors possession – “occupants, commands agents or access institutions” to the ings of: 50.7% at the University of Sydney, 55.9% at the - University of New South Wales, and 67.7% at Monash quant 1992, p. 97). Dezalay and Garth have focused on specific profits at stake in the field (Bourdieu and Wac 12 The Group of Eight comprises the University of Sydney, the University of Melbourne, the University of Queensland, the University of Adelaide, the University of Western Australia, Monash University, the University of New South Wales and the Australian National University 13 There was no rating amongst law students for support services at the University of Western Australia due to the number of survey responses being too low. 14 ANU is a three hour drive from Sydney.

Law in Context, Vol 37, Issue 1, 2020 117 ISSN: 1839-4183

- social space. Bourdieu argues that the legitimation of such mercial arbitration (Dezalay and Garth, 1996) and Asian symbols is not the product of a deliberate and purposive legalanalysing systems legal (Dezalay fields (for and example, Garth 2012)) international through com the action or symbolic imposition. Rather, it results from the lens of Bourdieu, postulating that an analysis of a legal fact that powerful agents in the social space (for example

of the social space to deal the aces that create a percep- 4).field Dezalay involves and examining Garth’s “issuessociological of power approach and hierarchy” to the le- tionlaw firms and appreciationpartnering with that the their LSS) picture apply the of thestructures world that are inbuilt in the field (Dezalay and Garth 2016, p. is evident. These agents, through their distribution of social change and stability (Dezalay and Garth 2016, p. capital, can determine a hierarchy of legal practice that gal field defines the field as a conceptual tool to explain is favourable to them because of their de facto monopoly (Bourdieu 1989) of career information in the law school is4). characterised Bourdieu’s “sociology as semi-autonomous of fields” has spaces been comprisingadopted to environment. “rulesresearch of differencesthe game and in the actors hierarchy oriented of the to legal the stakesfield that of 2.3. HABITUS actors in this social space, positioned according to their economic,particular socialfields.” and (Dezalay cultural and characteristics Garth 2016, (Anheier pp. 2, 3). et The al. 1995, p. 860), are assumed to compete for social positions phrases to impose symbolic dominance as critical in rein- giving rise to social structure. The actors are also placed Our research identified the deployment of linguistic law school space. Bourdieu theorised a “whole dimension the power and capital the actors represent (Dezalay and offorcing authorized perceptions language” favoured including by top its syntax,tier firms exists within “purely the Garthto shape 2016, or reformp. 4). In institutions our research, in the the law field school in relation environ to- to underline the authority of its author and the trust he ment represents a social topology where we analyse the demands” (Bourdieu 1991, p. 76). Dominant actors such relative positions and objective relations between these - actors (Bourdieu 1989, p. 16), invariably comprising law ogy from their deck of cards to maintain their ex cathe- students, the law students’ society (LSS), and faculty staff. draas top advantage tier firms in are the experienced social space. in We deploying submit thatterminol these repetitive terms become the dialect of law students, even capital (for example money, human resources, status), are Other actors, such as law firms who possess the necessary if some disavow the thought of a career with these firms. Bourdieu draws an analogy to “aces in a game of cards, 2.4. THE BOURDIEUSIAN INFLUENCE IN NORTH AMERICA ininvited the competition to exert their for influencethe appropriation within the of scarcecultural goods” field. AND THE UK in the social space (Bourdieu 1989, p. 17). Our research indicated a perpetuation (and even infatu-

2.2. CAPITAL and status, even amongst some students who refuse to Within this social space the fundamental powers are playation) the of thetop-tier notion game of top-tier in the firms social holding space. high Bourdieu’s prestige economic capital (in its different forms), cultural capital, concept of “habitus” may go some way to make sense of and social capital (Bourdieu 1989, p. 17). These species of the acceptance of this view in the social space. Habitus capital assume the form of symbolic capital when they are refers to the part-conscious “taking in” of rules, values “perceived and recognized as legitimate” (Bourdieu 1989, and dispositions, which adapts and installs itself within p. 17), that is legitimizing social class and social impor- the social space through “regulated improvisations” tance (Webb et al. 2002, p. 152). This becomes relevant (Bourdieu cited in Webb et al. 2002, p. 36). To Bourdieu, habitus is a dual system: one being a scheme of production as the culmination of a very competitive card game in the of practices, and one being a system of perception and lawin the school way prizedspace. jobOur offers, research from points top-tier to students firms, are strug seen- appreciation of practices (Bourdieu 1989, p. 19). Habitus therefore implies a “sense of one’s place” and a “sense of according to Bourdieu, symbolic relations of power tend the place of others” (Bourdieu 1989, p. 19). Bourdieu’s togling reproduce. to achieve The the offer field’s of a perception top-tier position of success as a symbol which, habitus – “the feel for the game” (Bourdieu 2004, p. 597) of high achievement and prestige reinforces the power – is a pattern of dispositions, practices and inculcated and hierarchal relations constituting the structure of the values recognised by individuals as a requirement to play (Pinnington and Gray 2007, p. 154; Cook et al. 2012, p.

118 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

1748). For Bourdieu, habitus can only function effectively and waiting staff. The team also held semi-structured if students lack awareness about what Bourdieu calls interviews at other times with key informants such as “the forgetting of history which history itself produces.” members of the Law Students’ Association and some law (Bourdieu 1977, p. 78). professors. Habitus was explored to explain how McGill’s law students’ identities and educational aspirations were impact of the information gap, through a Bourdieusian transformed even if these students insisted at the time lens,Our on research law students fills ain gap Australia. in the However,literature relevant about the to that nothing was going on. Through the Coffee House, our research to varying degrees is literature from legal student identity was developed through performances education scholars in North America and the UK that has and the accumulation of symbolic capital until students felt their future career path was their destiny rather than a matter of choice (Manderson and Turner 2007, p. applied Bourdieu’s tools to stthe activity and influence of 649). In contrast, some of the students in our research thelaw globalfirms on resurgence law students. of Bourdieu’s Their research theories was conductedoutside of Francein the first (Truong decade and of Weill the 21 2012). century, Their possibly research reflecting covered presence in the law school space through sponsorship - andrecounted recruitment that whilst activities, major it lawwas firmsnot laden had awith dominant lavish tices, student identity, perceptions of lawyers, and career entertainment that occurred with regularity at McGill aaspirations range of aspects – all pertinent relating to thelaw researchfirm recruitment in this article. prac University’s Law Faculty. Another point of dissimilarity Manderson and Turner’s ethnographic study at Canada’s with Manderson and Turner’s study is that most of the law McGill University explores, through a Bourdieusian students in our research did not consider a career with a framework, the conscious and subconscious impact of some of the students’ perspectives about the pinnacle in the top 50 of the QS World University Rankings (all but ofmajor career law success firm as theirwas shapeddestiny. byHowever, the notions in our of research career onemajor of lawthe lawfirms students on law instudents our research in a university were enrolled ranked at ANU, also in the top 50 of these rankings).15 Bourdieu’s environment. We concur with Manderson and Turner’s “thinking tools” (Wacquant 1989, p. 50) such as habitus success promoted by major law firms in their law school were utilised, in part, to research the transformation of unconscious transformation through the repetitious in- Canadian law students’ identities in a social space spon- finding that even if law students underwent a personal still existed opportunities for changing the meaning of the p. 649). This space was the “Coffee House”, a weekly social socialfluence space of major and the law performances firms in the law(Manderson school space, and Turner there sored by major law firms (Manderson and Turner 2007, 2007, p. 649). We will later on in this article detail how during the recruitment period that consumes many event at McGill University’s Faculty of Law, specifically: of the broader law student community. - the law school space can be transformed for the benefit In the UK, Sommerlad took a longitudinal and socio- aggressivelyfinal-year law compete students to throughout "sponsor" CoffeeJanuary House. and Febru In ex- logical approach by researching: ary, major law firms from Montreal, and even Toronto, professional identity formation, at both the individual waiters and live music, free food and – rather more to and organizational levels, and the dialectic between indi- change for a prominent presence, the law firms supply the point – free drink, at a cost of $5,000 to $10,000 per vidual processes and the social trajectory of organizational week (Manderson and Turner, p. 650). reproduction (Sommerlad 2007, p. 190). Manderson and Turner’s categories of participants was Through Bourdieu’s social analysis, Sommerlad surveyed wider than our research cohort, with the Canadian team legal practice course (LPC) students at an unnamed “new” conducting over 80 informal, unstructured interviews university16 in a large UK provincial city. This is similar to with attendees at 10 Coffee Houses held during 2005, our research whose participants included PLT students - (equivalent to LPC students), although Sommerlad’s col- tendees comprised law students, lawyers, academic staff, lection of data, which commenced in 2004, is more layered, six of which were sponsored by major law firms. The at

15 QS Top Universities Ranking https://www.topuniversities.com/university-rankings 16 A university established in post-1992.

Law in Context, Vol 37, Issue 1, 2020 119 ISSN: 1839-4183 as she surveyed two cohorts of part-time and full-time these class differences with reference to the capital that LPC students.17 Sommerlad’s objective was to track the law students and lawyers possess (for example, social, development of career aspirations, perceptions of the cultural, economic and symbolic capital) (Jewel 2008, p. - 1156), noting: in Bourdieusian terms, the myth of merit creates a researchlegal professional in contrast field, to our and study levels is of the attainment social class (Som and habitus that causes law students to internally arrive at merlad 2007, p, 196). A significant aspect to Sommerlad’s individual expectations and goals based on the legal pro- gender, working class, Asian British, Black and Ethnic Mi- nority)ethnic profiles to inform compiled her research from thethrough students the Bourdieusian (for example from top schools are more likely to perform high-prestige framework. Our research did not formally seek this level workfession’s that existing focuses hierarchy…Within on securities and general large firms, corporate graduates law, of detailed background information from the law students whereas graduates from lesser schools will specialize in and lawyers volunteering in the study (apart from their lower-status work such as personal injury defence, family gender, legal work experience history, and in the case of PLT law, or general practice (Jewel 2008, pp. 1185, 1194). students and lawyers the university where they obtained their LLB or JD). Whilst the law students (particularly We concur with Jewel’s proposition that a perceived lack of credentials thwarts many law graduates of some by their gender, issues of social class or ethnicity did not law schools from starting their career in commercial or arisefemales) in the in focusour research groups or provided semi-structured responses interviews. influenced18 corporate law practice.20 However, we differ with Jewel’s Whilst Sommerlad’s in-depth research did not reveal the assertion that “a career in public service” is equally pres- LPC students’ exposure to the recruitment activities of tigious as a corporate law practice due to the credentialed prestige of US government sector positions (Jewel 2008, p. 1205). Our research highlights that amongst many law 2007,law firms p. 205). in the The law encounters school field, the it students did highlight in our theresearch close links some firms had with select universities (Sommerlad to be of higher prestige than most government practices. Wherestudents Jewel’s in Australia, research careers is very in top-tier pertinent firms to are our considered research perceptionshad with the ofrecruitment legal practice activities is relevant of top-tier to Sommerlad’s firms in the is her observation of the devaluation of the career plans of insightslaw school into space Bourdieu’s and the cultural firms’ influencecapital.19 Sommerlad’son students’ law students “shut out” (for lack of suitable capital) from level, remained largely closed to “outsider lawyers” (for majority of law teachers in the USA whose background in example,finding that Black the and legal Ethnic profession, Minority especially graduates at and the those elite elite“high-status” educational jobs. institutionsJewel identifies created the complicity a gulf between of the vast the with working-class backgrounds) resonates with our experiences of law-status students and the experiences research. In particular, the information gap has a power- and expectations of their teachers (Jewel 2008, p. 1206). Jewel’s recommendation to recognise a broader range of about the hierarchy of Australian legal practice, in effect career trajectories as of equal value with “high-status” ful influence on shaping the perceptions of law students jobs (Jewel 2008, p. 1209) is endorsed by this article as a legal profession. rankingUtilising top tierBourdieu’s firms as theories,being superior Jewel, to in the her rest lengthy of the essay, researched the issue of class as it relates to legal consequence We contend of theour influence research large contributes corporate to thefirms body wield of education. Through the Bourdieusian lens, Jewel examined literaturein the law withschool its field. focus, through a Bourdieusian lens, on

17 In total 63 part-time students and 120 full-time students participated in Sommerlad’s study. Each student completed two questionnaires: one at the beginning of their course and one at the end of the course. Sommerlad followed up the first questionnaire with focus groups that around one-third of the students participated in. The focus groups were then followed up by semi-structured interviews with the students, largely drawn from the full-time students, with some part-time students participating. Selected members of the cohort were later interviewed at staged interviews during their training contract and into their first year two years of legal practice post-quali- fication, see Sommerlad 2007, p. 196. 18 Interestingly, some of the ANU LLB students in our research expressed concern about the impression that because the ANU’s law program has a predominant international law focus, their counterparts at commercially-oriented law schools in universities in Sydney and Melbourne will have greater opportunities for career paths in large commercial firms. 19 For example, Sommerlad noted that LPC students’ cultural capital includes their secondary school grades, their school attended, and sporting abilities, see Som- merlad 2007, p. 205. 20 Jewel’s essay consistently uses the term “corporate law” without any reference to “commercial law”, even though corporate law is a subset of commercial law.

120 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 top tier recruitment practices in the Australian law school of recruitment, we slightly relaxed the pre-requisite requirements: one of the undergraduate students was from PLT students in the Australian law school environment, the University of Canberra,22 and one of the lawyers had andfield. of We early also career present commercial the voices oflawyers the undergraduate who were only and a seven years commercial practice experience. The prin- few years earlier actors in that social space. cipal researcher moderated six focus groups (two focus groups for each of the three categories of volunteers) in the second half of 2014, with each comprising between 7. PART III: METHODOLOGY AND LIMITATIONS four and six volunteers. An assistant moderator was also present. Focus group sessions were held at the ANU “Though this be madness, yet there is method in't.” College of Law. Due to the inability of some prospective William Shakespeare, Hamlet volunteers to participate in the focus groups, we expanded the project by organizing semi-structured interviews with

3.1 RECRUITMENT the circle of volunteers with an individual perspective. We decided to recruit law students enrolled at ANU to Twenty-fourindividual volunteers, semi-structured giving us interviews more flexibility were conducted to widen minimise any proximity issues, as our school is part of the with seven undergraduate students, eight PLT students, ANU College of Law, where we teach in its PLT program. and nine early career commercial lawyers over two peri- ANU is located in Australia’s capital city, Canberra. The ods: between July and December 2014, and between July ANU College of Law has around 1,400 students, many and September 2015. All interviews were conducted in of who study for their LLB or JD degree, with the PLT person at the ANU College of Law or by telephone/Skype. course delivered online to students across Australia and Each volunteer received a $25 Co-op Bookshop voucher internationally. ANU is ranked in the top 50 globally, for their participation. and number one in Australia, in the QS World University Rankings. We decided to recruit undergraduate students who had completed two of the core compulsory subjects: 3.2. DATA COLLECTION Corporations Law and Lawyers, Justice and Ethics. This Focus group sessions ran between ninety minutes ensured students volunteering for the research were more and two hours, with interviews running between thirty informed about commercial law and ethics. Similarly, we also decided to recruit PLT students who had already pre-determined questions to stimulate discussion and completed the core compulsory subjects Commercial follow-upto forty five exploratory minutes. questionsVolunteers in thewere focus asked groups a set and of Practice and Legal Ethics in the PLT program.21 Whilst interviews. Volunteers were questioned about their experience of studying ethics at law school and the role of we decided that early career commercial lawyers eligi- legal ethics in law studies or commercial legal practice. blethere to isvolunteer no official for designation the project of would an early have career experience lawyer, We explored the volunteers’ impressions of commercial law, with the lawyer volunteers asked to compare their proximity and convenience considerations, we decided impressions as students with their current attitudes. toin commercialrecruit commercial practice lawyers of up to fromfive years. the Canberra Again, due legal to Volunteers were asked which subjects they preferred community of which approximately 2,400 lawyers practice studying in their law studies; students were asked which areas of law they were interested in practising, and lawyers legal centres, along with some as in-house counsels or were asked how their law studies prepared them to be a solein private practitioners. firms, government agencies and community commercial lawyer. After the initial reading of the focus Volunteers in the research were recruited in various group transcripts, it became apparent that wellbeing and stages from June 2014 to September 2015 via social me- work-life balance emerged as key themes requiring further dia and email. Originally, we sought a minimum of four investigation. As a result, we included questions about volunteers for each focus group, but due to the challenges wellbeing and work-life balance in interviews conducted

21 To avoid any potential conflict of interest during the principal researcher’s teaching of Commercial Practice in the PLT program, PLT student volunteers were recruited by a school administrative officer rather than the principalesearcher r directly. The officer revealed the names of the PLT volunteers only after their Commercial Practice grades were released. 22 The University of Canberra’s law school was established in 1993.

Law in Context, Vol 37, Issue 1, 2020 121 ISSN: 1839-4183 between July-September 2015. This led the principal We were able to obtain from the participant lawyers researcher to email follow-up questions to volunteers details of where they studied their LLB. This provided who participated in focus groups or interviews in 2014, diverse insights which we could compare with the law with a written response rate of about 50%.23 All focus student participants. Six of the participant lawyers were groups and interviews were audio-recorded and tran- ANU law alumni, and three were law graduates of the scribed verbatim by the principal researcher. We re-read University of Canberra. Three were alumni from law and re-analyzed the transcripts and follow-up written schools in Sydney, the most populous city in Australia: responses to identify elements to categorise into themes and connections, and develop a consensus-based coding scheme for data interpretation (Harding 2013, Litosseliti wasthe University a law alumna of Sydney of the (n University = 1), University of Tasmania of Technology (whose 2003). The principal researcher used the scissor-and- lawSydney school (n = was 1) and established Macquarie at the University end of the (n 19= 1).th century), Another sort method to code phrases, sentences, and participant one was a law graduate of Bond University (to date the exchanges relevant to the categorical themes, as part of the qualitative analysis that occurred over several months degree), and one was an alumna of the law school of New to allow immersion and familiarity with the transcripts. Zealand’sfirst and only largest private university, university the in University Australia ofto offerAuckland. a law We acknowledge the participant lawyers, all of whom 3.3. LIMITATIONS practice in Canberra are not representative of the legal profession Australia-wide. As Canberra is the seat of national The research has some limitations. The research government, it has the highest proportion of government

(Law Society of New South Wales 2018, pp. 22-23). This inwould their have family benefited to study from law more and theirbackground secondary data school of the lawyers (46%) compared with the nationwide figure (12%) backgrounds.student participants, However, such this as was whether beyond they the were scope the of first the research, although such data could have provided further Anclosely unintentional reflected the counterbalance percentage of was government that more lawyers private practitioners(n = 8, 42%) who volunteered comprised for the the lawyers semi-structured in the research. inter- students (all but one were students at ANU) in our research views than the focus groups,24 arecontext a representative to the findings. sample We also of lawdo not students claim thatin universi the law- and even slightly over-representative, of the Australia- ties around Australia. Yet, we submit that the data which wide composition: 69% of lawyers thus beingAustralia-wide more reflective, are in emerged from the sample provides insights and themes that are relevant and should resonate with law student interviews were private practitioners. Females constitute 61%private of lawyerspractice, (across 77% of all lawyers practice (n areas) = 7) volunteering nationwide who for the literature review, the SES results in some universities locatedcounterparts in two in ofother Australia’s Australian largest universities. metropolitan As flagged areas in Society of New South Wales 2018, p. 20). In comparison (Sydney and Melbourne), are relatable to our research femaleshave practised comprised for up 89% to fiveof the years lawyers after volunteeringadmission (Law for with respect to information from careers advisors. The the research. Both the lawyer focus groups were all-female literature review also highlights some contributions to a 2020 collection of scholarly literature focusing on well- for the semi-structured interviews (compared to seven ness in law (Marychurch and Sifris 2020), providing a females).(n = 4, n There= 6), and was onlyno apparent two male explanation lawyers volunteered for the large more contemporary relationship to our research data disparity between female and male lawyer volunteers, especially in relation to students being informed of career but this may explain the dominant themes of work-life paths beyond law school. balance arising from the lawyer focus groups and interviews.25 Amongst the undergraduate and PLT students

23 These follow-up questions sought volunteers’ responses to wellbeing, work-life balance and further insights about impressions of commercial law and the role of ethics. 24 One of the lawyer focus groups comprised two government lawyers and two private lawyers, the other comprised four government lawyers and two private lawyers. Two government lawyers and seven private lawyers volunteered for interviews. 25 At the end of an all-female undergraduate student focus group, one of the volunteers offered an explanation to the moderator on the paucity of male student volunteers, “You appeal to our emotions when you send the email, like could you help us”, eliciting laughter from the fellow volunteers. 26 The growth was measured from 2011 to 2016, see Law Society of New South Wales 2016, p. 4.

122 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

4.1 DOMINANCE AND DISPOSITION: THE LAW SCHOOL SPACE AND STUDENT HEADSPACE whichvolunteering, showed females females (ncomprising = 20) outnumbered of around two-thirds males (n A strong consensus emerged amongst the volunteer of= 13).law graduatesThis is consistent (Hart 2017). with26 2017 Again, national the prevalence estimates of work-life balance issues arising amongst the students early career commercial lawyers about their reflections of volunteers (one of the undergraduate law focus groups ofthe competing dominance positions of top-tier (Dezalay firms in theirand Madsenlaw school 2012, space, p. was probably reflective of the higher number of female 441).representing In one of the the Bourdieusian focus groups, cultural three of field the lawyers,composed in Another limitation was the potential of selection bias response to the moderator’s question about their expo- arisingwas all femalefrom the (n challenges= 5)). of recruiting the minimum sure to law careers marketing in their law student days, number of volunteers for focus groups and interviews. We deemed necessary a minimum of four volunteers per focus openedall reflected the narrativeon their concern by recalling: about the pervasiveness of top-tier firms. L2, a lawyer in mid-tier private practice, lawgroup, students and a minimum and early of careerfive volunteers commercial in each lawyers participant was We had a lot at the law student’s society, they’d often acute,category leading of interviews. to the principal Difficulty researcher recruiting reaching undergraduate out to colleagues in the ANU College of Law and the Canberra have seminar series where the firms would come and legal community to extend invitations to students and speak…most of them were top tier, mid-tier firms…they lawyers in their networks. The risk related to a skewing of rarelyL3, another get a non-commercial lawyer in mid-tier firm. private practice, agreed the data. However, the challenges of recruiting, especially with L2. This led the moderator to ask L3 about her ex- for focus groups, is cross-disciplinary (MacDougall and posure to marketing in law school: Fudge 2001, p. 120; Dyas 2009, pp. 65-66). We just had essentially nights dedicated to the clerkship programs and so it was just commercial, just top 8. PART IV: DATA ANALYSIS tier…there was nothing below that…you might be told “There are lots of options out there” but there was never “All the world’s a stage, and all the men and women merely players. They have their exits and their entrances.” any nights spinning the benefits of working in a mid-tier or a boutique firm. William Shakespeare, As You Like It experience in law school. L3 continued, “We never had nightsL4, thata government were about lawyer,working confirmed for government this wasor NGOs her As legal educators who teach PLT students as they ap- or anything like that.” L2 added the coda, “or even work- proach admission to practice, this study of the Australian ing with a law degree as a non-lawyer”, driving home the law school environment is timely. We seek to understand how the information gap exists in that space, and its im- their law school space. pact on students, some of whom are still discerning their pointIn aboutanother the focus dominance group, of L8, large a government commercial practice firms in career path. Our analysis seeks meaning and context in lawyer, noted that her impressions of the law and lawyers: the volunteers’ responses by applying Bourdieu’s tools, appreciating that whilst a volunteer’s responses could be snapshots not necessarily representing their whole pitched to you about how great their careers are and experience, they provide the basis to research the infor- started early in university where major law firms mation gap. All volunteers referred to in this article have you earn big money, and also only in commercial law. they’re all commercial law firms pretty much…[where] L8 continued by highlighting the lack of broader opportunities: been allocated identifiers: the undergraduate law students- are identified “U1”, “U2”, and so on, the PLT students are identified “P1”, “P2”, and so on, and the early career com mercial lawyers are identified “L1”, “L2”, and so on. to universityYou never and had tell any you family how law great firms commercial come out law to pitch was. the universities…it was five or six of them just got back

Law in Context, Vol 37, Issue 1, 2020 123 ISSN: 1839-4183

When the moderator asked the lawyers if they could gap to the detriment of students who desire information about a broader range of career options. Big impressions law school, several said they could not remember such have the power to foment narrow perceptions of real law. occurrences.recall criminal, family or wills firms attending at their The experiences of the student participants went some

27 L8 lawyers experienced in their student days. U1 in his focus laterL8’s in thereference focus group to “five used or six”the termof them Bix wasSix, asa reference did other way to confirming the information gap the participant volunteers.to the “Big Six” U13 – usedthe six this top-tier phrase firms in her in Australia.interview. U9 used a similar variations in her focus group (“the Top 6 in focusgroup group observed commented that “mostly that the the bigger careers firms…with fairs comprised some medium firms” were present at careers fairs. P4 in her clerkship applications to “all the top six” (he also recalled theSydney”). careers L12, fair a hall lawyer “packed in a global out” with firm, “commercial reflected on law his “all theit’s big a shame firms.” that U5 there’s in her not focus much group more representationlamented that: - provide their impressions of “commercial law” and “com- of the lot of mid-tier firms in Canberra…they’re consid mercialand the lawyers”,big firms”). some When of the the descriptors volunteers epitomised were asked the to exposed to them through my work. But it would be nice toered see very them good, on goodcampus law as firms well in because Canberra the and majority I’ve been of students, that’s where we’ll be going. U1, in his follow-up written response, thought of “big power and influence of top-tier firms. U5’s perceptions about the aspirations of her fellow students were heightened by her experience as a former business”,firms…trying and their P14 inbest her to interview earn more thought money”, “big P1 compa in her- member of the LSS executive. U5 understood, through nies…lotsfocus group of money.”had a similar U15, who response had experience with “big working firms, big in her LSS role, how sponsorship buys visibility (Manderson and Turner 2007, p. 656). U5 pointedly observed that on the LSS executive “there’s a certain type of person” aim- mergers.”a mid-tier In firm, a similar responded vein, P7 in (who his interview, worked in “big a global top tierlaw ing for a top-tier job “so they assume everyone else is as law firms, big deals, the multinational deals, big corporate- well.” This might explain why the LSS did not encourage pany disputes, big agreements”. Volunteers’ perceptions werefirm) voidthought of any “companies, references companyto small or transactions, even medium-tier com commended U5 on her point about the responsibility the smaller firms to come and recruit students at ANU. U7 impressions were not necessarily grounded in reality. space, but suggested that despite “everyone seeming to However,firms or small the powerfulbusinesses. repetition Some volunteers of these admitted terms goes these to LSS has on firms’ presence in the competitive law school that “broader understanding that you know there may law school space, and in shaping the disposition in the be similaraiming opportunitiesfor these top-tier elsewhere.” firms in U9Sydney”, sympathetically there isn’t studenthighlighting body the psyche. impact Students’ of top-tier perceptions firms dominating about com the- responded by indicating it was an issue she tried to raise mercial law and the ideal commercial lawyer had a “big” when she herself was an LSS executive member. U9 com- and theirpowerful presence thread: in thebig lawfirms, school big business,space – the big cultural deals, “clerkshipmented that spots” the LSSat the sells LSS premium “career event sponsorship of the year”. for Thefive big agreements, big money. Top-tier firms are so big, highfigures cost in of return premium for sponsorshipsnaming rights rules and outthe most right smaller to sell dominance and hegemony. It almost becomes inevitable field – is so visible that they have the power to exercise practices, and virtually all family, criminal and general herfirms perceptions from being of able the to legal pay profession. to play the U9game. recalled Such thatwas that many medium-tier commercial firms, government the influence of sponsorship on U9 that it impacted on or lexicon of students in the same way the Big Six do. We could name were the ones that “spent money through the arguepractice that firms this do vacuum not generally contributes enter to the the consciousness information LSS”in her or first saw two “their years name of law up, school all their the signs only around”. law firms As she an

27 Collectively, Australian top tier firms are informally the “Big Six” (equivalent to the UK’s “Magic Circle”). Whilst legal profession opinion varies about the number of top tier firms particularly with global mergers or associations, the general consensus is there are six based on their number of employees and offices, international links, and clients’ size and profile, see Pinnington and Gray 2007, p. 154.

124 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

LSS executive member, U9 attended a clerkship evening jobs that I can get.” U9 empathically proposed that there in her second year where she was: is scope for the ANU careers centre to be more proactive, like being a body in the crowd…I heard from all the rather than leaving it up to the LSS who are student volunteers, but unfortunately for many students there “is a big information gap”. Ifirms, learned you of know opportunities I was like outsidewell all rightthat. so those are the firms I can work for. It was only once I got further along The observations of U9 warrant further research into 4.2 CULTURAL CAPITAL AND HABITUS: BODIES IN THE the activities of law student societies, particularly their CROWD - bution of career information to students. school space provides much context for information gap, sponsorship by law firms and the influence on the distri Whilst the dominance of top-tier firms in the law and Turner 2007, p. 651), and our research highlights dominance as a standalone concept. Similarly, we deploy Power is never absent from a social field (Manderson Bourdieu’swe seek to avoid habitus utilising – “the Bourdieu’s feel for the tool game” of cultural (Bourdieu field power, economic capital (sponsorship funding) and so- cialthat capital in the ANU (their law collaboration school space, with top-tier the LSS) firms to haveexercise the Madsen 2012, p. 442). Bourdieu’s habitus is a scheme of a gravitational pull on student crowds. The psychology production1991, p. 13) of – inpractices, combination and also with a thesystem field of (Dezalay perception and of crowds or mobs is beyond the scope of this article, and appreciation of these practices (Bourdieu 1991). We but the information gap is perpetuated by the truism that crowds begat crowds, where the fear of missing out observations of students adopting the habitus generated permeates as an emotional contagion. This ideal of legal benefited from some volunteers who recounted their wellbeing on students who lacked the capital to obtain LSS leaders who transpose their own top-tier ambitions aby coveted top tier clerkship dominance – the in theace field, in Bourdieu’s and its impact competitive on the bycareer pursuing success lucrative is reinforced sponsorships (or even influenced) that are out by of some the game of cards. Ingrained in law students seeking top tier jobs is the departments. Apparent from the research is that many understanding that the application process is, as U1 de- studentsreach of all are but drawn well-moneyed to the power firms andand some prestige government exuded scribed in his focus group, “fairly competitive”. L3 was

marketing efforts were like “shock and awe campaigns”, thatby top-tier is for another firms in article). their marketing Subconsciously, events these (pop students culture wheremore straightforward students were by “just recalling spammed” that the withtop tier pitches firm’s depictions of smart-suited lawyers is also influential, but 2007, p. 663) perpetuates “big” (yet paradoxically, at the samefind that time their narrow “taste and of realpossibly law” unrealistic)(Manderson impressions and Turner withfrom dismaytop tier that firms a lot proclaiming of her fellow their students benefits, “were further just of commercial law and commercial lawyers. straightentrenching commercial, the competitive like those atmosphere. were the only L2 reflectedjobs that This leaves little room in the law school space and in they ever applied for, they wouldn’t consider anything else.” L2 thought this limited their career options, with power and capital to be effective actors in the law school their perception that working in a top-tier or mid-tier environment.the consciousness The ofcrowd students can pressurefor law firms some lacking students, the who may have successful careers as commercial lawyers much laughter from her fellow volunteers when she made afirm revealing was the observation “pinnacle at that the theend sort of their of people study.” who U5 drew have practitioners, to question and re-align their values in pur- volunteered for the focus group are “probably a different suitin smaller of the capital firms or that government, being a top-tier or as lawyer family purportedly or criminal type of law student to the ones who…are probably doing a clutch of [clerkship] applications right now because also has the insidious effect of crystalizing the information they’re due soon.” A fellow volunteer intoned “hush” to gapbestows. for fellow Much students of the crowd’s who have embrace no interest of top-tier in a career firms further laughter from the group. This candid exchange with the Big Six. U8, who was in the focus group with U5, pointed to the open secret that applying for clerkships is U7 and U9, expressed concern that her admitted lack of so competitive that aspiring law students will put virtually motivation to attend LSS events, careers fairs or clerkship everything aside to prepare their application. L3 recalled evenings could limit her ability to discover “the kind of knowing students “who essentially stopped going to uni,

Law in Context, Vol 37, Issue 1, 2020 125 ISSN: 1839-4183 and stopped studying for a month, and instead dedicated Another way the information gap is further perpetu- that month to clerkship [applications].” L2 followed with her observations, “It was clerkship month and then that’s what they did instead of going to uni.” timeated inshe favour was awareof top-tier of other firms options is the absence was at ofthe a habituscareers Whilst none of the volunteers indicated that any top fairgenerated where by she other encountered firms. L3 recalled a booth that for the government only other departments. L3 bemoaned the lack of engagement for a month to submit their clerkship applications, it is no with the government representatives (“asked a couple surprisetier firms to encouraged hear some students towent suspend to these their lengths studies to of questions and take a brochure”) compared with that demonstrate they have the capital (willingness to work representatives would: tier lawyer. One consistent thread across all the volunteer of the essentiallytop tier firms. deliver L3 you recounted the rote learned that the like government answers, categorieslong hours was by putting the impression the firm about first) theto qualify very long as ahours top- “It’s a great place to work, you know”, “We do this kind of expected of commercial lawyers: stuff” but you don’t ever sort of get a feel for it. Given the large number of law students seeking future “12, 13, 14 hour days, weekends” (U9 in focus group) employment, it is possible government departments may “15 hours in a day” (L7 in focus group) “15 hours day” (U14 in interview) Perhapsattract enough government sufficiently departments qualified applicants(and medium without and “16 hours a day, 16 hour days” (P6 and P8 respectively having to outlay the capital to match the top-tier firms. in the same focus group) theirsmaller applicants. private firms) However, know L3’s they experience do not need highlights to play the by “work you around the clock” (U10 in interview) challengesthe same recruitment of chipping away rules at that the top ingrained tier firms habitus impose (Jewel on “90 hour weeks at the big firms” (U16 in interview) 2008, p. 1223) amongst law students who are drawn to “working very long hours” (U14 in interview) options. The information gap is not just about the lack of weekends” (P13 in interview) informationtop-tier firms in without the law schoolthe knowledge space but of also broader concerns career the “start early in the morning, finish late at night, go in on “work up to 10/11pm” (L10 in follow-up written response) lack of awareness concerning diversity of career paths. It also pertains to the wide gulf created by the high octane “there until 11, 11.30 at night and coming in very early the next day because of the expectations” (U13 in interview) with the less vigorous approach by the rest of the legal professionenthusiasm that displayed contributes by the to thetop-tier domination firms comparedof top-tier

These impressions were confirmed by two of the firms in the activities and minds of many students. lawyers“working (who past worked 10 o’clock, in intop-tier some situations firms): I was working 4.3 FAR FROM THE MADDING CROWD: COMPLICITY AND past kind of 4 o’clock in the morning” (L17 in interview) RESISTANCE “we worked 40 hours straight” (L12 in interview) Dezalay and Madsen observed that in the social space the structural centre of gravity is where the effects of the Long hours are more power could be inferred from these impressions and experiences. As discussed earlier diminish at the outskirts (Dezalay and Madsen 2012, p. in this article, perceptions of commercial law itself pre- 441).field are Yet more it is these strongly outskirts felt, while where the many effects students eventually are positioned, and where the information gap is wide because may be ingrained in the habitus of some students who of the lack of broader career opportunities presented to decidedominantly a month relate off to from big theirfirms. law The studies mantra is of a smalllong hours price students on the edges of this space. It appears a large number of students in the research were on these outskirts (by choice or otherwise), as they did not indicate they tierto pay lawyer to demonstrate by devoting their 16 abilityhour days to put on the their top-tier clerkship firm actively participated in their law school’s careers fairs that application,first. If it means then adoptingso be it. the habitus of a real-life top-

were dominated by top-tier firms. For example, in one of

126 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 the PLT student focus groups, the volunteers were tepid about embracing the careers fairs due to lack of interest or know and you say, well, you know, honestly and truly. It - comesyou find down out theirto status favourite and ego. subject is criminal law, you tion about the type of volunteers who were drawn to this Money and prestige could also have been underly- researchthe predominance possibly explainsof top tier their firms. disposition U5’s earlier to observabroader ing motivators for L12, U11 and U13 in their quest for a career perspectives or their askance at, or antagonism corporate career. However, U13’s insightful commentary speaks to the larger question of which actors in the cultural that it was “a bit offensive” how “impressionable” the towards, top tier firms. L2 mentioned in her focus group- also complicit in the emergence and strengthening of the informationfield (in addition gap. Theto the criminal top-tier law firms enthusiasts and the thatLSS) U13 are offirms promotional thought students items at were, careers and days in the with other their lawyer message fo pointedly refers to have every right to adopt the habitus “ohcus group,gosh, greatL5 cynically to work associated there.” However the firms’ L12, distribution a global the “more cool” and “impressive” things like the notepads by gravitating to top-tier firms. Perhaps students who firm lawyer, recounted his excitement at gravitating to becomeare initially corporate averse tolawyers. commercial Their lawmotivation may find to their practice true money on this branding and advertising, and he wanted calling and practice commercial law, or more specifically “toand be the a partUSB ofsticks. that.” L12 noted the top tier firms had the Some readers may interpret L12’s pull to branded with large commercial firms may be largely driven by the- inglyhabitus or generatednot) in the by development top-tier firms. of However,the information we submit gap atthat the students expense identified of others by who U13 arewant complicit to learn actors more (know about he,stationery along with and twotrinkets students as being who the aspired superficial to practice product in criminal law or family law careers. This complicity has the of the promotional abilities of top-tier firms. However, about commercial practice that transcended notions of recruiters. This might rankle students with a true passion wealthtop-tier and firms, dominance displayed exuded a sense at ofthe realism centre andof gravity passion in foreffect a top-tier of artificially commercial enlarging career, the crowdbut it adds around status, top-tier ego the social space. L12’s advice to law students thinking of - dentsand prestige adopt the to dispositionstop-tier firms. to Top-tiershape the firms habitus are ofmost the I wouldn’t recommend doing this sort of work for the working in a large commercial firm was: lawinfluential school space.in the lawIt can school also have environment the underhanded if enough effect stu lifestyle or the money. You do have to really love it. You of giving the LSS a crude and uninformed impression (by have to be passionate about contracts. glancing at the masses of bodies in the crowd circling the U11 indicated his focus at university had been “very top-tier recruiters or relying on potentially biased feedback much commercial courses” and he “loved” equity and from top tier recruiters) that more students are drawn to, trusts and restitution. U13 expressed frustration about and excited by, top-tier marketing activities. This could some students who attempted to leverage Bourdieu’s provide the LSS with less incentive to devote adequate habitus and play the top-tier game which they were ap- resources to informing students about the broader range of career options, further solidifying the information gap and perceptions of real law. The thing I cannot understand right, and it frustrates parently ill-fitted for: Despite the mass of excitement generated by the domi- me beyond belief is the people who go to law school and the volunteers displayed an unequivocal and a principled and you ask them why, and they say, oh because they’ll resistancenation of top-tier to the habitus.firms, in P1the stated law school in his space, focus onegroup of earnthey say,lots “I of want money. to work And forI say a top why tier don’t commercial you just firm”do a commerce degree, out in three years you know, and you’ll proud not to be “a victim of the siren’s call” to come climb onthat top he of wasn’t other influenced graduates andby the then top the tier corporate firms, and ladder was and accounting, in fact it’s a skill shortage area you know. Andearn theyheaps still more, go, oh,and but there’s I can’t no do graduate maths you glut know, in finance and I say well you know, at the end of the day you’re going to to “partnership of the bigger law firm.” P1 was already have to do maths if you want to work in commercial law confident of his career path as a barrister. P4, just before

anyway, you know what’s the end game here? And then

Law in Context, Vol 37, Issue 1, 2020 127 ISSN: 1839-4183 expressing concern about her lack of awareness of the submit that the information gap pushes a dominant and clerkship process, concurred with P1’s resistance. aspirational message, permeating through the law school

a student elite status. Select students, who demonstrate 9. PART V: IMPACT OF THE INFORMATION GAP aspace, feel forthat the being game, selected are deemed by a top-tier to be atfirm the bestows top of the on game in the competitive law school space. Anything less “Something is rotten in the state of Denmark.” means the metaphorical drop downstairs that L1 and L4 William Shakespeare, Hamlet are examples of the lexicon entrenching symbolic power offelt the was dominant inflicted in on the them. law school“Top tier”, space, “next and drop” forming – these part 5.1 SYMBOLIC VIOLENCE: TAKING THE NEXT DROP of the language of the habitus. A revealing and poignant dialogue between L1 and L4 (both lawyers in the same government agency) in their focus group drew attention to the lingering effect of the 5.2 LANGUAGE AND SYMBOLIC POWER: UPSTAIRS, law school information gap: DOWNSTAIRS

L4: “There’s only so many roles that top tier – ” information gap in the law school space because students L1: “Top tier, exactly.” areA provided top-tier withlaw firma copious can confidently amount of informationassert there about is no top-tier careers. From the top-tier perspective, this asser- L4: “And you don’t want to feel like, if you don’t get that – ” tion is accurate. Through the Bourdieusian framework, L1: “Yes.’ L4: “That you somehow haven’t – ” (Dezalaywhat the andtop-tier Madsen firms 2012, say and p. 448). how theyIt also say highlights it, reveals a L1: “Had to take the next drop.” symbolicto an extent struggle their toposition inculcate within a particular and outside world the view field on L4: “Somehow it’s a compromised position or, or you’re others (Webb et al. 2002, pp. 95-98). Bourdieu wrote that not, that you failed. ” language itself is an expression of the state of relations of Moderator: “Failed because you couldn’t get a job in a symbolic power (Bourdieu 1989, p. 20). Language is not in itself powerful, but it assumes symbolic power when it is “used in particular ways, or by particular groups and L4: “Yes.” top tier firm?” institutions” (Webb et al. 2002, pp. 95-98). We posit that This very candid exchange illuminated the strug- the language used by students and lawyers to describe gles and insecurities arising from a perception that non top-tier commercial law careers are a counterfeit when compared to the careers of their top-tier counterparts furtherlarge law buttresses firms reproduces the information symbolic gap power at the of expensetop-tier and maybe some in the legal community at large. L1 offirms students and marginalises who despair mostabout of the the void other of informationfirms. This and L4 did not suggest aspiring to a top-tier career, but concerning other career paths. The use of language is so they were resigned to a belief that a non-top-tier career ingrained that even students and lawyers who oppose represented a failure to become “real lawyers”. Through top-tier domination in the law school space are unaware the Bourdieusian lens, their submission to the top-tier their casual use of these terms reinforces the domination at their expense. Bourdieu’s “symbolic violence” of the dominant (Bourdieu Acclaimed American linguist Samuel Hayakawa wrote: andnotion Wacquant of success 1992, confirmed p. 167). Dominance their complicity becomes as victims a “means of of reproducing an exclusionary identity” (Francis and A word generally has meaning in relation to its op- Sommerlad 2009, p. 79). This results in L1 and L4 perceiv- posite; you need to have “rich people” in order for “poor ing their oppression (for example, inferior feelings about people” to mean anything (Hayakawa and Hayakawa their law jobs) as the “natural order of things” (Webb et 1992, p. 123). al 2002, p. 25). It is unclear if L1 and L4 became subject For Hayakawa, inferences are important and judgment to symbolic violence during their time in the law school is implied in language (Hakayawa and Hayawaka, 1992, space or after their graduation to enter legal practice. We

pp. 25, 29). The term “top-tier firm” was one of the most

128 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 common terms expressed by the research volunteers. “below” the top-tier. As previously mentioned, L1 and L4 Six undergraduate law students, three PLT students, and referred to their government law experience as being the “next drop” from top-tier practice. U11 reinforced the variation “top-tier commercial”. Of these, only two of the “Upstairs, Downstairs”30 perception of the legal hierarchy undergraduatesten lawyers used had the completed term “top-tier a clerkship firm” with or aits top-tier slight when he recounted being told at a clerkship interview

to go from top tier to lower-tier” but it’s harder to do the isfirm, also none interesting of the PLT to notestudents that hadclerkship clerkship was experience, formerly a reverse,that the benefitakin to theof choosing “analogy a of top-tier how you firm can’t is “it’sski uphill” easier genericand only term three under of the the lawyers superseded worked articles in top-tier system firms. of legal It (further supplementing Bourdieu’s sporting metaphors). training in Australia. The term clerkship has now been One implication of this is that top-tier graduates are eagerly appropriated exclusively to represent a position that is sought after by the rest of the legal community (maybe there is some truth to that) who desire to recruit top-tier theunique connotations to the top arising.tier firms. This The creates implication the impression of “top-tier” of a rising slightly higher in the sports table (but never enough sportis that league the rest ladder of the (Bourdieu firms are wasnot theknown best to firms write with about all tocapital reach into top-tier their levelranks, of with course). the endowed An elitist firm perspective proudly is that the only members of the top-tier club are those and “feel for the game” conjure up sporting metaphors). with the fortune of commencing their careers in top-tier sports (Clement 1995, p. 147), and his tools of “field”- ing conditions, top prestige, and top opportunities. Any their legal career on any other rung of the ladder (that is In this ladder top-tier firms have top clients, top work firms. A demeaning interpretation is that anyone starting even second-rate. To put it frankly, a more sinister impli- be pitied in their aspirational struggle to reach upstairs firm below the top are of lesser-value, less prestigious or andanywhere join the below exclusive the top ranks rung of oftop-tier the top-tier life. firms) will 28 at the top of the table. Whilst Google Scholar may not We are not suggesting that the linguistic usage of “top- cation is that firms that are not “top-tier” will never be the legal community, at one level it highlights the clear be unrealistically extreme and will most likely offend the be an accurate reflection of the hierarchical language in principlestier firms” beof freecurtailed speech. in the In fact,law school we consistently space. This use would the - dominance of top-tier firms in shaping the linguistics of munications with law students in the teaching sphere. 29 law students. A search The of caveat“top-tier to firm”these bringssearches over is that600 However,descriptor it “top-tier is important firm” for in thethis law article student and community,in our com results compared with nine for “large-tier firm”, and one- including law teachers, to recognise the underlying mean- for “big-tier firms”. ing and impact of the terminology they employ, even if a “top-tier firm” could include accounting or other pro fessional firms. Nevertheless the massive gulf between the search numbers highlights the influence and power- it is utilised to critique the domination of top-tier firms. 10. PART VI: BROADENING CAREER INFORMATION mentedof the top-tier by some firm volunteers nomenclature in their across perspectives several levels. (and The notion of top-tier firms being at the top is aug TO LAW STUDENTS believed that experiencing the world of commercial law waseven an criticisms) opportunity of top-tier to “skill-up”. firms. U5U14 in in her her focus interview group “To sleep, perchance to dream.” William Shakespeare, Hamlet andexpressed “do all theconflict big cases” on whether or whether to pursue she would a career be happier with a “big law firm” where she would be “at the top of my game” in a mid-tier firm. L3, in criticising the predominance of top-tier firms at careers fairs, noted the absence of firms 28 A search engine for scholarly literature and academic publications. 29 Google Scholar search conducted on 3 June 2020. 30 British TV drama Upstairs, Downstairs (1971-1975) portrayed a townhouse where the servants represented the “downstairs”, and the family (their employers) resid- ing “upstairs” symbolised the upper class. 31 Ooh La La, Performed by The Faces (1973).

Law in Context, Vol 37, Issue 1, 2020 129 ISSN: 1839-4183

6.1 “I WISH THAT I KNEW WHAT I KNOW NOW, WHEN I 6.2 “ALWAYS BE A FIRST-RATE VERSION OF YOURSELF, WAS YOUNGER.”31 INSTEAD OF A SECOND-RATE VERSION OF The moderator asked the lawyers in the focus groups SOMEBODY ELSE.”32 if there was anything they wish they had known before We strongly concur with the volunteers that bridging they became commercial lawyers. In one focus group, the the information gap in the law school space is essential responses of L5 and L8 emphasised the mindset when to give law students greater cognisance of career choice and to achieve their dreams. One option is to adopt a recommendation of ANU’s own LSS Law School Reform space.they graduated, L8 responded: strongly influenced by the dominance Committee that: pushedI thought by top-tier being firms a successful in their commercialerstwhile law lawyer school

moreMentors active fromrole in a therange law of school, fields outside with a particular the academy fo- ofmeant my friend’s working experiences, in one of the people big six firmsI work and with. now There’s I kind cusand onmainstream raising the corporate aspirations law of firms later should year students play a much and aof lot realise more that’s to commercial definitely not law the than case working listening in toone a lotof showcasing the opportunities available to them (ANU those places. LSS Law School Reform Committee 2010, p. 59). L5 concurred: The full impact of the information gap on the wellbeing 33 I certainly came out of law school thinking that go- of students is beyond the ambit of this article. Yet it is relevant to students who encounter anxiety, depression go for, best thing you could do. Now having been out in or stress in the law school space. U5 in her focus group practiceing into theand big I realise top firms it’s wasnot thereally best that option cracked you couldup to was critical that the law school failed to take steps later what it appears to be when you’re a student. in the law degree when it becomes necessary to: alleviate those problems of competitiveness or depres-

getting really, you know, tough for students. These reflections capped off the concerns expressed by sion…especially for example in fifth year when things are representationsome of the volunteer (or misrepresentation) students. It confirms of real an environment law can nar- While a variety of factors can affect a law student’s rowwhere career the ongoing horizons dominance quickly and of drastically top-tier firms (Pryor and 2008, their wellbeing which are unrelated to their studies or pros- p. 49). U5 in her focus group highlighted the importance pects of employment, a survey into the wellbeing and of being ambitious, but it was a shame that ‘realistically law school experience of students at the University of there are so many options for graduates not exposed to Melbourne found that: them on campus.’ L9, in the same focus group as L5 and The broad message was that students were stressed L8, was one of the strongest proponents of bridging the about their employment prospects and many would information gap by encouraging the entrance of a broader welcome more proactive counselling about opportunities “in and out of law” (Larcombe et al 2013, p. 426). slate of firms into the law school space: However, if we were to take a more cynical approach, mid-tier,I definitely let’s try would to get have some had boutiques, more variety, let’s try rather get some than criminal,just the top let’s tier also commercial try to get…government firms come to us. departments. Let’s focus on (for example, government or community legal service providers)we submit to impartthe introduction information alone to studentsof new actors about in broaderthe field L9’s advocating a wider range of actors, in the cultural career options, may lack the desired initial impact due

(or enthralled) by top-tier dominance. Bourdieusian field, is especially pronounced given she had practised as socialto the theoryingrained tends habitus to assume amongst the “habitusthe crowd will influenced adapt or a lawyer in two top-tier firms.

accommodate itself to the field”, and the “habitus cannot 32 Quote by Judy Garland. alter the field because of the external, objective status 33 We further address the link between the information gap and wellbeing, see Yau B., Catanzariti D., and Atkinson J. 2019. 34 We are prepared to speculate that if the LSS were to promote top-tier jobs as “alternative careers”, the top-tier firms may withdraw premium sponsorships.

130 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

options in the legal community and beyond (Pinnington and Gray 2007, p. 164). thethat point is attributed that by theto the time field” a student (Adkins seeks 2003, other p. 36). options The suchpower as of a top-tiercareer in firms government, manifests NGOs itself orin theother habitus sectors, to 11. PART VII: CONCLUSION these choices are considered “alternative” (Pryor 2008, p. 49). The thought of an alternative career can be a “strange idea” (Manderson and Turner 2007, p. 665). “If you build it, they will come.” As Hayakawa wrote, a word generally has meaning in Field of Dreams relation to its opposite, and in the law school space the 34 opposites to “alternative” and “strange” just happen The famous quote from the movie Field of Dreams is to be “top-tier” and “normal”. The top-tier activity in the actually, “If you build it, he will come.” However, it has been popularly re-interpreted over time to adopt an inclusive law which comes to have “normative currency” in the law cultural field presents a powerful corporate image of real school space (Turner and Manderson 2007, p. 786). Even of the career aspirations – in and out of law - of all law meaning. Ideally the law school field should be inclusive in her focus group derided the commercial section of a themselves there, just as much as every other legal service U9 – who decried the domination of top-tier firms – later students. Whilst top-tier firms have every right to establish “regarded as a bit of joke” as “they’re not making money of career information within the law school space that leading Australian government law firm because it was for people.” excludesprovider, thewe cannotdreams remain of many silent students. about theWe narrowconcur withflow Jewel’s caution to law teachers that the danger of silence by reproducing their version of a model lawyer, it is not is the strengthening of the habitus which lends itself to so simpleWhilst fortop-tier other firms actors dominate to compete the bylaw replicating school space the reproducing existing inequalities amongst students in shot-for-shot movie remakes are not as successful as 2008, p. 1207). We value the importance of respecting 35 theactions original, of top-tier firms in the cultural field. Just as most- the lawability school of students field and to the make wider informed legal profession choices (Jewelabout ate “directly and aggressively on the home turf of the their careers with a wider range of information at play dominant” risks any failing attempt to gain by alternative traction (Pinnington firms to oper and Gray 2007, p. 160). It may even appear embarrassingly aim to be dealt the top-tier ace and pursue their notion ofin realthe field.law. Equally, That is nota student’s to disrespect choice their to seek choice a career if they in attract, which could have the unintended effect of turn- government, mid-tier, smaller private practice or other awkward for these firms and the students they seek to sectors must also be respected and valued as the practice the law school environment needs to emerge with actors of real law rather than an alternative. The observation of ing away prospective students. Rather, a new field within Professor George Williams, Dean of Law at the University power (Dezalay and Madsen 2012, p 443). For example, of New South Wales align with our arguments. Professor dedicated to question and redefine social hierarchy and- Williams, who seeks to show his students the full range sion of real law. In more contemporary literature on law studentlaw teachers wellbeing, can influence it is recommended students to havethat lawa broader teachers vi “very good at dominating the landscape for law students” play a role in helping students “identify their goals in re- whichof career can paths, “create commented a narrow, linear in 2019 path that to a career”law firms (Price are lation to legal works, or at least to clarify their values in relation to their career options.” (Swannie 2020, p. 127). capital inevitably exercise the power to make their pres- ence2019). felt Our in article law schools. demonstrates With much that firmsat stake with for substantial students law school space and assert their control and autonomy across Australian law schools, every effort is required byA bridge establishing can be theirestablished perspectives for other and firms value to of enter real lawthe to address and eventually eradicate the information gap and real lawyers. This would give more students a broader and re-imagine this space to open the minds of students and realistic understanding and choice of their career pursuing their dreams to become real lawyers.

35 The shot-for-shot remakes not as successful as the originals were Psycho, The Omen, and Funny Games, see Salt 2016, p. 467.

Law in Context, Vol 37, Issue 1, 2020 131 ISSN: 1839-4183

12. REFERENCES an analytical framework.” BMC Medical Research Methodol- , 9(1): 65–74. 1. ogy Theory, Culture & Society, 20 (6): 21–41. 17. Evers, M., Olliffe, B., and Pettit, R. 2011. “Looking to the past to plan for the future: A decade of practical legal training.” 2. Anheier,Adkins, L. H., 2003. Gerhards “Reflexivity: J., and Romo Freedom F. 1995. or Habit “Forms of ofGender?” Capital 45(1): 18–44. and Social Structure in Cultural Fields: Examining Bourdieu's The Law Teacher, Social Topography.” American Journal of Sociology, 100 (4): 18. Fisher, R. 2015. “The legal community's struggle with mental 859–903. health.”, 6 October, Law Report, Radio National. https://www. abc.net.au/radionational/programs/lawreport/mental- 3. ANU LSS Law School Reform Committee 2010. “Breaking the health-in-law/6828762#transcript Accessed 3/2/19 Frozen Sea.” (2010) https://www.anulss.com/wp-content/ uploads/2016/05/Breaking-the-Frozen-Sea-Report.pdf 19. Francis. A. and Sommerlad, H. 2009. “Access to legal work Accessed 4/3/19. experience and its role in the (re)production of legal professional identity.” International Journal of the Legal Profession, 4. Bourdieu, P. 1977. Outline of a Theory of Practice. Cambridge, 16(1): 63–86. UK: Cambridge University Press. 20. Harding, J. 2013. . 5. Bourdieu, P. 1984. “The Market of Symbolic Goods.” In R. Qualitative data analysis from start to finish London, UK: SAFE Publications. Johnson (ed.) The Field of Cultural Production: Essays on Art and Literature. New York, USA: Columbia University Press. 21. Hart, C. 2017. "Law Council of Australia tackles gender bias with new workshops." Australian Broadcasting Corpora- 6. Bourdieu, P. 1989. “Social Space and Symbolic Power.” So- tion, Sydney, 17 February 2017. https://www.abc.net.au/ ciological Theory, 7 (1): 14–25. news/2017-02-17/law-council-of-australia-tackling-the- 7. Bourdieu, P. 1991. Language and Symbolic Power. Cambridge, gender-bias/8280186 Accessed 1/06/20 UK: Polity Press. 22. Hayakawa, S. I. and Hayakawa, A. R. 1990. Language in 8. Bourdieu, P. and Wacquant, L. 1992. An Invitation to Reflexive thought and action. Sociology. Cambridge, UK: Polity Press. 23. 9. Bourdieu, P. 2004. “The peasant and his body.” Boston, USA: Houghton Mifflin Harcourt. Ethnography, learning environment case study.” The Law Teacher, 49(1): 5 (4): 579–599. 92–121.Hewitt, A. 2015. “Can you learn to lawyer online? A blended 10. Cook, A., Faulconbridge, J., and Muzio, D. 2012. “London’s 24. Holland, C. and Taylor, D. 2020. "Hopeful Perspectives: legal elite: recruitment through cultural capital and the re- Incorporating Hope Theory in Australian Law Students' production of social exclusivity in City professional service Academic Experience." In J. Marychurch, J. and A. Sifris (eds.). 44 (7): 1744-1762. Environment and Planning A Wellness for Law: Making Wellness Core Business, Chatswood, 11. Clément, J. 1995. “Contributions of the Sociology of Pierre Australia: LexisNexis Butterworths, pp. 111-122. fields.” Bourdieu to the Sociology of Sport.” Sociology of Sport Jour- 25. Jewel, L. A. 2008. “Bourdieu and American legal education: nal, 12(2): 147-157. 12. Dezalay, Y. and Garth, B. 1996. Dealing in virtue: International hierarchy.” Buffalo Law Review, 56: 1155–1224. commercial arbitration and the construction of a transna- 26. Kritzer,How law H. schools 2013. “Law reproduce Schools social and the stratification Continuing Growthand class of Chicago, USA: University of Chicago Press. tional legal order. the Legal Profession.” Oñati Socio-legal Series, 3(3): 450–473. 13. Dezalay, Y. and Garth, B. 2010. Asian legal revivals: Lawyers 27. Lane, C. 2020. “It's Your Journey: Overcoming the Adversity Chicago: USA: University of Chicago in the shadow of empire. of Law School.” In J. Marychurch, J. and A. Sifris (eds.). Well- Press. ness for Law: Making Wellness Core Business, Chatswood, 14. Dezalay, Y. and Garth, B. 2016. “‘Lords of the dance’ as double Australia: LexisNexis Butterworths, pp. 43-44. Journal of 28. Larcombe, W., Tumbaga, L., Malkin, I., Nicholson, P., and Professions and Organization, 3(2): 188-206. Tokatlidis, O. 2013. “Does an Improved Experience of Law agents: elite actors in and around the legal field.” 15. Dezalay, Y. and Madsen, M. 2012. “The Force of Law and School Protect Students against Depression, Anxiety and Stress: An Empirical Study of Wellbeing and the Law School Law.” Annual Review of Law and Social Science, 8: 433–452. Experience of LLB and JD Students.” Sydney Law Review, 16. Dyas,Lawyers: J., Apekey, Pierre BourdieuT., Tilling, and T., andthe Siriwardena,Reflexive Sociology A. 2009. of 45: 407–432. “Strategies for improving patient recruitment to focus 29. Law Society of New South Wales. 2018, National Profile of Solicitors 2018 Report. 2018. https://www.lawsociety.com. groups in primary care: a case study reflective paper using

132 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

(eds.). Wellness for Law: Making Wellness Core Business, Chatswood, Australia: LexisNexis Butterworths, pp. 39-42. au/sites/default/files/2019-07/2018%20National%20 30. Litosseliti, L. 2003. Using Focus Groups in Research. London, 39. Salt, B. 2016. “The exact remake: a statistical style analysis Profile%20of%20Solicitors.pdf Accessed 1/06/2020 UK: Bloomsbury. New Review of Film and Television 31. MacDougall, C. and E. Fudge. 2001. “Planning and recruit- Studies, 14 (4): 467–486. ing the sample for focus groups and in-depth interviews.” 40. Socialof six HollywoodResearch Centre. films.” 2020. “Quality Indicators for Learn- Qualitative Health Research, 11 (1): 117–126. ing and Teaching.” Melbourne. https://www.qilt.edu.au/ 32. Manderson, D. and Turner, S. 2006. “Coffee house: Habitus qilt-surveys/student-experience Accessed 1/06/2020 and performance among law students.” Law & Social Inquiry, 41. Sommerlad, H. 2007. “Researching and Theorizing the Pro- 31 (3): 649–676. cesses of Professional Identity Formation.” Journal of Law 33. Marychurch, J. and Sifris, A. 2020. Wellness for Law: Making and Society, 34 (2): 190–217. Wellness Core Business. Chatswood, Australia: LexisNexis 42. Swannie, B. 2020. "From Alienation to Empowerment: Butterworths. How Legal Educators Can Support Student Wellbeing.” In J. 34. O’Loughlin, D., Sifris, A., and Batagol. B. 2020. “Wellness and Marychurch, J. and A. Sifris (eds.). Wellness for Law: Making Wellbeing at Monash University: The First Semester Law Wellness Core Business, Chatswood, Australia: LexisNexis Experience”. In J. Marychurch, J. and A. Sifris (eds.). Well- Butterworths, pp. 123-130/. ness for Law: Making Wellness Core Business, Chatswood, 43. Truong, N. and Weill, N. 2012. “A decade after his death, Australia: LexisNexis Butterworths, pp. 39-42. French sociologist Pierre Bourdieu stands tall” The Guard- 35. Pinnington, A. H. and Gray, J. T. 2007. “The global restructuring ian, London, 21 February. https://www.theguardian.com/ world/2012/feb/21/pierre-bourdieu-philosophy-most- International Journal of the Legal quoted Accessed 11/10/18 Profession,of legal services 14 (2): work? 147–172. A study of the internationalisation 44. Turner, S. and Manderson, D. 2007. “Socialisation in a space 36. of Australian law firms.” of law: student performativity at ‘Coffee House’ in a univer- Maybe you should consider your options.” Sydney Morning sity law faculty.” Environment and Planning D: Society and Herald,Price, J. Sydney,2019. “Banking 6 September. on a career https://www.smh.com.au/ at a top-tier law firm? Space, 25(5): 761–782. 45. you-should-consider-your-options-20190905-p52oc5.html Workshop with Pierre Bourdieu.” Sociological Theory, 7(1) Accessednational/banking-on-a-career-at-a-top-tier-law-firm-maybe- 1/06/20 SociologicalWacquant, L.Theory, 1989. 7(1): “Towards 26–63. a Reflexive Sociology: A 37. Pryor, L. 2008. The Pinstriped Prison. Sydney, Australia: 46. Webb, J., Schirato, T., and Danaher, G. 2002. Understanding Picador Australia. Bourdieu. Thousand Oaks, USA: SAGE Publications. 38. Roberts, E. 2020. “Pragmatic Concern for Law Student 47. Yau, B., Catanzariti, D., and Atkinson, J. 2019. “The Informa- Wellbeing: Financial Security, Autonomous Learning and tion Gap: A comparative study of the paradigms shaping Compassionate Teaching.” In J. Marychurch, J. and A. Sifris perceptions of career success for law undergraduates and professional legal training students in Australia and the latent implications of non-professional legal career opportunities for law graduates in England.' In R. Field and C. Strevens (eds.) Educating for Well-Being in Law, Abingdon, UK: Routledge, pp. 74–88. 48. You’ve Entered Law Land, “Legal clerkships: what are

youveenteredlawland.com/legal-clerkships-what-are-they- and-why-are-they-important/they, and why are they important?” Accessed 22 June 2/12/18. 2015. http://

Law in Context, Vol 37, Issue 1, 2020 133 Received: November 18, 2020, Date of acceptance: December 17, 2020, Date of publication: January 14, 2021, DOI: http://doi.org/10.26826/law-in-context.v37i1.125

A Law School Course in Applied Legal Analytics and AI

By Jaromir Savelka,1 Adjunct Professor of Law at the University of Pittsburgh and a Postdoctoral Fellow in the School of Computer Science, Carnegie Mellon University, Orcid: http://orcid.org/0000-0002-3674-5456 Matthias Grabmair,2 Assistant Professor in the Department of Informatics at the Technical University of Munich, Orcid: http:// orcid.org/0000-0001-6586-2486 Kevin D. Ashley,3 Professor of Law and Intelligent Systems at the University of Pittsburgh and a Fellow of the American , Orcid: https://orcid.org/0000-0002-5535-0759

Association1 Carnegie Mellon of Artificial University, Intelligence USA 2 Technical University of Munich, Germany 3 University of Pittsburgh, USA

ABSTRACT

natural language processing (NLP) enable new legal apps that, to some extent, can analyze contracts, answer legal questions,Technological or predict advances the in outcome artificial of intelligence a case or issue. (AI) While are affecting it is hard the to legalpredict profession. the extent Machine to which learning these techniques (ML) and will change law practice, two things are certain: legal professionals will need to understand the new text analysis techniques and how to use and evaluate them, and law faculties face the question of how to teach law students the required skills and knowledge to do so. At the University of Pittsburgh School of Law, the authors have co-designed a semester-long course entitled, Applied Legal Data Analytics and AI, and twice taught it to combined groups of law students and students from technical departments. The course provides a hands-on practical introduction to applying ML and NLP to extract information from legal text data. It demonstrates applications of text analytics that support the work of legal professionals, researchers, and administrators and techniques for evaluating how well the new tools work. The course culminated in joint projects engaging small teams of law and technical students in applying machine learning and data analytics

efforts to incorporate instruction on computer programming and machine learning in legal education. It then pro- videsto legal an problems. overview Thisof the article course introduces and explains the newhow textwe taught analytic law techniques students skillsand briefly of programming surveys law and schools’ experimental current

designKeywords that –prepared artificial them intelligence for the andfinal law, course legal projects. text analytics, machine learning, natural language processing, legal technology education

134 Law in Context, Vol 37, Issue 1; 2020 ISSN: 1839-4183

Acknowledgments. The authors gratefully acknowledge the University 6.3 Text Annotation Activities of Pittsburgh School of Law for offering this course and supporting 7. Students’ Comments and Instructors Lessons Learned the preparation of this article. We thank the reviewers and editors 7.1 Reactions to Course and Instructors of Law in Context for their very valuable comments and suggestions 7.2 Covid-19 Response Disclosure statement – No potential conflict of interest was reported 7.3 Reactions to Readings and Abstracts by the author. 7.4 Reactions to Programming Instruction and Assignments License – This work is under Attribution-NonCommercial-ShareAlike 7.5 Reactions to Course Projects 4.0 International (CC BY-NC-SA 4.0) https://creativecommons.org/ 7.6 Planned Revisions re Course Projects licenses/by-nc-sa/4.0/ 7.7 Planned Revisions re Teaching Programming Skills to Law Students Suggested citation: Savelka, J., Grabmair, M. and Ashley, K., “A Law 8. Conclusions School Course in Applied Legal Analytics and AI” , Law in Context, 37 (1): 134-174, DOI: http://doi.org/10.26826/law-in-context.v37i1.125 9. INTRODUCTION Summary are affecting the legal profession. Machine learning (ML) 1. Introduction andTechnological natural language advances processing in artificial (NLP) enableintelligence new legal(AI) 1.1 Changing the Focus of a Course on AI and Law apps that, to some extent, can analyze contracts, answer 1.2 New Legal Apps Change Legal Practice legal questions, or predict the outcome of a case or issue. 1.3 Article Overview While it is hard to predict the extent to which these tech- 2. AI and legal data analytics in law school curricula niques will change law practice, two things are certain: 2.1 Why teach AI in law school and how? 2.2 Teaching law students about challenges of relying on the new legal professionals will need to understand the new text technology analysis techniques and how to use and evaluate them, 2.3 Teaching law students how to evaluate the new technology and law faculties face the question of how to teach law 2.4 Preparing law students to interact with technologists students the required skills and knowledge to do so. 2.5 How much programming to teach law students? At the University of Pittsburgh School of Law, the au- 3. Course Description thors1 have co-designed a semester-long course entitled, 3.1 The Students Applied Legal Data Analytics and AI, and twice taught it 3.2 The Instructors to combined groups of law students and students from 3.3 Course Overview technical departments. The course provides a hands-on 4. Course Content by Parts practical introduction to applying ML and NLP to ex- 4.1 Part I: Introducing Python, Machine Learning, Natural Language tract information from legal text data. It demonstrates Processing and AI’s Effect on Legal Practice applications of text analytics that support the work of 4.2 Part II: Computationally Modeling Case-based Legal Argument legal professionals, researchers, and administrators, and and Improving Legal IR techniques for evaluating how well the new tools work. 4.3 Part III. Applying Text Analytics to Legal Texts This article introduces the new text analytic tech- 4.4 Part IV: AI in Legal Domains 5. Final Projects incorporate instruction on computer programming and 5.1 Suggested Project Ideas and Datasets (First Project Meeting) machineniques and learning briefly surveysin legal laweducation. schools’ Then current it describes efforts to 5.2 Project Proposals (Second Project Meeting) 5.3 Work on the Projects (Bi-weekly Stand-ups) the 2020 version of the course, including the students, 5.4 The Resulting Projects (Final Presentations and Project Reports) instructors, and course sessions in overview. We explain 6. Comparison with Spring 2019 how we taught law students skills of programming and 6.1 SCOTUS Prediction experimental design and engaged them in assignments 6.2 Fairness in Machine Learning that involve using Python programming environments

1 Matthias Grabmair and Kevin Ashley taught the course in spring 2019. The course was jointly offered at Carnegie Mellon University (CMU) and the University of Pittsburgh (Pitt). Jaromir Savelka and Ashley taught the course in spring 2020, when the course was offered solely at Pitt.

Law in Context, Vol 37, Issue 1, 2020 135 ISSN: 1839-4183 to analyze legal data. The course culminated in joint such as bail recidivism data, to support decision-making. projects engaging small teams of law and technical stu- Another source of information comes from citation dents in applying machine learning and data analytics to networks, graphs of the relations among legal cases or legal problems. The article explains how the instructors statutory provisions, that can be created automatically - based on citation information in legal texts such as cases ning early in the term with project ideas and databases or patents. (Zhang and Koppaka 2007). prepared the students for the final course projects, begin of text, forming teams, working on the projects as a team As the focus of research and development in AI and Law has changed, so should the focus of a course for law students the projects and reporting results. We draw some salient like this one. Our course differs from a more traditional and obtaining interim feedback, and finally completing comparisons between the 2019 and 2020 versions of the one on AI and Law in that it introduces students to the new course and report what worked well and what did not, techniques for processing legal texts that are now being the students’ reactions, and lessons learned for future applied in legal practice tools. Although it covers some offerings of the course. prior work on computational models of legal reasoning with rules and cases, it focuses on how text analysis tools 1.1 CHANGING THE FOCUS OF A COURSE ON AI AND can extract legal information from the texts of cases and LAW statutes, which may enable automatically acquiring the legal knowledge with which those models can reason. in which researchers build computational models of As the learning outcomes, we intend that students Artificial intelligence is a subarea of computer science understand the new AI and Law techniques for represent- and law (AI and Law), researchers have been building ing legal knowledge and argumentation, practice basic computationalintelligent behavior. models In ofthe legal-reasoning field of artificial behaviors intelligence for techniques for applying machine learning to legal data decades. While they have made great progress, for ex- including case texts, learn to develop and assess research ample, in modeling case-based arguments that account hypotheses in legal data and text analytics, and participate for analogical reasoning and underlying legal values, the in designing, planning, and evaluating a legal data analyt- work has remained largely academic. ics project. In addition, the course introduced students to With the advent of legal text analytics and citation some topical legal issues arising in the context of AI, Big network analysis, however, AI and Law techniques increas- Data, and machine learning technology, including effects ingly affect the legal profession. Legal text analytics are of electronic datamining on privacy, intellectual property computational techniques that apply natural language rights in data, and bias in ML-based decision-making. processing, machine learning, and other methods to automatically extract information (e.g., summary statistics, 1.2 NEW LEGAL APPS CHANGE LEGAL PRACTICE keywords, names of entities, citations, or topics) from The new techniques in AI and Law have enabled new text archives of legal case decisions, contracts, or statutes (Simon et al. 2018, p. 253; Ashley 2019). Machine learning practice. Based on these techniques, new legal apps can, refers to computer programs that induce or learn models totools some that extent, promise analyze (or threaten?) contracts, to answer upend traditionallegal questions, legal from data (given a set of assumptions about data repre- or predict the outcome of a case or issue. sentation and modeling), typically using statistical means, Programs like Ravn,2 Kira,3 and LawGeex4 apply text with which they can, for example, classify a document or analytics to contracts, identifying types of provisions, predict an outcome for a new case. (Kohavi and Provost semi-automating the review of contracts for routine 1998; Bishop et al. 2006). General legal analytics apply approval, highlighting issues, and referring apparently ML methods to other kinds of data in the legal domain,

2 RAVN, imanage.com/product/ravn/ Accessed 6/7/2020 3 KIRA, www.kirasystems.com/ Accessed 6/7/2020 4 LAWGEEX, https://www.lawgeex.com/resources/aivslawyer/ Accessed 6/7/2020

136 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 unusual provisions for human review. Given a virtual which the judge may be more familiar, thus improving the data room full of contracts for review, they can generate chances of a favorable ruling. (Crichton 2015) a spreadsheet cataloguing provisions by types, such as Some new legal apps are using text analytics and cita- non-disclosure agreement or disclaimer of warranties, tion networks to obtain more information from the citation links between citing and cited cases with which to infer When answering a legal question, the Ross system,5 why a case is being cited. They may identify the topic of greatly improving the efficiency of due diligence searches. based on IBM Watson (Ferrucci et al. 2010), searches the paragraph where the citation appears or that of the a large collection of texts to locate sentences or short cited case. For example, the “How Cited” tool in Google excerpts from cases and other documents that appear to Scholar Cases,8 employs such topic information to group answer a user’s question. For example, upon inputting a into equivalence classes cases that cite a particular case for question in plain English such as, “In New York, what is the same reason. CaseText’s CARA A.I. system9 is another secondary liability with respect to copyright infringement such app. When users input a written legal memo, CARA A.I. suggests additional cases to cite in support of argu- sentence from a federal district decision from the South- ments in the memo based on text analytics and citation ernand Districthow is it of established?” New York to it theresponds effect withthat its“. . top-ranked. A party is liable for contributory infringement if, ‘with knowledge of memoranda supporting particular types of motions, for of the infringing activity,’ it ‘induces, causes, or materi- example,networks. motionsThe CARA to A.I. “quash Compose a subpoena, tool generates exclude first expert drafts ally contributes to the infringing conduct of another.’. . . .”, including citation, suggested readings, and updates. discovery or disclosure.” (Hudgins 2020) testimony, file a motion for protective order or compel Lex Machina,6 now part of LexisNexis, predicts out- Sophisticated legal apps like these encourage the belief comes of new cases based on information about litiga- that legal practice is changing dramatically, threatening to tion participants and their behavior gleaned from a large shrink the profession and the prospects of law students repository of past cases (Surdeanu et al. 2011). Another and young attorneys. While this may be true to some program uses machine learning to predict the outcomes extent, the commercial press has fanned expectations of decisions of the European Court of Human Rights based for the abilities of the techniques far beyond the reality.10 on the cases’ textual descriptions of case facts (Medvedeva The legal academy should realize – and should teach 7 2020). Ravel, also acquired by LexisNexis, employs cita- law students – that, despite these advances, the text ana- tion networks to present visual maps of citations of U.S. cases regarding a legal concept input by a user. If the user prediction programs do not represent substantive features lyticof a legal applications dispute, theyare subject cannot toexplain significant their predictionslimitations. inIf citation map of cases cited by or citing the U.S. Supreme terms with which legal professionals would be familiar. inputs “campaign finance”, for instance, Ravel outputs a Court’s 2010 Citizens United decision, which permitted Legal question answerers may map answers on to ques- corporations to make independent political expenditures. tions, but at present they understand neither the ques- In addition, it offers a “Judge Analytics dashboard” that tions nor the answers. Similarly, contract review systems highlights cases a judge has cited in the past and with cannot read contracts as human attorneys would. These tools rely on matching, not reading. They cannot reason

5 M ROSS INTELLIGENCE, https://rossintelligence.com Accessed 6/7/2020. On December 11, 2020, Ross Intelligence announced it was shutting down its business due to a copyright infringement claim brought against it by Thomson Reuters. The claim alleged that Ross copied Westlaw’s copyrighted materials for use in machine learning, an interesting example of intellectual property limitations on datamining. Dipshan, R. 2020 “Ross Shuts Down Operations, Citing Financial Burden from Thomson Reuters Lawsuit.” https://www.law.com/legaltechnews/author/profile/Rhys-Dipshan/ Accessed 14/12/2020 6 LEX MACHINA, https://lexmachina.com Accessed 6/7/2020 7 RAVEL, https://ravellaw.com Accessed 29/7/2019 8 GOOGLE SCHOLAR, https://scholar.google.com Accessed 6/7/2020 9 CASETEXT, https://casetext.com Accessed 6/7/2020 10 For some examples, see (Simon et al. 2018, p. 253).

Law in Context, Vol 37, Issue 1, 2020 137 ISSN: 1839-4183 about how small changes in the question would affect experimental design and engaged them in assignments the applicability of the answer. As a result, these tools that involved using Python programming environments require supervision by knowledgeable attorneys who to analyze legal data. understand what they can and cannot do and know how to evaluate their results. current efforts to incorporate instruction on computer These limitations in the current state of the art also programming,Section 2 of text the analysis, article brieflyand machine surveys learning law schools’ in legal raise interesting challenges for research and development. education. Section 3 then describes the 2020 version of With more research, automated contract review might the course at Pitt, including the students, instructors, be able to draw inferences about due diligence across and the course sessions in overview. Along the way, we the content of multiple provisions in a single contract or explain why we selected the readings and discuss various across multiple contracts. It might also integrate other pedagogical techniques employed, including requiring stu- kinds of data available in a virtual data room such as tax dents to prepare one-page abstracts of selected readings. returns, leases, and employee information. Also, of interest Section 4 discusses each of the course’s four parts in are the (still) primarily academic research efforts around some detail, beginning with the ways in which the in- “argument mining”, which apply legal text analytics to structors used programming environments to introduce 11 identify argument-related information in case corpora. Python, ML, and NLP. The section describes the introduc- Rather than focus exclusively on computational models tion to AI and Law concepts and research with a focus on of legal reasoning that have been developed in AI and Law, computationally modeling case-base legal argument, the this course focuses on the legal text and data analytics more detailed treatment of how to apply text analytics to employed in the new legal apps that are changing legal legal texts including statutes and cases, and discussions practice. Increasingly, the legal profession employs – and of issues of fairness, privacy, and liability affecting AI. relies upon – text analytic tools. Since, as soon as they Section 5 explains how the instructors prepare the begin practice, many law students will confront the need to employ these tools and to rely on their outputs, they the term with project ideas and databases of text, forming need to understand how these tools work and what their teams,students working for the onfinal the course projects projects, as a team beginning and obtaining early in limitations are.12 In addition, as attorneys who have some knowledge of the technology, they can better participate reporting results. interim feedback, and finally completing the projects and in working with technical personnel and researchers in Finally, in Section 6 we draw some salient comparisons addressing the limitations and improving the tools. between the 2019 and 2020 versions of the course. In Section 7, we report what worked well and what did not, 1.3 ARTICLE OVERVIEW the students’ reactions, and lessons learned for future Given the intensity of the law school academic com- offerings of the course. munity’s interest in how to prepare students for legal practice with the new data analytic technologies, this 10. AI AND LEGAL DATA ANALYTICS IN LAW SCHOOL article describes the course, its goals and implementation, CURRICULA and lessons learned. A continuing focus of the article is A substantial number of law school curricula include how we taught law students skills of programming and courses that address AI. “As of the 2018 Spring Semester,

11 See, for example, the seminal work of Mochales and Moens (2011). Argument mining has since become the focus of a research subcommunity and its ARGMINING workshop series (e.g., https://argmining2020.i3s.unice.fr/). 12 There may also be professional ethical responsibility. American Bar Association Rules of Professional Conduct Model Rule 1.1 states that “A lawyer shall provide competent representation to a client. Competent representation requires the legal knowledge, skill, thoroughness and preparation reasonably necessary for the representation.” Comment 8 to Rule 1.1 states (emphasis added), “To maintain the requisite knowledge and skill, a lawyer shall keep abreast of changes in the law and its practice, including the benefits and risks associated with relevant technology,….” https://www.americanbar.org/groups/professional_responsibility/publications/model_ rules_of_professional_conduct/rule_1_1_competence/comment_on_rule_1_1/ Accessed 10/12/2020

138 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

... roughly 10% of ABA-accredited law schools offer at used to deliver legal services better, faster, and cheaper.” - (Perlman 2017, p. 6) One way some law schools expose students to these least one course explicitly concerning artificial intel new tools and methods is by embedding examples and ligence.” (Dalton 2019) This reflects a recognition in the changing the practice of law.” (Savkar 2019) lessons in substantive or skills-oriented law school legal academic community that, “Artificial intelligence is courses. In substantive courses, AI tools may be “woven 2.1 WHY TEACH AI IN LAW SCHOOL AND HOW? into traditional classes, such as product liability classes There probably is no consensus about why AI should and regulatory classes, administrative classes and the be included in the law school curriculum and how best to like.” (Council 2019, quoting Robert Kantner, Jones Day) include it. “A broad range of ideas are taking root, some- Indeed, “some law schools have begun the integration what experimentally, across the law school community.” of technology within the law school curriculum by add- (Savkar 2019) ing a few days of eDiscovery to Civil Procedure, offering One reason cited for offering a course on AI is to help courses in the law of technology, such as Cyber Law, or law students understand how the new technologies are touching upon data and communication security during changing models for delivering legal services and to en- Legal Profession.” (Eicks 2012, p. 5-9) able them to participate in and even design those changes. AI tools and techniques have also been embedded in skills-focused law school courses. Oklahoma University If we can help students understand that technology, (OU) College of Law “had a transaction law practicum where [they]’re showing them the current smart contract and specifically AI, can create a much more streamlined, software and contract analytic software out there.” (Dalton legal services, and reorient or recalibrate what it means efficacious means of connecting lawyers to consumers of 2019, quoting OU College of Law Professor Kenton Brice) to provide legal services by lawyers, then that’s an enor- The fact that these skills include performing legal research provides another natural opening for introducing AI since students to the value and capacity of law to provide access mous benefit for us as legal educators in educating our “every single one of these platforms [such as Westlaw to justice. (Dalton 2019, quoting Daniel B. Rodriguez, Dean, and other research tools] to some degree or another is Northwestern Pritzker School of Law) using a form of AI implementing legal research.” (Dalton This reason includes preparing law students to be 2019, quoting OU College of Law Professor Kenton Brice) “practice-ready” for the new types of legal work the tech- AI may also be addressed in courses focusing on busi- and related cutting-edge technologies, law schools face ness, tech start-ups, and innovation “that train students nology enables. “With the explosion of artificial intelligence in the business of law and operations, get them to think like entrepreneurs, and have them improve processes, huge opportunities to create graduates who efficiently and gather data, and use technology.” (Dalton 2019, quoting courseconfidently affords rely an on opportunity technology to “to better expose serve students their clients to the Daniel Linna, now Director of Law and Technology Initia- sameand run tools more practicing efficient attorneys practices.” use” (Dalton (Dalton 2019) 2019) An or AIto tives and a Senior Lecturer at the Northwestern Pritzker School of Law) For example, a legal clinic at Cardozo Law of legal jobs in the future—perhaps legal data analyst or aims to take “students through the life cycle of an actual machineprepare them learning to fill legal the expectedspecialist.” “entirely (Miller new 2019, categories quoting area technology startup, from formation through IPO, and University of Colorado Law Professor Harry Surden) More beyond.” (Dalton 2019, quoting Professor Aaron Wright) generally, the goal is said to be to produce lawyers for the Law courses incorporating AI may aim to expose students to the engineering side of legal processes and sophisticated, commercially astute, hybrid professionals,” modes of delivering legal services. A goal at Suffolk Law (Dalton21st century 2019, who quoting are “flexible, Richard team-based, Susskind) “able technologically- to identify School is to expose students to “concepts like legal project how technology and other innovative methods can be management and process improvement, legal design ...,

Law in Context, Vol 37, Issue 1, 2020 139 ISSN: 1839-4183 automated legal document assembly, expert system tools, or criminal justice, the issue of data bias is a particularly electronic discovery, and other areas as well.” (Perlman important focus. “These predictive tools need to be ex- 2017, p. 7) amined to determine any algorithmic bias and what data Once process improvement has begun and continuous was originally provided to make such an assessment.” (Reid 2018, p. 481) At Stanford Law School, Professors Daniel Ho and David Engstrom teach a class on federal improvement is part of an organization’s culture, we find and quality and improve outcomes through document agencies’ use of AI. many opportunities to exponentially increase efficiency It’s hard to engineer a solution for algorithmic bias systems, and other readily available, basic technologies. without understanding how decades of anti-discrimination automation, workflow automation, rules-based expert Done properly, getting control of processes also means law have dealt with these challenges, nor is it possible to creating legal-services delivery metrics and capturing understand what the future of anti-discrimination law data from these processes .... [t]his ... also helps position organizations to identify opportunities for data analyt- the sources of algorithmic bias. (Council 2019, p.2, quoting should look like without a fine-grained understanding of Professor Daniel Ho, Stanford Law School) quoting Daniel Linna) An important focus in some of the courses is thus on ics and data-driven artificial intelligence. (Dalton 2019, At the University of Minnesota, “the coding for law- teaching students how to assess the extent to which one yers course presents an overview of the changing role of can appropriately rely on the technology in the practice lawyers, the role of lawyers as project managers, the key of law. “Inappropriate reliance on algorithms is certainly responsibilities of a project manager, and the organiza- a theme we weave into more than one course, and there's tional and operational structure of a legal tech project.” a constant reminder that technology is there to serve (Contreras and McGrath 2020, p. 323) people; and that when it can result in injustices, we have Several law school courses are addressing AI in order to be vigilant about how it's used.” (O’Grady 2018, quoting to help law students to better understand “the emerg- Professor Gabriel Tanenbaum) ing legal issues around AI” (Council 2019) and how to regulate AI in society. 2.2 TEACHING LAW STUDENTS ABOUT CHALLENGES OF The law schools at Columbia University, Harvard RELYING ON THE NEW TECHNOLOGY University and Stanford University have within the past Biased data and algorithms are not the only causes of two years introduced courses that touch on AI topics problems confronting appropriate reliance on the new including algorithmic bias and civil rights; liability for autonomous vehicle accidents; and regulating AI-powered will disrupt legal ethics and professional responsibility, applications and platforms. The goal is to give future Murphytechnologies. and Pearce In their (unpublished) article on how pointartificial out intelligence that today, lawyers, judges and regulators greater exposure to the attorneys are in the unenviable position of relying on technology, which is expected to become increasingly new technologies even though they are unable to verify embedded in society and raise new questions around the quality of the services. “[H]ow do lawyers know that matters of legal responsibility, due process and regula- they can reasonably rely on the legal search results of tory frameworks. (Council 2019) These societal issues include protection of privacy and to understand the algorithms....” (Murphy and Pearce these businesses? Lawyers lack the software expertise data bias. “If students want to work on the cutting edge unpublished) of things like AI, privacy, blockchain technology et cetera, A course on legal data analytics and AI can help law you ... have to begin to really understand how these things students to understand the assumptions underlying the work.” (Dalton 2019, quoting Professor Aaron Wright) new machine learning and text-analytic technologies and Given the increasing reliance on data analytics and to learn how to assess the capabilities and limitations of these methods as they are applied generally in society machine learning by agencies administering social benefits

140 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

the information about the features’ weights is distributed appropriate for applying machine learning to texts where theand techniquesmore specifically may seem in legal very practice. counter This intuitive. is especially Harry an explanation. across the network’s nodes and difficult to fashion into Surden characterizes machine learning as “producing Compounding the problem of reliance are the extraor- intelligent results without intelligence.” dinary claims made on behalf of the technology by com- For example, when a machine learning-based computer mercial providers and the press as documented, such as: system produces a translation, it usually does so using The question is how—and when—machines will enter statistical associations. However, such a pattern-based the legal economy in full force and render lawyers obsolete. machine learning approach—while often producing de- (Simon et al. 2018, p. 237) cent translations—does not actually involve the computer There is a lot of legal work that is routine…. But that “understanding” what it is translating or what the words routine work, sifting through documents for relevant mean in the same way a human translator might. (Miller information, is wrapped in language, which had pro- 2019, quoting Professor Harry Surden) tected lawyers from the effects of automation, but no This is true as well of applications of machine learning longer. Indeed, language is no longer a barrier because to legal documents and texts. According to Miller, Surden of advances in `natural language processing’ (NLP) technologies, as emphasized in the most recent McKinsey & from current machine learning approaches,” including Company report on technology and employment. (Simon cautions that “a limited number of legal tasks may benefit tasks in “e-discovery document review, litigation predic- et al. 2018, p. 271) tive analysis, and legal research.” “Core tasks still require a great amount of problem solving and abstract reasoning believe that we are on our way to a future where at least Behind the numbers, many private law firm leaders that pattern recognition or machine learning is unable to the younger lawyers will be replaced by robots. (Simon 13 replicate.” (Miller 2019) et al. 2018, p. 285) That places lawyers relying on ML technology in an odd The claims are even more remarkable given that the position. Increasingly in legal practice, they will need to new legal apps cannot read legal text like human lawyers rely on ML tools that draw inferences from textual data, can (Ashley 2019). but those tools have very limited information about what the texts mean. It would help if the ML technology could 2.3 TEACHING LAW STUDENTS HOW TO EVALUATE THE adequately explain the basis of its predictions and results NEW TECHNOLOGY in terms that lawyers could understand. Lawyers would then be able, at least, to assess the system’s reasons for In such circumstances, assessing whether and how its predictions, to decide if they accept those reasons, and much to rely on inferences based on machine learning is to factor the reasons into a determination of whether to rely on its advice. Most of the AI programs that employ can help law students to understand the extent to which difficult. A course in applied legal data analytics and AI machine learning, however, cannot explain their results such claims exaggerate the current state of the art. in terms lawyers would understand. Most employ neural Most everyone would agree that the impact of algorithms, network architectures for which generating explanations is problematic. Much depends on the features that the on everyday life, institutions, and society will only grow, computational technologies, and artificial intelligence machine learning considers. Neural networks may dis- and rapidly. Yet most law students and lawyers lack the cover features that are statistically weighty but that do foundational knowledge to generally explain how these not correspond to legal concepts lawyers would expect to technologies work, much less assess them. (Linna 2018) see in an explanation. Even if the features do make sense,

13 See (Surden 2014, p. 101) “there are a subset of legal tasks often performed manually today by attorneys, which are potentially partially automatable given techniques such as machine learning, provided the limitations are understood and accounted for.”

Law in Context, Vol 37, Issue 1, 2020 141 ISSN: 1839-4183

It would help students to learn in some detail how the legal inferences are generated and how machine learning can be evaluated. As it appears today, There is no doubt that LegalTech will automate “legal work,” such as contract drafting, legal risk management, and dispute resolution. If legal work will be dependent on and performed by algorithms in the future, it is crucial for the future lawyers (and lawyers who are at the start of their careers) to get a better understanding of “data

analytics” and “artificial intelligence.” Since all these of understanding code will be of value in the medium to near-benefits technologies are code based, the benefits long-term. (Fenwick et al. 2018, p. 17). FIGURE 1. Assessing AI as per D. Linna’s class on AI and Legal Reasoning In some of the law school courses focused on AI, the instructors and students are pioneering approaches for on the new technologies and their applications, a kind assessing text analytic techniques as applied in the legal of activity which, it is anticipated, will become increas- - ingly important in legal practice. The courses provide gal Reasoning class, for example, an aim is to develop a a venue for law students to learn how to communicate domain. In Daniel Linna’s Artificial Intelligence and Le with professionals who do not necessarily share their legal-services delivery like that in Figure 1. assumptions or methods and who do not understand framework for assessing artificial intelligence as used for - their legal language. formance suggests, students need to understand that the AI courses focused on ethical and regulatory issues, for degreeAs the and first criticality question of theabout reliance the required depends level on the of “useper example, engage multidisciplinary students in evaluating case”: how the technology is being used and the kinds of the technology. See, e.g., Harvard Law School’s course problems that it is being used to solve. called “The Ethics and Governance of AI,” which is part of The Dutch ethicist Pim Haselager notes a distinction a joint effort with Massachusetts Institute of Technology between two types of “use cases”: I. AI as a component of (Council 2019, p.2). The course of Stanford law professors a modular system delivering results that assist a human to Daniel Ho and David Engstrom on federal agencies’ use make a legal decision, and II. AI as the overall legal decision of AI also combines law students and computer science maker or recommender. The former keeps humans on (or students (Council 2019, p.2). in) the loop, while the latter keeps them “under the loop” The multidisciplinary interactions may focus on design- as Haselager put it, where they are much more likely to be ing and implementing the new technologies. forced to rely blindly and less critically on the AI system. “Legal tech” will profoundly disrupt the legal profession (Haselager 2019) Understanding the implications of these and since these technologies are code-based[,] lawyers different types of use cases is crucial for establishing how need to be able to understand and talk about code to much reliance would be required, and for appropriately participate in the design of such legal technologies and to designing methods for keeping attorneys in the loop as maximize their usefulness in supporting all legal work…. they apply the new technologies. [A] combination of lawyers, mathematicians, and coders taught the course, and the students are encouraged in do- 2.4 PREPARING LAW STUDENTS TO INTERACT WITH ing their assignment to partner up with students in other TECHNOLOGISTS sense, the kind of multi-disciplinary teams of the future Law school courses in AI help students to engage with faculties to put together their final assignment. In that students and faculty from other disciplines who focus is re-created in the classroom. (Fenwick et al. 2018, p. 29)

142 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

... “Co-creation” involving partnerships between analytics perform. Some argue that “Law students need developers and nondevelopers will be crucial to build- to learn programming skills to understand how technolo- ing a “better” digital future. And understanding coding gies can optimize their work and make the process of enables lawyers to constructively engage with coders, providing services easier.” (Pivovarov 2019) This is often programmers, and other software developers. (Fenwick characterized as “coding for lawyers” but may be subject et al. 2018, p. 11, 29, 30f) to some disclaimers illustrated below. In a similar vein, a Legal Analytics and Innovation [T]he “Coding for Lawyers” course is … not about Initiative at Georgia State University School of Law aims teaching students “how to code,” but about making them to enable “law students … to collaborate closely with realize how important it is to think about our relationship computer science and business students … to design with new technology and technology experts. In this way, complex technologies that solve previously unsolvable students can see the new opportunities that technology legal problems (such as predicting to a high degree of creates, but also think about the new issues (practical and ethical) that such technologies create. For this reason, we by a large set of parameters).” (Savkar 2019) plan to introduce the basics of machine learning and deep accuracy how a particular judge will rule in cases defined 15 While some students may participate in design, per- learning to the program. (Fenwick et al. 2018, p. 30) haps more will be called upon to oversee or supervise Beyond teaching coding to help law students under- the application of the technology in practice, activities stand the practical and ethical issues the new technolo- which also will involve them in multidisciplinary interac- gies raise, practice with coding may help law students to tions with clients and staff members. “Students entering communicate with technologists. the current and future legal employment market must [W]e are not suggesting that law students should become understand the implications and impact of AI and related professional coders. After all, coding is complicated and to technologies on the practice of law and be prepared to become a serious coder does take time. However, we do oversee their implementation and the resulting processes.” believe that the ability to understand and communicate 14 (Dalton 2019) with coders is a necessary skill for the lawyer of the future. [S]tudents should be taught to question the data produced by an Al program and avoid over-reliance on technology only by reading or hearing about it but by participating As such, we think that law students will benefit … [n]ot alone…. [O]verly broad tasks asked of Al programs can in and experimenting with coding projects in a classroom result in inaccuracies. Such operations need review and environment. (Fenwick et al. 2018, p. 30) Our answer to how much programming law students need is enough to enable at least some of them to evalu- significant human oversight. … [J]unior lawyers will now lead or assist employees assigned to the IT or Al depart- ate the new technologies as they are applied in legal find themselves in supervisory roles as they are asked to practice. Murphy and Pearce (unpublished) warn of the skills in the classroom and encourage group projects and growing need to teach students how to evaluate the new ment… Law schools need to find ways to emphasize such team-oriented case study discussions. (Reid 2018, p. 484) technologies. [L]egal services workers, including lawyers, will not 2.5 HOW MUCH PROGRAMMING TO TEACH LAW need to code but they will need to understand how to STUDENTS? apply and evaluate new methodologies powered by arti- A still open pedagogical issue concerns how much programming a law student needs to know in order to marginal parts of the curriculum, whether a class session ficial intelligence. Pieces of these skills only exist today in understand how and how well machine learning and text or course on electronic discovery, or a clinic on developing

14 See also, (Contreras and McGrath 2020, p. 323) 15 See also, (Contreras and McGrath 2020, p. 324, “Students are not expected to be fluent coders by the end of the course, but to have an appreciation and understanding of the capabilities of coding.”)

Law in Context, Vol 37, Issue 1, 2020 143 ISSN: 1839-4183

apps the Department of Electrical and Computer Engineering, central to the provision of legal services it will have to and one Pitt undergraduate student from the School of for consumers. Once artificial intelligence becomes become central to legal education. (Murphy and Pearce Computers and Information. A Duquesne University Law unpublished) School professor with a longtime interest in AI and Law 16 We agree that law students need to engage closely also attended on a regular basis. with computer code in order to understand how the technologies work and how to evaluate them. In the Ap- 3.2 THE INSTRUCTORS plied Legal Analytics and AI course, we stepped law stu- In order to teach this course successfully, the instructors dents through programming with Python notebooks and need to combine an ability to: (1) teach students how to organized teams combining law students and technical create computer programs in Python that apply machine learning and natural language processing to textual and the students implemented some Python code, designed other legal data, (2) help students design experiments students to work on final projects. In each of the projects, empirical evaluations, ran the experiments, obtained the evaluating applications of ML and NLP to such data, (3) results, and then engaged in error analysis to see what explain how these experiments relate to research in AI and the programs missed and why. Law and to legal practice tools, and (4) instruct students about the legal issues that arise in contexts involving AI 11. COURSE DESCRIPTION and machine learning. As noted, the most recent version of the course in It is likely that two instructors with complementary Applied Legal Analytics and AI took place at Pitt from skills and knowledge will be required to cover all four January through April 2020. abilities, and it is also helpful for the instructors to be able to share the responsibilities of teaching and guiding 3.1 THE STUDENTS students. That was certainly our experience in teaching the course. As noted, Jaromir Savelka17 and Kevin Ashley18 One of the unique features of the course was a mix 19 of students from law backgrounds and technical back- co-taught the spring 2020 course; Matthias Grabmair grounds. The course comprised a total of ten students: and Ashley co-taught in spring 2019. Their combined skills and knowledge covered the course’s wide-ranging demands. As we learned by experience, it was also crucial five Pitt law students, four Pitt graduate students from 16 The previous version of the course was offered at Pitt and CMU from January through April 2019. That version of the course comprised a total of fifteen students: eight Pitt law students, three Pitt graduate students from the School of Computing and Information, one Pitt graduate student from the Katz Graduate School of Business and three CMU undergraduate students. 17 Jaromir Savelka is an Adjunct Professor of Law at the University of Pittsburgh and Postdoctoral Fellow in the School of Computer Science at Carnegie Mellon University. For four years as a Lecturer at the School of Law of Masaryk University in Brno, the Czech Republic, he taught Legal Informatics and related legal technology courses at the graduate and undergraduate levels. In 2013 he enrolled as a graduate student in the University of Pittsburgh Intelligent Systems Program where he developed expertise in computer programming and applying machine learning techniques to legal texts. In April 2020, he successfully defended his dissertation, entitled “Discovering Sentences for Argumentation about the Meaning of Statutory and Regulatory Terms.” For the last three years, he has worked as a Data Scientist at Reed Smith, LLP, Pittsburgh, where he gained experience in evaluating and developing legal AI systems supporting eDiscovery and due diligence. He has a graduate degree in law and an undergraduate degree in computer science, both from Masaryk University. 18 Kevin Ashley, a Professor of Law and Intelligent Systems and senior scientist at the Learning Research and Development Center at the University of Pittsburgh, has expertise in artificial intelligence and law and is a Fellow of theAmerican Association of Artificial Intelligence. He is co-editor in chief ofArtificial Intelligence and Law, the journal of record in the field of AI and Law and the author ofArtificial Intelligence and Legal Analytics: New Tools for Law Practice in the Digital Age (2017). He has a JD degree from Harvard Law School and a PhD degree in Computer and Information Science from the University of Massachusetts. For five years before beginning his graduate studies in computer science, he practiced law as an associate at White & Case in New York City. His research has focused on developing computational models of case-based legal argument and ways to automatically analyze texts of legal decisions to populate those models. 19 Matthias Grabmair is an Assistant Professor in the Technical University of Munich Department of Informatics. At the time of the spring 2019 version of the course, he was a Systems Scientist at Carnegie Mellon University's Language Technologies Institute teaching in CMU's Master's Program in Computational Data Science. He worked on solving problems in domain-specific question answering and knowledge engineering, including analysis of legal documents. His expertise reflects studies and experience in artificial intelligence and law, knowledge representation and reasoning, computer programming, natural language processing, applied machine learning, information retrieval and computational models of argument. He holds a PhD degree from the University of Pittsburgh’s Intelligent Systems Program. He has an LLM degree from the University of Pittsburgh and a Dipl.-Jur. degree from the University of Augsburg, Germany.

144 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 for two instructors to be able to share the responsibilities Law Practice in the Digital Age (Ashley 2017) would cover of teaching students, providing feedback, and guiding most of the subtopics in the second area, AI and Law. As student teams working on course projects. For additional to the third area, Legal Practice and AI, we selected law - review articles to illustrate practical applications of text viders lectured the class on applications in legal practice suchhelp, asguest automating speakers contractualfrom law firms due or diligence legal service or patent pro in selected areas of the law. analytical tools and the legal ramifications of applying AI review. It also makes sense to invite more traditional As far as ordering the topics and readings, there were legal scholars to address such topics as the use of machine learning in empirical legal research or legal issues time-consuming for law students to learn programming affecting machine learning such as intellectual property skills.two driving In addition, factors. the We students knew that needed it would to learn be difficult such skills and rights in data, effects of electronic datamining on privacy, in time to apply them in working on their course projects. bias in ML-based decision-making, and product liability Hence, teaching programming skills and providing an inof autonomous systems. tuition of how they would be relevant became the initial and extended focus of the course. 3.3 COURSE OVERVIEW As illustrated in Table 1, we organized the lectures In overview, the course content comprised three ma- and discussion sessions around a progression of themes jor topics: (1) Text Analytics: introducing programming in four parts. The readings and lectures associated with with Python, machine learning and natural language, and each part are discussed in the next section. focusing ultimately on using text analytic techniques to In Part I, the goal was to introduce the law students extract argument-related information from case texts and to some selected aspects of programming with Python, to predict case outcomes. (2) AI and Law: introducing and to basic concepts in machine learning and natural computational models of legal decision-making, arguing language programming. We also sought to convey to the with rules and cases, and predicting case outcomes and non-law students some intuitions about legal reasoning how these models could interact with text analytics to and provide all of the students with a glimpse of the im- improve legal information retrieval. (3) Legal Practice pact of AI on legal practice. and AI: introducing students to legal text analytic tools Part II provided a foundation in prior research in the that have been applied in legal practice and to substantive legal domain topics that affect AI such as data privacy, decision-making, modeling case-based legal reasoning, and machine learning bias and fairness, and liability of au- usingfield of the Artificial models Intelligence to predict and Lawexplain on supportinglegal outcomes. legal tonomous vehicles. It described how legal information retrieval works and Designing a syllabus for conveying this content presented suggested how to make legal IR more intelligent. at least two challenges: selecting appropriate readings Part III focused on techniques for technology assisted and deciding on the ordering and duration of each topic. review, extracting information from statutory texts, ex- Since no textbook appeared to cover all of the targeted tracting argument-related and other information from case texts, predicting outcomes directly from case texts, area, Text Analytics, we designed workbooks of program- and segmenting case decisions by function.20 content, we settled on the following plan. For the first ming exercises supplemented with accessible research In Part IV, the last lectures in the course focused on papers to illustrate applications of the techniques to legal domain topics that affect AI including patent law, various legal tasks and use cases. Selected chapters from Artificial Intelligence and Legal Analytics: New Tools for

20 The University shutdown due to covid-19 occurred at the end of Part III and continued through Part IV. As a result, classes were cancelled during the week of March 16. Upon resuming the course during the week of March 23, all lectures were delivered via Zoom or Panopto

Law in Context, Vol 37, Issue 1, 2020 145 ISSN: 1839-4183

TABLE 1. Course Sessions in Spring 2020

legal regulation of data privacy, machine learning bias presenting preliminary proposals. Two sessions at the and fairness, and legal liability of autonomous vehicles. 21 Table 1 also highlights two other aspects of the course: projects. The course projects are discussed in Section 5. end of Part IV were devoted to team presentations of final the four homework assignments and four project sessions. The homework assignments all occurred in Part I and 12. COURSE CONTENT BY PARTS served as the primary means for introducing students to a The course content was presented primarily through tool for legal question answering, introducing the non-law weekly lectures and readings. In support of the lectures, students to legal reasoning and a bit of legal practice, and in most weeks students were assigned to read one or two engaging the students in working with Python programs. topical AI and Law research papers per week. Table 1 shows when the homework was assigned; it was For selected readings students were asked to submit due a week or two later. These homework assignments a brief, one-page abstract of the reading a day or two are discussed in Section 4, Part I. The second aspect was a semester-long focus on the of the abstracts: a three-sentence summary of the read- course projects. As illustrated in Table 1, two project ing,ahead three of the positive class aspectstime. A oftemplate the paper specified as well the as threeform sessions at the beginning and end of Part I focused the criticisms (about one sentence each), three questions students on identifying projects, assembling teams, and the student would like to ask the authors because the

21 The spring 2019 course covered similar topic matter. The primary differences were a more abbreviated treatment of AI and Law research, an extended treatment of the application of machine learning to, and quantitative models of, Supreme Court Justice voting (three sessions), fairness in machine learning, and the disparate impact of ML in bail recidivism prediction and credit scoring (one session), and a guest lecture about automated contract review for due diligence (one session).

146 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 student either did not understand something or would the court system and to such concepts as burden of proof, like more information, and a brief statement of how the choice of law, types of motions to dismiss, the role of dis- reading relates to the student’s research or other interests. tinguishing, and the limits of precedential constraint.22 The abstracts were due a day or two before the session In Lola v. Skadden, for example, the court held somewhat so that the instructors could include answers and target controversially that attorneys do not exercise “legal discussion to address the students’ points in the abstracts. judgment” when they perform services that a computer could provide while reviewing documents (Simon et al. 4.1 PART I: INTRODUCING PYTHON, MACHINE LEARNING, 2018, p. 253). We looked at a subsequent decision from NATURAL LANGUAGE PROCESSING AND AI’S EFFECT the same federal district, in which a district court readily ON LEGAL PRACTICE distinguished the decision in Lola v. Skadden despite the An interdisciplinary course requires introducing students similarity of the fact situations. The Dynamo decision il- to many novel subjects. In Part I, we introduced them to lustrated how courts are adapting to the role of machine examples of legal text analytic tools and their effects on learning in determining the relevance of documents in legal practice, to some selected aspects of programming e-discovery. with Python, to basic concepts in machine learning (ML) Homework assignment 1 introduced the students to and natural language processing (NLP), and to a brief his- using one of the new legal apps for an exercise in legal research and memo writing and the non-law students to some basics of the legal system. The LUIMA system, a tory of research in Artificial Intelligence and Law. 4.1.1 INTRODUCING LEGAL TEXT ANALYTIC TOOLS prototype legal question answering system developed by AND THEIR EFFECTS ON LEGAL PRACTICE Matthias Grabmair, accepts questions dealing with veterans’ An initial goal in Part I was to survey the new legal text claims for compensation for posttraumatic stress disorders analytic tools and methods. The readings (Ashley 2019) (PTSD) and retrieves answers (i.e., extracted sentences similar to the question) from a database of decisions of and lecture illustrated some techniques for applying 23 machine learning to legal texts, using citation network the Board of Veterans Appeals (BVA). diagrams, and question answering in the apps described The students began by reading a brief but informative in Section 1. Their limitations were also introduced: the cases, using statutes, synthesizing cases, and drafting introduction to the legal24 system, reading and briefing the new machine learning techniques have in generating a legal memorandum. The readings introduced basic explanations,fact that apps and cannot the constraints read like lawyers, imposed theby the difficulties need for legal reasoning concepts (e.g., standards of proof, pre- manually annotating training sets for machine learning. sumptions, and preponderance of the evidence). Using These limitations, we suggested, counteract the hype sur- these as a guide, students read a decision of the Board of rounding the impact of AI on the practice of law and its Veterans Appeals and prepared a brief of the case. They predicted effects on legal employment (Simon et al. 2018). then read a problem scenario and a list of questions to Two recent decisions concerning e-discovery and pre- guide them in identifying some of the issues that a legal dictive coding helped introduce the non-law students to analysis would need to address.

22 See Lola v. Skadden, Arps, Slate, Meagher & Flom LLP, 620 F. App'x 37, 620 Fed. Appx. 37 (2d Cir., 2015). Dynamo Holdings et al. vs. Commissioner of Internal Revenue, U.S. Tax Court, Docket No. 2685-11, 8393-12; July 13, 2016. 23 The BVA is set in a reasonably limited yet intuitive legal universe. In performing administrative adjudications, it produces court-like opinions that follow typical legal argumentation, writing, and citation patterns. Running the assignment in the experimental LUIMA system rather than a commercial search engine had several advantages. First, it reduced the complexity of the interface and focused students on the research task by not providing headnotes or other forms of topic-based linking across decisions. Second, it made sure that only sentences from BVA decisions were retrieved as answer candidates, thereby preventing exposure to cases that use similar vocabulary but stem from different jurisdictions. Third, the student search queries and behavior in the system could be tracked. 24 Excerpts from (Brostoff and Sinsheimer 2013).

Law in Context, Vol 37, Issue 1, 2020 147 ISSN: 1839-4183

For each of the questions, the students were asked to

source[s] for the answer. Students could submit these questionsfind and write or variations a very brief of themanswer directly and to to cite the the LUIMA case search tool, using the interface shown in Figure 2. The program returned a list of sentences (left) rank-ordered by their responsiveness, a summary of the case in which the sentence appeared (right), and a link to the full case

similar prior decisions by navigating to the decisions from whichtext (center). the candidate The students answers were had asked been to extracted, focus on finding and to identify patterns of arguments with which the BVA decides legal issues like those in the scenario. The students could also mark answers as more relevant to the query or less, a kind of feedback to LUIMA. FIGURE 2. Legal question answering with luimasearch Finally, the students were asked to imagine that they The scenario involved a mechanic in an Air Force base worked in a Veterans Legal Clinic. The director has asked who witnessed a fatal crash of a helicopter on which he them to prepare a legal memorandum assessing legal was scheduled to work. This led to alcohol abuse, a gen- arguments for an appeal to the BVA of the VA’s rejection of the client’s PTSD claim. The instructors advised the Veterans Administration (VA) dismissed the veteran’s students to employ the results of their legal research with claimeral discharge, for disabilities morbid as flashbacks,not substantiated. and drug He abuse. has asked The LUIMA in order to prepare their memos and provided a an attorney (the students) to prepare arguments for ap- list of general criteria for assessing legal memoranda. peal to the BVA. In order to prepare arguments for an appeal, the students a chance to work with a basic legal question answering In this way, the first homework assignment gave students - system and to experience using the system to perform ards for proving issues related to PTSD claims and how legal research and generate a legal memorandum. Upon toneeded meet tothem. find Eight out more questions about at the the relevant end of the legal scenario stand review of the students’ memos, the instructor found them guided their inquiries with the LUIMA system, such as: to be quite acceptable. 1. What is the overall legal standard for determining if 4.1.2 INTRODUCTION TO PYTHON PROGRAMMING - 2. What is the legal standard for proving that a veteran a veteran is entitled 25to compensation for disability? ming with Python. No prior experience with program- The first two lectures focused on introducing program currently has PTSD? comfortable reading small pieces of code. Later we asked ming was assumed. Our goal was first to make students

25 The other questions are: 1. What are the special regulations concerning proving service connection for PTSD? 2. What kind of evidence has been accepted to show a veteran has PTSD? 3. What kind of evidence has been accepted to show PTSD was service connected? 4. What if the evidence for and against service connection is evenly balanced? 5. What if the veteran is currently an alcoholic? 6. When is a veteran found to not be credible?

148 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 them to edit the code in order to adapt provided examples as we tried our best to inform them as to the necessary to their needs. Both lectures were organized as a guided knowledge and skills for successful course participation. walkthrough of prepared Jupyter notebooks26 hosted on By including programming assignments in the required the Google Colaboratory (Colab) platform.27 The idea was homework assignments, however, we did try to motivate to provide students with small snippets of executable code them to invest the necessary time and effort. with which they could explore coding outside of class, as - gramming was the second homework. It was based on a well-knownThe final Spelling activity Corrector focused ondescribed teaching by PythonPeter Norvig pro well. The first lecture covered several foundational concepts The second lecture addressed additional concepts, such (2007). This simple program learns the frequencies of including code layout, primitive types, and flow control. as modules, functions, and data structures (lists, tuples, words from the document corpus and utilizes the model 28 sets, and dictionaries ). to correct the spelling of a word provided on input. The We did not expect that students with no prior program- homework was designed to introduce students to a de- ming experience would be able to learn enough from just tailed analysis of a small-sized computer program. The two in-class lectures. While we did not aim to turn such goal of the exercise was to teach them how to follow the that it would require considerable effort for them to ac- in each step of the execution. This, in turn, would enable quirestudents the intoknowledge full-fledged and skills programmers, necessary weto successfully recognized flow of the program and to understand what is happening participate in the rest of the course. Hence, we assigned or recommended several other activities to further the “Howstudents It Works: to reason Some about Probability potential Theory” weaknesses section or offlaws the aim. First, the students with no prior programming programin the program. webpage Students (Norvig were 2007) instructed and then to to first think read about the experience were required either to peruse a textbook the program in general. They answered questions such covering the basics of Python programming29 or take as: How should a spellchecker behave if it sees a perfectly an online course covering the same material during the with pointers to the PEP 8 – Style Guide for Python Code, correct word? What should a spellchecker ideally do in first few weeks. Furthermore, we provided the students30 case it sees an incorrect word? What does it mean for a the documentation to several modules (input/output, The second part of the homework was designed to 31 32 33 word to be correct or incorrect? pickle, json, and xml ), resources covering encoding teach students how to read the source code. The assign- 34 (Zentgraf 2015) and regular expressions, and an extensive - 35 list of Python resources for non-programmers. As the tions related to the role those pieces of code play in the students’ backgrounds varied greatly, it was ultimately ment pointed them to specific lines while asking ques up to them to decide how much effort they would invest them to investigate the role of the imported modules and program. For example, one of the first questions asked

26 Jupyter notebooks (https://jupyter.org) are web-based, interactive environments for programming code in Python (and other languages), executing code, loading and inspecting data, conducting analyses, and plotting the results. Using Jupyter notebooks, an instructor can set up a classroom exercise or homework assignment in a file (an .ipynb file) providing students with code, a series of tasks, and instructions how to perform them with a provided file of data. Students complete the assignment in the notebook and turn it in to the instructor for grading. 27 WELCOME TO COLABORATORY, https://colab.research.google.com/ Accessed 6/7/2020 For a similar pedagogical purpose Contreras and McGrath (2020, p. 330) employed the Repl.it integrated development environment. 28 A Python dictionary is a collection of items which is unordered, changeable and indexed. 29 We recommended the freely available (Halterman 2018). 30 THE PYTHON TUTORIAL: INPUT AND OUTPUT. https://docs.python.org/3/tutorial/inputoutput.html Accessed 7/7/2020 31 PICKLE - PYTHON OBJECT SERIALIZATION. https://docs.python.org/3/library/pickle.html Accessed 7/7/2020 32 JSON – JSON ENCODER AND DECODER. https://docs.python.org/3/library/json.html Accessed 7/7/2020 33 THE ELEMENTTREE XML API. https://docs.python.org/3/library/xml.etree.elementtree.html Accessed 7/7/2020 34 REGULAR-EXPRESSIONS.INFO. https://www.regular-expressions.info/ Accessed 7/7/2020 35 PYTHON FOR NON-PROGRAMMERS. https://wiki.python.org/moin/BeginnersGuide/NonProgrammers Accessed 7/7/2020

Law in Context, Vol 37, Issue 1, 2020 149 ISSN: 1839-4183 where they are being used in the program. This tied back model (or function) from labeled training data. We paid to the lectures covering the concept of modules in Python. special attention to the experimental setup and evaluation Other questions successively led the students from the of ML, explaining the use of training, validation, and test lowest level functions toward the more complex ones. sets, as well as the use of k-fold cross-validation.36 We Eventually, the students had to understand the whole also covered several evaluation metrics (e.g., accuracy, program in order to complete the homework successfully. precision, recall, F1, AUC, and Jaccard similarity).37 We They were also asked to run the program with carefully explained the importance of more detailed inspection of selected input and to explain the results. Interestingly, the experimental results using a confusion matrix38 and the program learned its model from the textual corpus. via means of detailed error analysis. Hence, it provided a nice prequel for NLP and ML, topics With the Jupyter notebook, we demonstrated how one covered later in the course. turns a text into features (e.g., a bag-of-words39). This

4.1.3 INTRODUCING MACHINE LEARNING much AI and Law research and was frequently discussed is important as text classification plays a major role in Two lectures focused on introducing ML. The aim was throughout the subsequent lectures. The lecture concluded to explain selected ML concepts for practical use in the with a brief introduction of several ML models selected context of the course, not to provide a comprehensive based on their conceptual appeal (i.e., logistic regression, 40 introduction to ML. Over the remainder of the semester, decision tree, k-nearest neighbors ). we planned to revisit and expand upon many aspects of The second lecture was a guided walkthrough of a prepared Jupyter notebook. The goal was to provide papers. The two lectures were a mix of a presentation students with fully functional code that performs an ML andML, asome guided in walkthroughthe context of of specific prepared AI Jupyter and Law notebooks. tasks or The idea was to leave the students with example code with evaluation of the results. As the data set, we used that they could use as the basis for the third homework statutoryexperiment interpretation starting from data data that preparation are publicly and available. finishing (more details below). (Savelka 2019) We chose to work with scikit-learn41 for the ML experiments as it enabled writing the code in the model and discussed training and optimization. After way that largely corresponded to the concepts covered The first lecture introduced the concept of an ML reinforcement, or transfer learning), we focused mostly students through the stages of (1) downloading the data in briefly introducing other paradigms (e.g., unsupervised, JSONin the format, previous (2) lecture. exploring The the first data section (3) and of preparing code walked the on supervised learning, that is, inferring a classification 36 k-fold cross validation is a standard procedure for evaluating an ML program in which the data are divided into k subsets or “folds.” In each of k rounds, a different one of the k subsets is reserved as the test set. The ML model is trained using the k - 1 subsets as the training set (Ashley 2017, p. 395). 37 Accuracy is the ratio of correct predictions over the number of all predictions. Precision is the ratio of the number of positive predictions that are correct over the total number of positive predictions. Recall is the ratio of positive predictions that are correct over the number of instances that were positive. F1 is the harmonic mean of precision and recall where both measures are treated as equally important. (Ashley 2017, pp. 393, 396, 400) AUC is an ML metric for evaluating a binary classifier; AUC relates to the probability that a classifier will rank a randomly chosen positive data point (e.g., relevant provision) higher than a randomly chosen negative one (non-relevant provision). (Ashley 2017, p. 393) Jaccard similarity measures the similarity between two sets as the ratio of the size of the intersection and the size of the union of the sets. DeepAI, Jaccard Index, https://deepai.org/machine-learning-glossary-and-terms/jaccard-index Accessed 16/12/2020 38 A confusion matrix is a table that contains information about a classifier’s predicted and actual classifications. (Kohavi and Provost 1998). 39 A bag-of-words is a representation of a document as a collection of its terms that ignores the sequential order of the terms in the document (Ashley 2017, p. 394). 40 Logistic regression is a statistical learning algorithm that predicts the odds of being an instance of a category based on the values of independent variables (predictors). It employs an iterative statistical procedure to estimate weights for the predictors. A decision tree is an ML technique that learns a tree-like set of questions or tests for determining if a new instance is a positive instance of a classifier. Each question is a test: for example, if the weight of a feature is less than a threshold value, branch one way, otherwise branch the other way. A k-nearest neighbor or k-NN algorithm compares a problem with cases to base a prediction on those that are most similar. One measures the similarity or dissimilarity between the features of the cases in terms of some metric. Then one predicts that a new case will have the same outcome as its closest neighbors. (Ashley 2017, pp. 108, 396, 398) 41 The scikit-learn library provides an open-source set of software tools for predictive data analysis. scikit-learn, https://scikit-learn.org/stable/index.html Accessed 7/7/2020

150 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 data to be used in the ML models (including the division 4.1.4 INTRODUCING NATURAL LANGUAGE into training, validation, and test sets). As the task was PROCESSING Two lectures were dedicated to introducing NLP. Again, (4) training decision tree, k-nearest neighbor, and logistic the goal was to explain selected concepts for practical use simple text classification, the second section focused on in the course context, not to provide a comprehensive evaluating the three models using accuracy, precision, treatment of NLP. We also planned to revisit and expand recall,regression F1, as models. well as Theconfusion final part matrices. was dedicated to (5) upon certain aspects of NLP over the remainder of the Three related required readings supported the two in-class lectures on ML. Two of them introduced facets papers. As in the case of the introduction to ML, the two of ML concerning choice of appropriate representations lecturessemester were in the organized context ofas specific a mix of AI presentation and Law tasks and or a and algorithms for learning and the potential for unsu- guided walk through prepared Jupyter notebooks. Simi- pervised learning, that is, automatically clustering data to larly, the idea was to leave the students with an example reveal patterns that may correspond to concepts (Halevy code that they could use as the basis of the fourth (last) et al. 2009; Domingos 2012). An early article about Lex homework (more details below). The NLP lectures sup- Machina illustrated applying machine learning to predict plemented the ML lessons by focusing on the initial stage outcomes of cases from texts quite independently of their where texts are transformed into features. substantive facts (Surdeanu et al. 2011). The goal of the required reading was to create a situation where the methods using Jupyter notebooks. We introduced NLP as students would need to apply the concepts discussed in aimingThe firstto develop lecture methodsinterleaved to examplessolve practical of applying problems NLP class in order to understand the articles. involving language (e.g., search, information extraction, Finally, the third homework assignment related to the summarization) and contrasted it with computational two lectures introducing ML. The set of exercises was in- linguistics, whose goal is to understand properties of hu- tended to reinforce students’ understanding of the basic man language. We then discussed several core analytical steps in carrying out a supervised ML experiment. The goal sub-tasks, such as tokenization, segmentation, grammatical was to help them practice each step of such an experiment, and syntactic parsing, or named entity recognition. After beginning with acquisition and preprocessing of data and a brief introduction of each such sub-task, we provided a practical demonstration with the spaCy NLP library.42 Practically, the students were supposed to start from the We concluded the lecture with a case study on sentence Jupyterfinishing notebook with the evaluationused in the of second the experimental intro to ML results.lecture boundary detection and its challenges in legal texts, spe- and adapt it so that they could solve the problems from 43 the homework assignment. We also included two optional The second lecture presented two more case studies exercises that went beyond what was provided in the cifically case law. notebook from the class. One focused on using a different entity recognition in legal texts using a mapping of a special- izedvia required legal ontology readings. onto The a general first was one. an 44example The second of named case other involved improving a model’s performance using study concerned automatic recognition of facts and legal gridclassifier search from to optimize those presented hyper-parameters. during the lecture. The principles from case law.45 Additionally, the students were also assigned to read a paper on an automatic analysis of jury verdicts which was not covered during the in-class lectures (Conrad and Al-Kofahi 2017). The second part of

42 SPaCY, Industrial Strength Natural Language Processing, https://spacy.io/ Accessed 6/7/2020 43 The case study was based on (Savelka et al. 2017). 44 An ontology is a general, formal specification of the objects in a domain, their relations and properties. The case study was based on (Cardellino et al. 2017). 45 The case study was based on (Shulayeva et al. 2017).

Law in Context, Vol 37, Issue 1, 2020 151 ISSN: 1839-4183 the lecture was a guided walk through a prepared Jupyter 4.1.5 INTRODUCING ARTIFICIAL INTELLIGENCE AND notebook. This walkthrough was like the one in the second LAW introduction to ML. The goal was again to provide the - students with fully functional code that performs an ML of legalA final practice goal of basedPart I was on theto convey readings, a basic two understand chapters of with evaluation of the results. The focus, however, was Artificialing of the Intelligence field of AI andand Legal Law Analyticsand its role (Ashley in the 2017). future51 muchexperiment more on starting text feature from datarepresentation. preparation Three and finishingdifferent The lecture introduced students to computational models approaches were covered (i.e., bag-of-words, static word of legal reasoning, including legal expert systems, their embeddings,46 and transfer learning with contextualized uses and limitations. We discussed applying logic program- word embeddings47). As the data set, we again used part ming to represent statutory rules as in a classic program of the statutory interpretation data that is publicly avail- where the researchers expressed the British Nationality able (Savelka 2019). Act (BNA) in the form of logical rules and effectively “ran” the statute as a computer program (Sergot et al. 1986). to the two lectures introducing NLP. The set of exercises We focused on some problems for formalizing legislation enabledThe fourth students and to final practice homework performing assignment the basic related steps with logic. These include semantic ambiguity and vague- of integrating NLP components into a supervised ML ex- ness, syntactic ambiguity caused by the absence in natural periment. The goal was to reinforce students’ awareness language statutes of parentheses demarking the scopes of all the steps, but especially of the choices concerning of logical connectors, and legal indeterminacy: even if feature representation. Practically, the students were sup- adversaries agree on the facts and applicable law, they posed to adapt the Jupyter notebook used in the second can still generate reasonable pro and con arguments, and introductory NLP lecture to solve the problems from the courts come to different conclusions (Berman and Hafner homework assignment. In order to complete the assign- 1986). This is important because classical logic cannot ment, students could use the scikit-learn library and the readily support proving a proposition and its opposite, same model as in the previous homework. Optionally, they something that contending legal arguers frequently do. could also train the model using static word embeddings, Finally, we introduced a theme of the book and of the such as GloVe48 course: the role of text analytics in cognitive computing. tune a transformer-based deep neural network language Cognitive computing involves designing collaborations model (Devlin et or al. word2vec 2018).49 The(Mikolov, example et al. code 2013), to perform or fine- between humans and computers in which each performs even the optional tasks was provided. This was intended the kinds of intelligent activities that they can do best. Le- to serve as a solid basis for the work the students carried gal text analytics, we hypothesize, can extract meaningful out in their course projects.50 information from legal texts that will enable computational models of legal argument to connect more directly with

46 Specifically, we used the GloVe embeddings described in (Pennington et al. 2014). 47 Specifically, we used fine-tuning of the BERT model described in (Devlin et al. 2018). 48 A static word embedding such as GloVe is a learned text representation that employs statistics about word co-occurrences across the whole text corpus to capture word contexts and meaning. “Static” indicates that the word embeddings are employed as a component of the machine learning model and not updated (Pennington et al. 2014). 49 A language model is a neural network that has been trained to predict the next word in a sequence of words. A transformer, a special kind of neural network architecture, has an attention mechanism; each layer of its network assigns more weight to some features of an input sentence than to others. It learns associations between words that might be relatively far away from each other in complex sentences. By making multiple, parallel connections between certain words while ignoring others, it develops a “treelike representation of sentences [which gives] transformers a powerful way to model contextual meaning….” (Pavlus 2019) 50 A course on teaching coding for lawyers at the University of Minnesota also combined lectures and “practical exercises based on Python and covered the fundamentals and practical exercises of machine learning and natural language processing techniques.” “[S]tudents complete a code-based guided tutorial of an end-to-end machine learning project to understand the steps involved in the creation of this technology.” (Contreras and McGrath 2020, pp. 325, 328-330) 51 Ch. 1, Introducing AI & Law and its Role in Future Legal Practice and Ch. 2, Modeling Statutory Reasoning.

152 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 the texts. This, in turn, will expand the capabilities of decision trees to model judicial decisions regarding bail, cognitive computing by addressing some of the limita- to predicting whether a U.S. Supreme Court Justice and the tions of machine learning for answering legal questions and explaining the answers (Ashley 2017, pp. 11-12). Katz, et al. (2017) applied extremely randomized forests ofwhole decision Court trees affirm to past or reverse decisions a lower represented court’s asjudgment. features 4.2 PART II: COMPUTATIONALLY MODELING CASE-BASED from the Supreme Court Database and information about LEGAL ARGUMENT AND IMPROVING LEGAL IR trends. In that reading students encountered another Building on this foundation, Part II provided a broader example of evaluating a machine learning program with framework by exploring AI and Law research on modeling case-based legal reasoning and argument and on The lecture and Chapter 4 also introduced work on stratified k-fold cross-validation as introduced in Part I. using the models to predict and explain legal outcomes. predicting case outcomes based on case-based arguments and taking value effects into account as a kind of theory explain its answers and predictions in terms attorneys construction.55 These knowledge-based prediction ap- canThis understand. is one way the field can assist machine learning to proaches can explain their predictions in terms of legal A lecture and readings52 introduced competing com- arguments. putational models of case-based reasoning in law – legal The reading, Chapter 5, Computational Models of 56 argument as theory construction, as a fortiori reasoning, Legal Argument, introduced students to a revolution- or as explanation mapping. Each one employs a different ary development as AI and Law researchers produced knowledge-based technique to represent legal cases: general models of legal argument into which models of in terms of prototypes and deformations, dimensions - and factors, or exemplar-based explanations. Each one tary modules. The new models, such as Carneades, use computes relevant similarity of cases in a different way, attackingrule-based (and/or and case-based supporting) reasoning arguments fit asinstead complemen of strict for example, as overlapping sets of factors or as ratios of logical inferences.57 Unlike relying solely on deductive criterial facts, and supports different kinds of counter- logic, the argument models accommodate two realities: examples.53 The competing case representations have legal reasoning is nonmonotonic and defeasible. That is, different implications for using text analytic techniques inferences change as information is added or becomes to connect the texts of legal decisions directly to computa- invalid (nonmonotonicity) and arguments contradict and tional models. So far, only the factor-based representation defeat each other (defeasibility). seems likely to be able to be extracted automatically from The lecture featured two examples of computational case texts. In addition, only the factor-based models have models of legal argument. Given a case as a list of fac- been adapted to account for teleological reasoning with tors, Grabmair’s Value Judgment-based Argumentative the values underlying the legal rules.54 Prediction (VJAP) program outputs a predicted outcome, Researchers in AI & Law have developed increasingly sophisticated knowledge-based methods for predicting justifying prediction. It uses rule-based, case-based, and outcomes of legal cases. The lecture surveyed these methods: value-baseda level of confidence reasoning in tothe argue prediction, if a decision and an is argument coherent from applying a nearest neighbor algorithm to tax cases with tradeoffs among value effects in prior cases and on capital gains represented with features, to applying bases its prediction on the arguments.58

52 (Ashley 2017) Ch. 3, Modeling Case-Based Legal Reasoning, Ch. 4, Models for Predicting Legal Outcomes. (Katz et al. 2017). 53 (Ashley 2017, Sec. 3.3) 54 (Ashley 2017, Sec. 3.4, 3.5) 55 (Ashley 2017, Sec. 3.5, 4.6) 56 (Ashley 2017) Ch. 5, Computational Models of Legal Argument 57 (Ashley 2017, Sec. 5.1 - 5.3) 58 (Ashley 2017, Sec. 4.6)

Law in Context, Vol 37, Issue 1, 2020 153 ISSN: 1839-4183

The second example is a model of evidence-based legal identify the topic and type of a statutory provision, for arguments about compliance with legal rules, the Default Logic Framework (DLF). As described below, it inspired knowledge engineered rules (or regex expressions) can the LUIMA type system featured in the next readings.59 A thenexample, extract whether relevant it is ainformation definition or from imposes these an provisions, obligation; type system is a kind of ontology for annotating training such as the obligor and obligee. The ML approach requires sets of texts in terms of a hierarchy of concepts and rela- human experts to annotate training sets for learning and tions so that machine learning in an annotation pipeline gold standard sets for evaluation, a labor-intensive and can automatically assign meanings to regions of text. The costly activity. Some research projects illustrated the utility LUIMA type system captures the argumentation roles of applying knowledge-engineered rules and templates to that sentences play in judicial decisions. These include partially annotate texts in support of human annotation. a LegalRuleSentence, which states an abstract legal rule In principle, this could make human annotators’ work without applying it to a particular case, or an Evidence- as well as experts. Indeed, students could successfully on whether or not evidence in a particular case proves that annotatemore efficient statutes and andenable cases non-expert in a crowdsourced humans to annotateactivity. BasedFindingSentence, which reports a factfinder’s finding A lecture pointed out that despite these ML techniques, we hypothesize that state-of-the-art legal information automatically extracting logical formulations of rules, a rule condition has been satisfied. As the lecture explains, retrieval systems could be more effective to the extent instead of manually constructing them as in the BNA that they consider argument-related information such program, still seems technically out of reach. Combin- as sentence roles, a point elaborated in Part III. ing text extraction and network diagrams, however, can help to partially automate the comparison of similarly 4.3 PART III. APPLYING TEXT ANALYTICS TO LEGAL TEXTS purposed regulations such as different states’ public Part III’s readings and lectures focused on applying health emergency statutes. This is of practical import, machine learning to three types of legal documents: since commercial organizations, for example, insurance, electronically stored documents produced in e-discovery, health care, computer security, and privacy regulation, statutes, and the texts of legal decisions.60 must comply with multiple states’ similar, but somewhat different, regulatory schemes.

4.3.1 APPLYING ML TO DOCUMENTS IN EDISCOVERY AND STATUTES 4.3.2 APPLYING ML TO LEGAL DECISIONS After recapitulating the machine learning methods, Three articles reported on successive experiments metrics and evaluation techniques introduced in Part I, predicting case outcomes based on their texts in a man- a lecture addressed their application in the e-discovery ner that accounted for some information about their domain that was introduced in the Dynamo Holdings case. substantive topics and the parts of the decision reporting Among other things, it featured a project applying network the facts: (Aletras et al. 2016; Chalkidis et al. 2019; Med- analysis of emails that could guide selecting witnesses vedeva et al. 2020). These papers support the feasibility for deposition and the possibility of constructing such of automatically analyzing legal texts and building suc- a network automatically from the sender and receiver cessful predictive models of judicial outcomes. The data header information in emails produced in the litigation. comprised decisions of the European Court of Human Another segment focused on applying machine learn- Rights. Given textual evidence extracted from a case, the ing and rule-based techniques to extract information task is to predict the main conclusion of the court, that is from statutory or regulatory texts. Machine learning can whether there was a violation or not of any article of the European Convention on Human Rights or, alternatively,

59 (Walker 2006) Ch. 6, Representing Legal Concepts in Ontologies, Ch. 7, Making Legal Information Retrieval Smarter and Type Systems. (Ashley and Walker 2013). 60 (Ashley 2017) Ch. 9, Extracting Information from Statutory and Regulatory Texts.

154 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 a violation of a particular article. The results support a decide if a case retrieved by an IR system is worth read- text-based approach for making ex ante predictions of case outcomes. Anonymizing information in the texts UK and Ireland dealing with civil matters, the researchers such as the name of the state involved in the case did not demonstrateding. Working with the fifty feasibility common of humanslaw case reliably reports annotatfrom the- much affect the prediction. On the other hand, predicting ing applications in cited cases of legal principles to facts outcomes simply from the names of the judges achieved and of an ML system learning to perform such annotation. lower rates of success, but not much lower. The third paper also introduced students to the application of new information, we returned to the LUIMA project of Part II language models for text-based prediction such as BERT andAs its a type final system example for ofannotating extracting case argument-related texts in terms of (Bidirectional Encoder Representations from Transformers) the roles sentences play. An experiment described in the and attention-based neural networks that can highlight readings,62 pitted LUIMA against a commercial legal IR the most predictive portions of the text. system. LUIMA’s search module employed cases annotated We supported the lectures on the court outcome in terms of certain sentence argument roles and retrieved prediction with a Google Colab notebook. As the basis them based on the annotations. LUIMA’s re-rank module we used a paper by Medvedeva et al. (2020) and their learns weights from a human expert’s “true” rankings of publicly available dataset.61 cases in the training set that are responsive to queries; students through the data exploration and preparation it uses these weights to re-rank responsive cases in the phases. Then, they could go through The notebook the same experimentsfirst led the test set. The LUIMA evaluation showed that employing as those reported in the paper. Finally, students were argument-related information in indexing and querying encouraged to experiment with the code and try their along with the re-ranking weights outperformed retrieval own experiments. This exercise was important because with a commercial IR system that did not use such infor- it introduced the students to the idea of reproducing the mation, at least for a restricted set of documents in the results from a paper and what such an effort requires. domain of vaccine injury claims. We then revisited the topic of extracting information By annotating sentence role types such as Evidence- from the texts of cases, this time using more knowledge- BasedFindingSentence, an argument retrieval system intensive techniques. Predicting case outcomes by them- cases supporting a legal proposition in the abstract, but explain its reasons. For explaining and for improving legal examplescould help of users applying find a “application proposition tocases”, concrete that facts is, not which just informationselves is not retrieval, sufficient, we especially maintain, ifone a program needs to cannotextract may be analogous to the user’s problem. We considered argument-related information from case texts. In Part II, how LUIMA’s semantic annotation and argument-based we spoke of factors and sentence roles in cases as a kind reranking could be applied externally to a commercial of argument-related information useful for modeling legal system’s output for a user query, potentially improving reasoning. Here we considered the extent to which those the re-ranking, which would be a nice example of cogni- features can be extracted automatically through machine tive computing. learning and related techniques. The class reviewed the The class examined various approaches to automati- classic argument mining work of Mochales and Moens cally segmenting case decisions into argument-related (2011) in extracting argumentative propositions and classifying them as premises or conclusions. In one of of rhetorical roles of sentences such as establishing facts, the readings, Shulayeva et al. (2017) annotated “cited casefunctional history, parts. arguments Two readings and analysis,focused on the the decision identification ratio, facts and principles in legal judgements”. That is the kind of information that could help attorney/users quickly Bhattacharya et al. 2019). The outcome is a partitioning and the final decision (Saravanan and Ravindran 2010;

61 https://ndownloader.figshare.com/files/8901952?private_link=6f7d9e7c375ff0822564 62 (Ashley 2017) Ch. 11, Conceptual Legal Information Retrieval for Cognitive Computing. (Grabmair et al. 2015).

Law in Context, Vol 37, Issue 1, 2020 155 ISSN: 1839-4183 of a text at the sentence level into several “buckets,” in- valuable. Indeed, data mining of consumer informa- formation that could be used to inform a summary of the tion is big business. Data mining discovers patterns in case. Another reading explored automatically segmenting distributed information regardless of how the data are court opinions into high-level functional parts such as formatted or of legal and ethical constraints on fairness. Introduction, Background, and Analysis as well as issue- We then surveyed the generally limited extent to which American law under the U.S. Constitutional, federal, and The results for functional parts were not very far from state regulatory systems protects against these threats. humanspecific partsperformance; with Conclusions the results (Savelka for Conclusions and Ashley 2018). were We contrasted these with the protections under Europe’s less successful. The lectures on case law segmentation General Data Protection Regulation (“GDPR”) and the new were supported with a Google Colab notebook based on California Consumer Privacy Act (CCPA). (Bhattacharya et al. 2019) employing the researchers’ Machine learning and big data can be subject to bias, published code for their experiments. and correcting for a lack of fairness has been a perennial Commercial IR providers, of course, are constantly concern. The ProPublica article addressed fairness in enhancing their own systems for cognitive computing. machine learning and the problems of disparate impacts Students read a paper describing Lexis Answers, a recent of ML in bail recidivism prediction, including the investi- addition to the Lexis Advance legal research platform gation of racial bias in the COMPAS re-offense prediction (Bennett et al. 2017). The paper presents the challenges program (Angwin et al. 2016). that the system developers are addressing, for example, The lecture raised the question of the extent to which the need to identify irrelevant text, such as text quoted fairness can be achieved given practical and theoretical in the documents, the need to disambiguate terms by constraints. For example, if a model is unfair, one possibility context, for example, distinguishing meanings of “fraud” is to blind it to that attribute. Indeed, a common safeguard in criminal and civil law, and the problem of creating an in privacy regulation is to avoid using “prohibited” vari- ontology and type system to support text understanding ables. A simple test for unequal treatment is to change the of Lexis users’ queries. For example, is a user seeking the group attribute for a data point, and check if the predic- “standard of review in an appellate court” for a family of tion changes. Trying to achieve fairness by blinding the model to certain problematic features, however, does not of how to evaluate the system’s utility and effectiveness. always work. Apparently “harmless” variables can serve claims or for a specific claim? Finally, there is the question as proxies for group membership since they may not be 4.4 PART IV: AI IN LEGAL DOMAINS statistically independent of the group membership. Imbal- In Part IV, the remaining readings and last lectures in ances in the dataset may still lead to different treatment of groups. In addition, one needs to use some prohibited including AI and patent law,63 legal regulation of data privacy variables such as Group membership information in order the(O’Connor course 2018)focused and on machinemore specific learning legal bias domain (Angwin topics et to test for equal treatment. Moreover, removing predictive al. 2016), as well as legal liability of autonomous vehicles information from the model may lead to lower accuracy. (AVs) (Reed et al. 2016; Federal Automated Vehicles Policy Ensuring fairness, then, can be complicated. One can 2016; Nooteboom 2017). employ an optimized equation, one that tries to ensure The class considered the effects of electronic data min- that fairness constraints are met. Optimizing equations ing on privacy. Today, data collected can be aggregated to for fairness, however, is still subject to a problem: there are multiple conceptions of fairness, such as statistical/ what they do and say on the Internet. The information demographic parity, accuracy parity, equalized odds, usersproduce generate telling profilesas a by-product of who users of this are, activity as revealed is quite by equal opportunity, and predictive parity / conditional

63 A guest lecturer presented a talk entitled “NLP Customized to Fields of Law: Promises and Challenges” in which he focused on using AI to analyze patents’ level of “indefiniteness” flagging a potential statutory violation.

156 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 use accuracy. Each fairness criterion has something to component (e.g., computer science or computer engineer- recommend it, but they can be mutually exclusive; one ing). This increased the likelihood that each team had the may not be able to achieve any two of them at the same necessary knowledge and skills to complete a success- time (see section 6.) ful project. Furthermore, it ensured that students with The readings and lecture on autonomous vehicle (AV) liability addressed the questions of how the American with each other in a meaningful and productive way. As different backgrounds had to find ways how to interact law of product liability and negligence will apply to the this is a very important professional skill, we consid- manufacturers of vehicles involved in fatal accidents. The Although it was largely up to the students to design the ered it to be one of the projects’ most beneficial aspects. safety overall, inevitably, they will still cause accidents projects, there were certain fundamental requirements. injuringlecturer peoplenoted that, and evenproperty, though and AVs victims will increasewill sue manu traffic- Most importantly, each project had to have an empirical facturers of AVs and AV software systems. Presumably, component such as surveying a dataset or evaluating a the legal standards affect the ways in which automated ML model. An exercise of reading literature and writing vehicles should be designed given certain ethical responsibilities. For example, any drivers, human or automated, an essay would not suffice. must be prepared to deal with children on the street and 5.1 SUGGESTED PROJECT IDEAS AND DATASETS (FIRST other so-called “outlier” scenarios. These standards also PROJECT MEETING) give rise to legal and practical issues that affect machine Based on the experience from the 2019 run of the learning, for instance, in determining whether negligence course, we made sure that the students had an oppor- or product liability applies and in proving causation. tunity to start working on the projects very early in the

5. FINAL PROJECTS in the second week of the course. During this session we semester. For this reason, we held the first project session - explained the basic project requirements and milestones. First, we introduced several selected projects from the 64 grade.The Thefinal goal projects of the were projects one of was the tomost provide important students ele last year. Second, we presented the three datasets withments an of opportunitythe course, accounting to apply what for 50% they of learneda student’s during final 65 the course in the context of a larger work that they were These included the Supreme Court Database (SCDB) that we had identified as a suitable basis66 for a project. partially responsible for designing. The projects were a with full texts of the Court’s opinions, Veteran Claims 67 team effort. The only requirement was that a team consist Decisions comprising disability-claim decisions of the of at least one law student and at least one student pur- Board of Veterans' Appeals ("BVA") of the U.S. Department of Veterans Affairs, and the Enron Email Dataset with

64suing In the springa technical 2019 version degree of the course,with athe significant students completed programming five final projects: 1. Web Privacy Policy Analysis (Goal: To inform the public about privacy policies, apply supervised ML and NLP to a corpus of privacy policies to determine the kinds of data collected, by whom and if certain practices may be performed.) 2. Whose Line is it Anyway? (Goal: To inform readers about bias, apply supervised ML to dissenting opinions of Justices to classify sentences in legal opinions as taking a conservative or liberal stance.) 3. Risk Mitigation Strategy for International Patent Litigation (Goal: To inform potential litigants in international patent suits about their prospects in domestic U.S. courts, use ML to predict the results of patent litigation involving international parties based on database of U.S. patent litigation cases.) 4. Discover Legal Education Predictors of Bar Exam Passage and Failure (Goal: To identify law students who may need extra preparation, use data analytics and ML to identify early predictors for failing the bar exam and how those predictors change as students’ legal education progresses.) 5. Tax Court Analysis (Goal: To assist tax professionals to advise clients on when to seek settlement based on assessed deficiencies, apply ML to a dataset of tax cases to determine the deficiencies assessed against individual taxpayers who lose their cases against the IRS and the most commonly cited sections of the Internal Revenue Code.) 65 The Supreme Court Database, http://scdb.wustl.edu/ Accessed 6/7/2020 66 Caselaw Access Project, https://case.law/ Accessed 6/7/2020 67 LLTLab VetClaims - JSON https://github.com/LLTLab/VetClaims-JSON Accessed 6/7/2020

Law in Context, Vol 37, Issue 1, 2020 157 ISSN: 1839-4183 more than 600,000 employee emails from the period just or limitation in their chosen legal domain or dataset that before the energy company's collapse in December 2001 - due to a mammoth accounting fraud. The Federal Energy posed solution. The focus was on demonstrating proof of Regulatory Commission generated the corpus as part of concept;would benefit the template from a data-driven asked the students solution andto explain their pro the its fraud investigation.68 incremental progress that the project would make towards This was a considerable departure from the previous the envisioned solution, either by creating and statistically year when we assigned students the task of identifying a exploring a new dataset to reveal interesting patterns or proper dataset as part of pitching a project and forming by applying analytical/ML models to an existing dataset a team. We had reasoned that it would be educational toward realizing some goal. To inspire project ideas, the for law students to identify domains and potential data instructors suggested reading a recent paper (Branting sets as part of the exercise. While it did lead to success- 2017) that surveyed some practical legal application areas. ful projects, we found that, at about the halfway point in the course, some students had not yet obtained enough 5.2 PROJECT PROPOSALS (SECOND PROJECT MEETING) well-informed intuitions to assess the suitability of data The goal of the second project session was for the for projects. In the end, organizing project pitches and teams to present their project proposals and have them supporting teams in their formation and technical setup discussed. Eventually, the students formed three teams: required considerably more time and effort than we had two teams with three members and one team with four. expected, and it drew too much attention away from The one smaller team was also joined by the Duquesne other tasks. University Law School professor who was auditing the lectures. Each group was allocated 20 minutes for the (via short code snippets in Google Colab notebooks), we presentation and 10 minutes for a discussion. Interest- suggestedAfter briefly several introducing project ideas the three for each prescribed of them. datasets These ingly, none of the teams opted for one of the proposed were very high-level ideas that the students could adopt project ideas and, hence, all the teams proposed their directly and develop into full proposals or use as an in- own projects. One of the teams even opted for using a spiration for their own ideas. For example, for the Enron different dataset from the three suggested ones. This Email Dataset we suggested identifying which emails are shows that students were motivated to put extra effort purely personal and express worry/anxiety, sadness/ into the projects. despair, or anger/agitation. For the SCDB, our suggestion focused on an area of criminal procedure: the constitu- tionalityThe first of ateam search proposed for and a seizure text classification of evidence project under was to predict if a justice will vote for or against affirming of the decision itself. Finally, an example of a suggestion the Fourth Amendment. They proposed to automatically a new decision involving issues identified from the text for the Veteran Claims Decisions was to automatically identify language in case decisions suggesting that a court identify the role a sentence plays in a legal decision and had applied a “totality of circumstances” rule versus a determine if it is helpful to understand the sentences as “bright line” rule in determining the validity of the search. tightly dependent ordered sequences or rather as inde- Here, the team had a clear idea of the work they wanted pendent units. to carry out including the technology they would use (a After the session, the students were expected to form transformer-based language model such as BERT). The teams and start working toward project proposals. We team also committed to assembling their own data set provided a template for proposing a project. It asked including annotating the data. students to describe the legal domain problem the stu- Our initial feedback on the proposal was to take extra dents proposed to tackle, including the process, task, or care in the data collection step. We suggested that the team use case. Students should identify the current problem

68 UC Berkeley Enron Email Analysis Project, http://bailando.sims.berkeley.edu/enron_email.htmlwork on Accessedexplicitly 6/7/2020 defining the type system. We provided

158 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 several examples of annotation guidelines and prompted the students to think about how to deal with the situation and for each we suggested the steps that would be involved. where a single decision contains language associated general ideas. We identified the two most promising ones with both (or different) types of rules. We advised the on an analysis of a disagreement of the Supreme Court students to use explicit, reasonable criteria to guide the withThe the first court one of would lower have instance. resulted The in analysis a project would focused be selection of legal decisions to include in their data set. We emphasized the importance of demonstrating that human (disagreement in terms of different aspects) or the annotators could agree on the task of labeling instances strengthdriven by of either a disagreement a faceted definitionon an ordinal of a scale disagreement (e.g., very of the types. Demonstrating such interrater reliability strong, strong, medium, mild). In carrying out the project (IRR) is a crucial component in assessing the system. Hence, we recommended that two different people look - at a certain number of the same decisions independently the students would likely have to filter the SCDB dataset in order to measure IRR by computing a kappa/alpha ato non-negligible cover specific agencies number andof cases issues, from come the up data with set a de in statistic. Finally, we asked the students to think about a termstailed definitionof the disagreement, of the disagreement, train an ML manually system usingannotate the benchmark against which they could evaluate the system, such as a random or majority class baseline. of a disagreement, and evaluate the system. textual features of a case to figure out the strength/type The second team wanted to investigate the concept of The second proposal was a project focused on an a swing vote in Supreme Court decisions. This phenom- analysis of the effects of a veteran being represented by enon relates to the justice who typically breaks the tie in a an attorney before the Board of Veterans’ Appeals. In decision that otherwise would involve an even number of carrying out the project the students would likely have to votes for and against, (i.e., conservative vs. liberal). Such a justice plays a vital role in decisions that would otherwise be deadlocked, often involving a clash of political ideolo- filter the BVA dataset we would provide them, explicitly- gies. The goal was to use empirical methods to predict allydefine annotate what it a means non-negligible for a veteran number to be of represented cases in terms by 69 which justice would likely play such a role in each case. ofan theattorney, attorney define representation the possible typesand outcome, of outcomes, train manu a ML The team would use the SCDB to carry out the project. system using the textual features of a case to determine Although, the proposal left certain aspects of the pro- the outcome and if an attorney represented the veteran, ject open we deemed that the students were quite close evaluate the model, apply the system to the full dataset, and analyze the effects of attorney representation of the give the effort more structure we provided the students veteran. The team chose to work on the second proposal, withto defining a list of the suggested scope of steps. a successful First, we project.asked them In order, to come to primarily because of its practical importance to veterans.

- 5.3 WORK ON THE PROJECTS (BI-WEEKLY STAND-UPS) tionup with we advisedan explicit the definition students toof comewhat itup means with afor strategy a case to be decided by a swing vote. Second, given the defini After the meeting where the students proposed their for distinguishing these cases from the others. Third, we projects, they were supposed to work toward their completion. In order to make sure the students were making steady progress during the semester, we held bi-weekly, beproposed to come that up with they a explicitly strategy for define identifying what it such means a justice for a justice to be a swing vote in a case. The final step would 30-minute meetings with each group. Students began the meetings by explaining what they had done and The third team struggled to come up with a proposal based on the definition. their plans for the next steps. We usually commented on for a single project. Instead, they came up with several the progress and made sure the team remained on track

69 Studies suggest there is substantial variation in the identity of the median justice across areas of the law. (Lauderdale and Clark 2012)

Law in Context, Vol 37, Issue 1, 2020 159 ISSN: 1839-4183 toward successfully concluding the project. Interestingly, points that we believed were important to be addressed each group required a very different approach. in the reports. The group working on automatically identifying the type of rule, “totality of circumstances” versus “bright line,” a court applied in deciding the constitutionality of limitedThe firstdataset group of shortsucceeded excerpts in delivering from US Supreme an exceptional Court a search turned out to be self-reliant. Even though the Fourthproject Amendmentexactly as proposed. cases retrieved Specifically, from they WestLaw. assembled They a group appeared to be able to complete the project unas- manually labeled the case texts as suggesting the use of a “totality of circumstances” rule or a “bright line” rule. outsisted, weaknesses we focused and on specificlimitations points of wheretheir work we could and help.help set they evaluated the performance of the ML models theSpecifically, students our think approach through with if and this how group to remedy was to them. point (decisionThe team tree,performed support two vector sets ofmachines, experiments. multi-layer In the perfirst- The second team that worked on investigating the ceptron) trained on simple representation of text features concept of a swing vote in Supreme Court decisions (bag-of-words). The best performance achieved by these struggled with how to include some non-trivial empirical models was an F1 of 0.72 (weighted; 0.68 for the bright line class and 0.74 for the totality of circumstances class), that their project appeared to be limited to a simple data a reasonably good combination of precision and recall. component as required. Specifically, we were concerned team focused on discussion of how to make the task more language model based on BERT (Liu et al. 2019). This In the second set they fine-tuned a transformer-based ambitious.filtering/analysis. Toward Hence, the end most of the of the semester, interaction the studentswith this approach turned out to be much more successful with F1 of 0.87 (0.88 and 0.86 respectively). justices would vote in individual cases. The project exceeded all expectations. From the view- responded by redefining the project as predicting how The third group, that worked on analyzing the effects point of fulling the course project requirement, there was of a veteran being represented by an attorney before the little to criticize. We suggested that the students consider BVA, plunged into a detailed inspection of the decision the possibility of continuing their good work with the goal texts looking for textual cues suggesting representation of presenting it at an AI and Law venue (either a confer- (or the lack of it). Our main effort with this group was to ence or a workshop). In order to do so, we suggested, help them see the whole picture and move beyond the initial step. The students responded well and brought - they would need to significantly extend the data set and the project to a successful conclusion. By the end of the ditionally, an inter-annotator agreement study should be define the annotation task with explicit guidelines. Ad semester, however, they had substantially scaled down their original plan. ableconducted to extend to confirm the system the to objectivity deal with ofall thethe taskdocuments’ and its text,level notof difficulty just the portionsfor humans. selected Ideally, by theyhumans. would also be 5.4 THE RESULTING PROJECTS (FINAL PRESENTATIONS AND PROJECT REPORTS) The students performed some of these improvements The students presented the results of their projects dur- during the summer of 2020, submitted an expanded version of their report, entitled “Transformers for Classifying for the presentation and 10 minutes for the discussion. It Fourth Amendment Elements and Factors Tests” to the wasing two up tofinal the project students sessions. to decide Each how group to deliver had 20 the minutes pres- Jurix 2020 conference, one of the two main conferences entation. Although there was no requirement that all the in AI and Law, which accepted it as a full paper (Gretok, members of the group present, all three teams decided to split the presentation among all the members. The The second team completed a successful project, as well. et al. 2020)! presentations took place about a week before the project The team decided to abandon the original plan to predict reports were due. Hence, we focused the discussion on the a swing vote in Supreme Court decisions. Instead, they

160 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 conducted a very detailed analysis of the SCDB features as focused on fair machine learning, and (3) engaged in a to their power for predicting justices’ voting (i.e., liberal or multi-stage annotation and text analysis homework as- conservative) with the intention of training a multi-layer signment and in-class text analytics workshop. perceptron model on the subset of most predictive features. The 2020 course version afforded less time for these The team reported that features regarding issue, issue area, exercises because of the way in which we reorganized and lower court decision appear to be the most predictive of a justice’s vote. A multi-layer perceptron model using the most predictive features, however, did not outperform earlierthe process in the of semesterpreparing and for focusedfinal projects. the student As noted, teams we the baseline (majority class). As regards this project we onbegan selecting formulating among final three project legal text teams corpora, and topics70 which much we appreciated the students’ detailed feature analysis. It made available early in the semester. We tried to focus was important for them to learn that some features are all the technical assignments on programming tasks that substantially more predictive than others. Ideally, the students would have detected the limited predictiveness (as opposed to solely the SCDB or other nontextual data). of some of the features earlier in the semester in time to Thus,would we support revised final the projects course curriculuminvolving legal to introduce text analytics stu- steer their project toward a more productive outcome. dents earlier to programming with Python and involving The last group struggled the most but completed the natural language processing and machine learning. This project successfully, as well. The original plan was to deter- - mine from a case decision if a veteran was represented by cient and contributed to their quality, but at the expense an attorney before the BVA and then correlate the variable madeof leaving the processout technical of organizing exercises final with projects non-textual more legal effi with outcomes. In the end, however, the team was not able data such as the SCDB and bail recidivism prediction to carry out the second step of the plan. We appreciated data. In addition, we devoted more lecture time to plac- how the students delved into the actual texts looking for ing legal text analytics within the context of research in patterns indicating if the veteran was represented or not. We believe this effort was a substantial educational exercise; and text analysis task. AI and Law, leaving insufficient time for the annotation it helped them to understand some of the rewards and Nevertheless, since these three spring 2019 activities challenges of working with legal texts at scale. The team seemed to be valuable pedagogical exercises, we include spent so much time and effort in identifying the project brief descriptions here. aim and in exploring, annotating, and preparing the data, 6.1 SCOTUS PREDICTION representation variable with the outcomes. Nevertheless, This in-class Python programming effort tied into a evenhowever, with that its limited they left scope, insufficient the work time was to acceptablecorrelate the as series of sessions devoted to predicting outcomes of U.S. Supreme Court cases. It was originally intended to include the final course project. a homework assignment, but the instructors decided to 6. COMPARISON WITH SPRING 2019 keep it in-class given the students’ mixed levels of technical Over the course of the spring 2019 semester, students engaged in three activities that were not included in spring with the dataset and models. expertise and the higher technical difficulty in working 2020. They: (1) participated in a series of exercises using As noted, in some highly publicized work, Katz, et al. the Supreme Court Database (SCDB) and working in class (2017, p. 2) applied machine learning to predict out- with prediction code based on that developed by Katz, comes of SCOTUS decisions and the votes of individual et al. (2017), (2) performed a programming assignment justices. They applied a decision-tree-based learning

70 As noted in section 5, these included the Enron Email Dataset, Supreme Court Data Base with the addition of full texts of the opinions, and the Veteran Claims Deci- sions. The three spring 2020 projects dealt with the latter two corpora, although one of the SCDB projects did not make use of the case texts. 71 The Supreme Court Database, http://scdb.wustl.edu/ Accessed 7/6/2020.

Law in Context, Vol 37, Issue 1, 2020 161 ISSN: 1839-4183 model (so-called extremely random trees) to cases in meaningful statements and structured queries in the the SCDB.71 The model correctly forecast 70% of case database format. This would then enable them to move outcomes and 71% of Justice-level vote outcomes over on to the more technically demanding part: selecting parts a sixty-year period. Cases are represented in terms of of the database and adapting the numerical encoding specially designed features from the dataset according toward training vote prediction models. to a codebook as well as aggregate features representing As the last increment, Grabmair conducted an in-class 72 “trends”. Textual features are not included. workshop on Supreme Court prediction demonstrating One learning objective was to give students some ini- code he had written to partially reproduce the results of tial practice in using Python code to query the Supreme Katz, et al. (2017) based in part on their code.75 Court Database for information about the justices and step was to select the parts of the database that should certain decisions. Some initial exercises for students to be used as features for the prediction. This is important The first perform included: Retrieve all votes where J. Scalia cast as the vote direction is coded in multiple variables of the a vote coded as liberal AND the whole Court's decision numerical vector (i.e., the row representing an individual was coded as conservative AND J. Scalia dissented AND vote) in the database. The second was to add to the dataset drafted a separate opinion. Interestingly, that really did some additional features that Katz, et al. had included in their model, such as the political parties of the president of the SCDB. Then students were asked to pick one of who had nominated each justice, circuit court information, thosehappen! cases, In fact, use there the citation were 64 to such retrieve cases the in ourfull releasetext of and aggregate trend information regarding the justice’s the opinion, read J. Scalia’s dissent and consider what and court’s probability of voting one way or the other. An information was missing from the dataset that would be important constraint here is to compute these aggregates necessary to capture the dissent’s reasoning. based only on the votes of prior years. This ensures that Some exercises in sensemaking about justices’ votes the model does not have access to temporally implausible focused on six cases for which students were asked to use information during training and testing. The third step is Python code to identify information such as the decision then to train random forest prediction models using the date, the basis for the Supreme Court’s jurisdiction, the scikit-learn Python library, similar to the original work by types of legal issues, the legal sources cited, the lower Katz, et al., and evaluate them. This includes comparing court’s outcome, the Supreme Court’s decision, and the the prediction accuracy of models over the years in the division of votes of the Court. dataset using different sets of features. One can also use For the programming, students needed to gain some built-in functions to assess each feature’s contribution facility with the numerically coded features of the SCDB,73 to the overall prediction. The trained model came very focusing on the justice-centered dataset at the issue close to the performance reported in Katz, et al.’s paper. level.74 To that end, the exercises allowed the students One interesting observation was that, while the aggregate to practice translating back and forth between legally trend features contributed most to the prediction of the

72 The case representation includes features such as case‑origin circuit, lower-court disposition, issue and issue area, the background of the justices and Court at that time, and historical trends. Background information includes justice, justice gender, Segal-Cover score, and party of appointing president. Trends include overall- historic Supreme Court, lower-court trends, current Supreme Court trends, individual Supreme Court justice, and differences in trends. The Segal-Cover score measures a justice’s “perceived qualifications and ideology” based on expert analysis of newspaper editorials prior to confirmation. The behavioral trends and trend differences are human-engineered features. They track “the ideological direction” of individual and overall justice voting behavior. Differences in these trends “include general and issue[-]specific differences between individual justices and the balance of the Court as well as ideological differences between the Supreme Court and lower courts.” (Katz, et al. 2017, pp. 6, 7, 14) 73 See the coding handbook here: http://scdb.wustl.edu/documentation.php?s=1 Accessed 27/11/2020. 74 The particular data of interest included, e.g., the case name, petitioner, respondent, petitioner’s state, case court of origin, lower court’s disposition of the case, the direction of the lower court’s disposition of the case, the issue and issue area, the direction of the Supreme Court’s decision, the authority cited for the decision, the type of law, the vote, and the direction of each justice’s vote. 75 MJBOMMAR / SCOTUS-PREDICT-V2, https://github.com/mjbommar/scotus-predict-v2/blob/master/src/legacy_model.py Accessed 7/7/2020

162 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 full model, removing them reduced performance only by was made publicly available (Larson et al. 2016) and a very small margin, and the model was able to leverage greatly facilitated technical research around algorithmic an almost as strong prediction signal from other features. fairness and its application in machine learning problems. As mentioned, the original plan was to assign some of For purposes of this assignment, the features available the SCOTUS prediction model training and evaluation in for prediction were limited to a subset of those originally a homework assignment. While the students were able employed,77 - to query the database and analyze the retrieved results, the instructors had the impression that the class would of fairness were and made the files available needed in to the complete posted slides the home from thework lecture were distributedon fairness to in the machine students. learning. Various definitions notebook was almost completely prepared and the actual The deliverables for this homework included the Python benefit more from a guided exercise. While the Python amount of code to be written was not substantial, we saw notebook with three completed tasks and a brief report that many students in mid-February 2019 had not yet gained enough ability to see the analytical method “be- included surveying the data, predicting recidivism with hind the code.” Additionally, in preparing the assignment, non-blindwith textual and observations race-blinded on models, specified and topics. training The atasks fair model. The data survey involved plotting histograms (us- gained from adding the non-standard features was not ing a Pandas plotting tool) of the COMPAS risk score data we observed that the benefit in prediction performance large enough to produce a satisfying experience suitable for the whole population and separately for Caucasians for a homework exercise. and African Americans. The task of predicting recidivism comprised two 6.2 FAIRNESS IN MACHINE LEARNING subtasks: creating a logistic regression model that is not This programming assignment involved training logis- race blind and one that is blinded and comparing the tic regression models on bail recidivism prediction data, fairness of their predictions on the whole population thereby focusing on a contemporary question of fairness and for Caucasians and African Americans separately. in machine learning, an issue and debate affecting many Students began by training a logistic regression model78 aspects of AI. The most plausible angle for this course was on all the data and assessing its accuracy. This was a the line of writing, work, and research initiated by Pro- Publica’s well-known 2016 survey (Angwin, et al. 2016) model separately for Caucasians and African Americans on the workings of the COMPAS system for predicting a andnon-blinded calculated race-specific the true negative model. and Then false they positive applied rates the defendant’s risk of committing another crime if released on bail. The inquiry found that the system’s prediction had across the groups, and considered which fairness criteria a higher false positive rate for African American defend- for each group, compared the fairness of the classifier ants than for Caucasian ones (i.e., it falsely scored them predictive parity. They then trained a “blind” recidivism as higher re-offense risks than Caucasians). A debate predictorwere satisfied: using demographic another logistic parity, regression equalized model odds, and or ensued between ProPublica and COMPAS’s developer Northpointe76 at the core of which was the question of how determine if the fairness changed. repeated the race-specific performance assessment to The third task involved using Microsoft’s Fairlearn is proprietary software, the dataset used by ProPublica library79 fairness can be defined quantitatively. While COMPAS itself to train a classification model whose behavior 76 A summary may be found in (Corbett-Davies et al. 2016). 77 That is, the defendant's gender, the age category of the defendant, the defendant's race, the number of juvenile felonies committed by the defendant, the number of prior convictions for the defendant, the category of crime with which the defendant is charged (misdemeanor, felony, or other), the risk score COMPAS assigned to the defendant (low/medium/high), the defendant's risk score on a scale of one to ten, and whether the defendant reoffended within two years. 78 As noted, logistic regression is a statistical learning algorithm that predicts the odds of being an instance of a category based on the values of independent variables (predictors). It employs an iterative statistical procedure to estimate weights for the predictors. 79 The assignment used an early version of the "Fairlearn" library: https://github.com/fairlearn/fairlearn.

Law in Context, Vol 37, Issue 1, 2020 163 ISSN: 1839-4183

- described above. Staff attorneys would like to ask the sible reduction in accuracy. In the assigned exercise, the system about the facts of the case, how the evidence was satisfies a desired fairness constraint at the cost of a pos assessed legally, and the case outcome. The system should odds fairness policy, ensuring fairness to non-re-offending also be able to answer abstract questions about Veteran’s defendantsstudents forced as initially the classifier advocated to adhere by ProPublica. to the equalized They law from the text of available decisions. computed the model’s performance on the whole data set and on the Caucasian and Afro-American subsets and sentences in BVA decisions according to their rhetorical compared the results for fairness (i.e., demographic parity, rolesAs ina firstthe fullstep, opinion, such a andsystem this would requires need annotating to classify a equalized odds, and predictive parity) as well as accuracy training set of sentences, as positive and negative instances across the groups and the tradeoffs between fairness of those roles. The instructor (Grabmair) conducted the and accuracy. This advanced exercise was optional for class sessions as an annotation workshop in which the law students but mandatory for students from technical students tried out an annotation scheme, wrote instruc- backgrounds. While bail recidivism prediction is an intuitions to guide the annotation, and annotated the cases. tive and legally very relevant task, we found that some of The relevant annotation types (i.e., the type system) the students struggled with the mathematical nature of comprised ten roles that sentences play in legal decisions: the programming exercise. 81

6.3 TEXT ANNOTATION ACTIVITIES citation, legal rule, evidence-based finding, and others. The third activity included a text annotation homework assignment and text analysis programming exercise extending over a span of three weeks. Students annotated sentences in the full opinions of Board of Veteran’s Ap- peals (BVA) cases according to their legal rhetorical roles using an online annotation environment, Gloss, to classify the sentences.80 A sample use case motivated the annotation exercise. Students were asked to assume that the BVA seeks ways to search their decisions more effectively. It would like to build an Alexa-like question answering system that system should receive a question about a case, and answer FIGURE 3. Gloss annotation environment with legal rhetorical basedcan talk on about one or BVA more cases. sentences As a first from pilot the study, decision, the dialog like role types the luimasearch function of the legal reasoning homework

80 As noted above, annotation means marking up portions of text that are positive instances of a concept or “type” of interest, which markups can then be used in annotating higher-level, more composite types. 81 The ten types and accompanying short explanations based on the LLT Lab BVA Type Explanations (https://github.com/LLTLab/VetClaims) include: Citation: Reference legal authorities or other materials in standard notation. Legal-Rule: Statement of legal rule(s) in the abstract, without application to current case. Evidence: Statements of facts/evidence in the case, without legal assessment. Evidence-Based Finding: Statement of authoritative findings, conclusion or determination of whether evidence satisfies a legal rule/standard. Evidence-Based Reasoning: Statements of reasoning, based on the evidence, in making the findings of fact. Legal Policy: Statement of abstract legal policies, principles or objectives without application to case facts. Policy-Based Reasoning: Statement applying legal policies to decide legal issues in given case, e.g., to decide whether to adopt or reject a legal rule. Conclusion of Law: Statement of ruling or holding about legal outcome of in case. Procedure: Statements of procedural facts and formalities in case. Header: Surface annotations for section headers.

164 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

The students had access to the Gloss annotation tool, TFIDF scoring words on the training dataset as a whole a convenient web-based annotation interface developed and per sentence type, and the performance of the clas- by Jaromir Savelka. They employed it to annotate two BVA cases, in one of which relief was granted and in the confusion matrix for the training data. The students ex- other it was denied. As shown in Figure 3, a type system sifieramined on mis-predicted the training and sentences, test set. focusingFinally, they on the plotted TFIDF a bar on the left side shows the 10 annotation types each with its own highlighting color. The text of the decision is kind of error analysis can lead to greater insights about thefeatures learned via model the index and howof misclassified to improve it.sentences. This by marking it with the cursor, selecting a type from a Clearly, these activities in the spring 2019 course shown in the center field. Students annotate a sentence pop-up menu, whereupon the sentence is highlighted in are worthwhile and would expand students’ skills and the color corresponding to the type. In the subsequent knowledge in ways very relevant to the course. The chal- class session, the instructor and students discussed the extent of the inter-annotator agreement with respect a semester’s activities. to their annotations of the two cases. They compared lenge, quite simply, is finding the time to include them in 7. STUDENTS’ COMMENTS AND INSTRUCTORS for clarifying the annotation instructions or revising the LESSONS LEARNED typenotes system. about anyBy the difficulties end of the encountered activity, a total and of the 16 casesneed If the course was a learning experience for the students, had been annotated by the class. it is also an opportunity for the instructors to learn. In These formed the dataset that would be used to de- this section, we report some of the lessons the instructors learned from this second experience of teaching the students in examining the cases included in the collection, velop and test the classifier models. This process involved loading each case’s sentences, putting it through preproc- project reports and in the anonymous student evaluations. course, based in part on students’ feedback in their final examining its performance, and analyzing any errors. The 7.1 REACTIONS TO COURSE AND INSTRUCTORS featuresessing and representing “vectorizing” the it, sentence and finally texts training were its a classifier,weighted major words minus the stop words, common small words According to the teaching surveys, students regarded the spring 2020 course in a generally favorable way. Six instructor transformed both training and test data into of eight responders in this course of ten students agreed TFIDFthat are vectors. filtered82 out. The In last the steps second involved step, trainingstudents decision and the or strongly agreed with the statement, “I would recommend this course to other students.” Two were neutral particularly suitable for this type of exercise because they and none disagreed. On a scale of strongly disagree (1) to cantree be classifiers easily visualized, for the sentence and students labels. can Decision assemble trees a arese- strongly agree (5), with respect to the “instructor's overall ries of decisions the model makes in assigning a sentence teaching effectiveness” the median response of the eight a type label based on the words it contains. Throughout responders regarding Prof. Savelka was 4.13 (.99 standard the programming parts of the exercise, students used deviation) and of the eight responders regarding Prof. 83 functionality provided by the scikit-learn Python library. Ashley was 4.50 (.53 standard deviation). The instructor and students began by examining the Some student comments described the course as a highest TFIDF features of an example sentence, the highest “Fantastic course, I learned a lot”, “really well–done course

82 TFIDF weights the terms in each document by frequency relative to a document and to the corpus. The weights are related positively to the term frequency in the document (TF) and inversely to its frequency in the whole corpus (IDF). To the extent a word appears often in a document or rarely in the corpus, the metric increases. To the extent it appears rarely in the document or frequently in the corpus, the metric decreases. 83 For the spring 2019 version of the course, seven of ten responders in the course of fifteen agreed or strongly agreed with the statement, “I would recommend hist course to other students.” while two were neutral and one disagreed. On a scale of strongly disagree (1) to strongly agree (5), with respect to the “instructor's overall teaching effectiveness” the median response of the nine responders regarding Grabmair was 4.11 (.78 standard deviation) and of the ten responders regarding Ashley was 3.80 (.79 standard deviation).

Law in Context, Vol 37, Issue 1, 2020 165 ISSN: 1839-4183

breadth of the subject. Depth is particularly important commented that both instructors were “knowledgeable, in a topic like ML, and much of the actual meaning and with fascinating content,” and a “Fun course!” Students- understanding is lost by going too quickly over the sta- evant research on combining AI and Law”. On the other tistics and modeling aspect. Even very basic examples hand,passionate, a comment and active complained, in the field” “We spend and “presented too much time rel recognition would be very helpful I think.” computer science for law students, and law for computer using pytorch or tensorflow in google colab with image on introducing entry level knowledges on two field[s]: students….” 7.2 COVID-19 RESPONSE Unfortunately, no description of a course in spring different backgrounds. Understandably, comments ap- 2020 would be complete without mention of the response The student reactions to the course reflected their parently from law students expressed some frustration to the covid-19 crisis. As noted, beginning with the week with the technical aspects of the course. "Honestly, this is of March 16, the University of Pittsburgh shut down to not a class for most law students, perhaps pre–requ[isite] avoid the effects of the virus. All classes were cancelled knowledge[] on computer should be included in course that week to enable instructors to prepare for online in- 84 description before we choose course.” “I realize there struction in the following weeks. Upon resuming classes, may be no good solution to this problem, but I found all lectures in the course were delivered asynchronously. the learning curve to be a bit steep. While I was able to Prof. Ashley recorded his lectures via Panopto; Prof. understand the basics and I was able to complete the Savelka used Zoom. Each Wednesday, Prof. Savelka held homeworks I still struggle to grasp many of the more tech- a twenty-minute session with each team to review their nical concepts we discussed.” Another comment, perhaps from an engineering student observing the law students, day’s lecture into three or four pre-recorded sessions of stated, “Much of the code and engineering aspect of the progress on the final projects. Prof. Ashley divided each course I felt could be taught in a way that did not leave Students commented as follows86: “The remote lectures between fifteen and twenty minutes each. the law students as behind as they were. Frankly, I'm not were a little slow, I will confess to listening to them at 2x entirely convinced that any understood what they were 85 speed, but they were clear and thorough.” “I personally coding or what they were doing.” struggle learning remotely but the lessons were helpful Overall, comments apparently from engineering stu- and the best of a bad situation. My group work became dents were enthusiastic about the course, although they expressed some frustration with what they perceived as no doubt suffered.” very difficult once we switched to remote and the project missing technologies: “I deeply appreciate your instruction Following recommendations of the administration to and am glad that I took this course. Even as a non–law instructors new to pre-recording lectures, Prof. Ashley embedded a small number of discussion questions in the and the ML/DL textual tools that apply. I will take away student, it has benefited my understanding of the field recorded lectures and solicited students to submit brief questions and concepts beyond the scope of the class, written responses. This seemed like an effective way to gauge how well students were following the lectures. “it may be prudent to narrow the actual content down your field, or my own.” A commenter recommended that … to a few case studies and algorithms rather than the an engineering student complained about the additional After the first week of pre-recorded lectures, however,

84 In spring 2019, some student commenters complained that the instructors had made “too many assumptions of what our prior knowledge is.” “It appears like we needed to have recently taken classes in philosophy, statistics, and computing programming to understand much of what is discussed in this course.” Indeed, that is one of the interesting aspects of AI and Law, that it brings all of these disciplines and law together! 85 See Section 7.7 Planned Revisions re Teaching Programming Skills to Law Students. 86 The University-provided form for student comments included the following question: “What do you think the University should know about your experience as a student in the current remote learning situation?”

166 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 time required to answer the discussion questions.87 Since abstracts provided the instructors with an indication of the students were also writing the one-page abstracts how well students understood the material. Given that the of assigned readings, the instructors decided to forego readings provided examples of how text analytic programs including discussion questions. have been employed in experimentation, also a focus of Of course, the question remains how best to encourage the course, the readings and the abstracts continue to student engagement with pre-recorded lectures. Recently, seem very important. we learned that the continuing pandemic would require Regarding the materials on AI and Law, as noted, in teaching the spring 2021 version of the course entirely spring 2020, the instructors employed chapters from remotely. In teaching another course remotely, Ashley Ashley’s Artificial Intelligence and Legal Analytics (2017) has distributed discussion questions to students prior to as assigned readings for seven sessions.89 According to publishing the pre-recorded lecture. Then, in a synchro- the comments, Prof. Ashley helped students learn about nous Zoom discussion session scheduled for the regular “realistic possibilities of integrating machine learning in class time, the students and he discuss the answers to those questions. The total duration of the pre-recorded and “clear and thorough applications to the law in an ac- lecture and the Zoom discussion session is no longer than cessiblethe legal mannerfield and forhow those it can of [be] other useful majors in building who we[re] a case,” a the scheduled duration of the class. Students actively bit green on the subject.” Two comments said he did “an participated in these discussion sessions. excellent job introducing legal concepts to students who had no former experience with the law,” and was “very 7.3 REACTIONS TO READINGS AND ABSTRACTS knowledgeable about the history of the subject and … very The instructors noted with interest that that there forthcoming [about] the good and bad areas of research were fewer comments about the readings and one-page and how they were being addressed.” abstracts than in spring 2019. Only two comments men- The focus on historical developments in AI and Law did tioned the readings or abstracts in spring 2020. Prof. not please everyone, however. A commenter complained, Ashley “assigned really interesting articles to read which - always related to class lectures and got us thinking more cally, the modeling lectures were perhaps state of the art in “some of the lectures and content felt a bit dated. Specifi about the implications of technology as it applies to the the legal research, but frankly, engineering disciplines have legal sector.” “I also enjoyed doing the reading abstracts. moved past many of the modes of modeling described in It changed how I read the articles and I found it deepened my understanding of them.” The abstracts were more are rarely used anymore. While they are indeed helpful the course. Specifically, logic models and decision models controversial in spring 2019.88 for subject understanding, I felt that they were taught in a way that might suggest that they would be valid for instructors assigned fewer abstracts and more conspicu- research today, which may or may not be the case.” ouslyThis integrated change maytheir reflect responses that to in the the students’ 2020 course, abstracts the Of course, Prof. Ashley did intend to suggest that these in the lectures. We again found the students’ abstracts former approaches in AI and Law might still have relevance to be quite perceptive and very useful. Beside prepar- today. In fact, rule-based decision and process models are ing students to discuss the research papers in class, the a very relevant topic in the discussion around the role

87 An anonymous comment also noted, “There is a great tendency to overload students when they go remote, and the university should offer at least some high–level guidelines as to how to adjust the course load to compensate.” 88 In spring 2019 students expressed conflicting opinions about the one-page abstracts. One student found them “very helpful for us to learn the material effectively.” Other students complained that they were too frequently assigned and too much work. “The abstracts are far too much busy work for law students.” Another commenter recommended that the instructors give the students more feedback on the abstracts or make them optional for extra credit. In that way, it would free up time for code- based assignments. 89 In spring 2019 chapters from this book were assigned readings in one session and suggested readings in 3 other sessions. One student in spring 2019 complained that “the book is terribly dense and is not helpful for beginners.”

Law in Context, Vol 37, Issue 1, 2020 167 ISSN: 1839-4183 of modern technology in legal practice.90 Indeed, that is [referring to Python programming environments]” and a major theme of the chapters assigned from the book. - To the extent text analytics provide the means to extract dents in the course. As a quid pro quo, the law students information from legal texts that can enable AI and Law couldnoted teachthat it tech would students be beneficial “how to to read have a casemore within tech stu the computational models of legal reasoning to connect directly to cases and statutes, those models may provide a basis technical expertise could “present in class and … help for the kinds of explanations that supplement machine first two weeks.” Alternatively, a teaching assistant with learning’s answers and predictions still lack. notebook etc.”91 students facing technical difficulties by setting up Jupyter As described in Section 4, Part I, in spring 2020, we 7.4 REACTIONS TO PROGRAMMING INSTRUCTION AND continued to use Python programming environments and ASSIGNMENTS homework assignments to teach skills. The classroom ses- If there was a point of consensus in student comments sions were more focused on leading the students step-by- from spring 2019, it was the recommendation to increase step through programming exercises with a few corpora the amount of class time spent on learning programming. relevant to the projects. In this way, we expected that the For example, commenters declared that, “It is okay to make skills for much of the data pre-processing and feature a few classes a Python class; we need it,” or that, “it is not a bad idea for law students to take a few more lessons accomplished during the semester. Thus, students would about Python.” Students perceived learning programming betterengineering learn the associated skills and with knowledge the final associated projects, withcould data be one commenter declared that “the assignment is helpful weeks in the semester to model building, experimenta- foras the us springto learn 2019 Python,” course’s another most difficultnoted that, challenge. “The Python While tion,wrangling error andanalysis, feature and preparation improving results.and devote the final assignment is extremely hard for me.” “Without enough Regarding the instruction in computer programming, foundations [in Python programming], it is really hard commenters credited Prof. Savelka with helping students for us to apply our knowledge into real life practices.” to learn “basic coding skills that can be used in a larger Students in spring 2019 suggested ways to divide program to achieve helpful machine learning tools for the programming activities into smaller, more learnable legal work,” and doing “an excellent job teaching computer chunks, for example by spending one class session per science to students who had little to no background in the subject.” Others noted that he provided “really useful notebooks to understand what the code is doing, what and helpful colab notebooks that made it easy to follow theweek limitations “as a working are, and day how manipulating it can be changed.” pre–filled Another Python along and understand each step” and “thorough lectures declared, “To learn computer code, one must experience it.” on techniques with example code in accessible forms via Small, digestible code–based assignments. Parsing excellent tools like Google Colab.” One comment stated that sentences. Counting words. Writing small computer he “made learning it fun, as fun as a wall of code can be.” functions. Watching [the instructor] do the same. Other comments suggested ways to improve the pro- These things all contribute to the "experienced based" gramming instruction. “Some more documentation to learning model. the provided code would be helpful in some cases. Some One student recommended “screening law school class provided code, though mostly from external resources participants earlier to determine the level of python exper- …, was not written for easy understanding by novices.” tise in the room.” Others suggested “having a tech student “Less condensed rewrites with more comments could teach the law student how to use Jupyter, anaconda, etc. be helpful for those struggling with the coding portion.”

90 See, for example, DECISIONS. AUTOMATED. https://bryter.io/ Accessed 7/ 28/2020 91 From the instructors’ viewpoint in spring 2019, the main problem was that the law students did not spend sufficient time on homework assignments. It seemed that the law students simply underestimated that amount of time that learning to program requires, a reality with which tech students per force are familiar.

168 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

“Code examples displayed on the screen were sometimes student projects now that there are some. This may give a good idea of what sort of questions could be explored examples with colab notebooks with legal datasets.” by the time the project really gets going. difficult to see well.” One comment suggested doing “more

7.5 REACTIONS TO COURSE PROJECTS preparation to make the process of preparing data from A second challenge involved sufficient training in data The Project Reports submitted in 2019 included sev- projects easier and less prone to errors. eral lessons learned, some of which recommended how [A]ll projects will deal with data coming from some to improve the timing and organization of the process for handling a Unicode Decode Error.92 If there had been a file and one of the biggest problems of the project was - final projects given certain challenges. manipulating dictionaries, and displaying results, the cially to obtain and process appropriate data: “[T]ime day given to general project help, such as opening files, wasChief the biggestamong thesebottleneck was a with lack thisof sufficient project. time,With espemore As can be seen, the 2019 Project Report discussions entire course would have likely benefited. better apply our results to [a] larger scope.” “[E]nough of lessons learned focused on recommendations for im- timetime, to… weencode could a haveuseful modified dataset how can webe crucialselected to data make to proving the timing and process for identifying teams and progresses in research projects.” “A future version of this preparing for projects. By contrast, in the two of the three 2020 Project Re- for their projects earlier on.” ports that discussed them, the focus was on substantive would benefit from students thinking about getting data One team suggested that: lessons learned from the team’s project experiments A future version of this course might allow students and activities. We take this as an indication that as in- to brainstorm and select projects earlier in the semester. structors, we had learned from the lessons of 2019 and As [students] learned techniques in class, [they] could improved the process for choosing projects, teams, and directly apply them to [the] project closer to when [they] learn them and with opportunities for better the 2019 comments led the instructors in spring 2020 to data, removing some causes of frustration. Specifically, understanding of concepts throughout the class. While begin organizing the teams and projects in the beginning this may require the project ideas to come more from of the semester, to assemble a dedicated text corpus for the professors, this could help with achieving a more complete project. However, it would still be feasible for involving these areas. From our perspective the changes three topic areas, and to suggest ideas for final projects students to come up with ideas based on their interests that we introduced in 2020 had positive effects. Starting and have more guidance from the professors about the the projects early in the semester and providing datasets methods to create the project. alleviated a lot of problems encountered in 2019. Somewhat similarly, another team recommended: This is a positive change, but the trade-off is the fact If there was a way to introduce all of the topics with they have still not been exposed to most of the course some of their use cases early in the semester, and have content.that students Possibly are thisdefining could their be the projects reason at why a time two ofwhen the the project topics due…early [in the second month of the three groups needed to change the scope of their projects course] (with the understanding that they might evolve), toward the semester’s end. this would give students a focus sooner that would help them explore data sources to use with their projects, giv- There are other lessons to be learned from 2020’s ing more time for the actual programming and analysis. project activities. One project report comment made a Another way to introduce what kinds of projects could be recommendation for future versions of the course. The done earlier (in lieu of the topics) would be to present past Fourth Amendment team’s project report stated:

92 https://wiki.python.org/moin/UnicodeDecodeError

Law in Context, Vol 37, Issue 1, 2020 169 ISSN: 1839-4183

Assignments and in-class guidance prepared the 7.7 PLANNED REVISIONS RE TEACHING PROGRAMMING authors well for this project. Additional opportunities SKILLS TO LAW STUDENTS for hands-on learning and comparison of different NLP Third, there is a lingering concern on how best to teach techniques or various transformers would be a welcome skills of programming and machine learning to prepare addition. The simpletransformers library makes these law students to participate actively and successfully in complex systems much easier to train and integrate into projects. While we have instituted more step-by-step a solution. Still, the approach of added bonus sections to instruction in the in-class and homework programming assignments is appreciated to prevent students from be- exercises, we still need to engage the law students more ing overloaded on tasks that may not be entirely relevant actively in the programming instruction and in program- to their project work. ming aspects of the projects. In addition, an anonymous student comments: “… the We have observed that it is more challenging for law law students do not have a background in the machine students with little technical background to reach a produc- learning…. My group primarily rested on technological tive level of programming for the course than it is for tech skills of an undergrad and our limited skills which I learned students to learn enough about law, legal reasoning and practice. Indeed, we may have been too accommodating very hard and compounded internal group frustrations.” to engineering students, while inadvertently leaving some for the first time this semester. This made the project law students behind. It may be a question of rebalancing. 7.6 PLANNED REVISIONS RE COURSE PROJECTS One way to rebalance may be to design different We will try to mitigate these effects in spring 2021. homework for each group, with more incremental in- First, as mentioned earlier, two of the three teams troductory exercises for the non-technical students. had to change the project’s scope towards the end of the Another is to take advantage of the technical talent in semester. This is not necessarily a bad thing. We noticed, the classroom. We will try to engage the tech students however, that students invested a lot of time and effort (e.g., the engineering students in spring 2020) to more in activities that led to realizing that the path they had actively provide feedback and assistance to law students chosen was likely not leading to success. While we think about the programming exercises. If the tech students get it is important that students have ownership of their pro- into the habit of helping the law students with respect to jects, in 2021 we would like to help them come to such a programming exercises, they may more naturally engage realization much earlier. Ideally, we hope to ensure that the bulk of time and effort would be spent on the substance of projects. Conceivably, law students could do the same for techthe law students students on legal in programming exercises. aspects of the final more assistance and feedback during the early stages of In any event, a goal of the course is to give law stu- the actual final project. We plan to provide students with dents an intense programming experience so that they their projects before they have been exposed to much of can better understand how the programs work, how to the projectcourse content.to overcome the fact that they need to define evaluate them, and the methods of the technical person- Second, as noted, despite our effort to create well- nel with whom they will need to collaborate in the future. rounded teams we detected problems with one of the That goal is met even if the law students do not learn to program effectively. aboutgroups membership where a bit inmore the technicalteams ourselves proficiency instead would of simply have 8. CONCLUSIONS settingbeen very the beneficial. ground rules. In 2021, We will we havewill consider to develop deciding an effective way to assess students’ backgrounds with respect Applied Legal Analytics and AI will address the lessons We hope that these modifications in the course on these two versions of the course, we have observed law to their potential contributions toward the final projects. learned from the course’s first and second offerings. In

170 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 students engaging with computer code closely enough engineering design and ethics tackling issues raised by to gain an understanding of how the technologies work the new technologies in the areas of professional respon- and how to evaluate them. We have stepped law students sibility, reliance, and liability of autonomous systems, and through programming with Python and participating in cognitive sciences concerning such issues as explanation in machine learning and cognitive computing. It would be each of the projects, the teams implemented some Python interesting to include academics and professionals from code,teams designedwith technical empirical students evaluations, to work on ran final experiments, projects. In these diverse domains in course-related collaborations obtained results, and engaged in some error analysis to or guest lectures. see what the programs missed and why. They have learned While the course focused mainly on U.S. law and legal how new text analytic tools work in legal practice and practice and on common law legal systems, there were about their limitations and gained experience in commu- nicating and working with technical personnel in applying are European academics trained in civil law, and one is and adapting the technology, a phenomenon which, we aboutstrong tointernational take up a faculty influences. position Two in of Germany. the instructors About expect, will become increasingly central to legal practice. one third of the students who took either version of the The experience of teaching this course has convinced course comprised international graduate students; two international law students were enrolled in Pitt’s program and conduct experiments with legal texts and data, and on International & Comparative Law. Nearly half of the thatus that the law skills students and knowledge can benefit they from learn learning will place to design them readings in AI and Law research were international in at the fore of a new generation of law students in tackling origin. The substantive legal issues affecting data mining the impact of technology on the practice of law. and machine learning such as bias, privacy, intellectual In meeting the challenges of teaching this course, our property limitations on data, and product liability, span experience demonstrates the importance of transdisci- international boundaries, where different jurisdictions plinary collaboration at multiple levels including among take different approaches. Although still quite different, instructors and students from very different backgrounds. the civil law and common law systems exhibit some The instructors’ expertise differs but together they cover convergences and present complementary opportuni- research and teaching in law, AI and Law, computer sci- ties for learning. We look forward to exploring how the ence, experimental design, and computer programming transdisciplinary collaborations so important to this with ML and NLP. The students have diverse skills and course can be extended across international boundaries, backgrounds, coming from multiple schools and depart- for example, through remote guest lectures, now feasible ments including law, computing and information, business, in our increasingly virtual academic world. electrical and computer engineering, and undergraduate computer science at Pitt and CMU. In guiding and par- 9. REFERNECES 1. necessarily collaborated with those from other disciplines V. 2016. “Predicting judicial decisions of the European Court outsideticipating of intheir final domains project of teams, expertise. instructors We also and encouraged students ofAletras, Human N., Rights: Tsarapatsanis, A natural D., language Preoţiuc-Pietro, processing D. perspective.”and Lampos, collaboration beyond academia; inviting guest speakers PEERJ Computer Science 2: e93. https://peerj.com/articles/ on applications in legal practice. cs-93/?utm_source=mandiner&utm_medium=link&utm_ from law firms or legal service providers to focus the class 2. Angwin, J., Larson, J., Mattu, S., and Kirchner, L. 2016. “Ma- The subject matter of the course is also inherently campaign=mandiner_201912 Accessed 27/11/2020 chine Bias,” ProPublica, 23 May. https://www.propublica. transdisciplinary, ranging beyond law, AI and Law, and org/article/machine-bias-risk-assessments-in-criminal- computer science to include political science in statisti- sentencing Accessed 7/8/2019 cally analyzing judicial decisions, criminology and sociology in addressing the effects of bias in machine learning,

Law in Context, Vol 37, Issue 1, 2020 171 ISSN: 1839-4183

3. Ashley, K. 2017. Artificial Intelligence and Legal Analytics. labeled biased against blacks. It’s actually not that clear.” New Tools for Law Practice in the Digital Age. Cambridge, The Washington Post, Monkey Cage. Oct. 17. https://www. UK: Cambridge University Press. washingtonpost.com/news/monkey-cage/wp/2016/10/17/ 4. Ashley, K. 2019. “Automatically Extracting Meaning from can-an-algorithm-be-racist-our-analysis-is-more-cautious- Legal Texts: Opportunities and Challenges.” Ga. St. U. L. Rev. 35: 1117-1151. 17. Council, J. 2019. “Top Law Schools Add AI Courses.” WSJ than-propublicas/?noredirect=on Accessed 7/7/2020. 5. Ashley, K. and Walker, V. 2013. “From Information Retrieval PRO Artificial Intelligence. https://www.wsj.com/articles/ (IR) to Argument Retrieval (AR) for Legal Cases: Report on top-law-schools-add-ai-courses-11555925401 Accessed a Baseline Study.” In K. Ashley (ed.), 26th Int’l Conf. on Legal 3/10/2019 Knowledge and Information Systems. Jurix-2013. Amsterdam: 18. Crichton, D. 2015. “With Judge Analytics, Ravel Law IOS Press pp. 29-38. Starts to Judge the Judges.” TechCrunch. April 16. https:// 6. Bennett, Z., Russell-Rose, T., and Farmer, K. 2017. “A scal- techcrunch.com/2015/04/16/who-judges-the-judges/ able approach to legal question answering.”, In Proceedings Accessed 26/11/2020 ICAIL-17. New York: ACM, pp. 269-270. 19. Dalton, B. “Cognifying Legal Education.” 2019. Above the 7. Berman, D. and Hafner, C. 1986. “Obstacles to the Develop- Law: Law 2020. https://abovethelaw.com/law2020/ ment of Logic-Based Models of Legal Reasoning.” In C. Walter cognifying-legal-education/ Accessed 2/10/2019 (ed.) Computer Power and Legal Language. Santa Barbara: 20. Devlin, J., Chang, M., Lee, K., and Toutanova, K. 2018. “Bert: Praeger. pp. 183-214. Pre-training of deep bidirectional transformers for lan- 8. Bhattacharya, P., Paul, S., Ghosh, K., Ghosh, S., and Wyner, A. guage understanding.” ARXIV Preprint arXiv:1810.04805. 21. Domingos, P. 2012. “A few useful things to know about Legal Judgments.” In M. Araszkiewicz and V. Rodríguez-Doncel machine learning.” Communications of the ACM 55: (10) (ed.),2019. “Identification32d Int’l Conf. of on Rhetorical Legal Knowledge Roles of Sentences and Information in Indian 78-87. Systems, Jurix-19. Amsterdam: IOS Press. pp. 3-12. 22. Eicks, J. 2012. “Educating Superior Legal Professionals: 9. Bishop, C. 2006. Pattern Recognition and Machine Learning. Successful Modern Curricula Join Law and Technology.” New York: Springer. In O. Goodenough and M. Lauritsen (eds.), Educating the 10. Branting, K. 2017. “Data-centric and logic-based models for Digital Lawyer 12-1: 5-1 – 5-14, https://www.academia. automated legal problem solving.” Artificial Intelligence and Law, 25 (1): 5-27. 8/1/2020 edu/9202158/Educating_the_Digital_Lawyer Accessed 11. Brostoff, T. and Sinsheimer, A. 2013. United States Legal 23. Federal Automated Vehicles Policy. 2016. Accelerating Language and Culture: An Introduction to the US Common the Next Revolution in Roadway Safety, NHTSA, US Dept. Law System. Oxford, UK: Oxford University Press. Transportation. (https://www.transportation.gov/AV/ 12. Cardellino, C., Teruel, M., Alemany, L., and Villata, S. 2017. federal-automated-vehicles-policy-september-2016) “A low-cost, high-coverage legal named entity recognizer, Accessed 26/11/2020, pp. 5-14, 17-19. Proceedings ICAIL-17. New York: 24. Fenwick, M., Kaal, W., and Vermeulen, E. 2018 “Legal ACM, pp. 9–18. Education in a Digital Age: Why 'Coding for Lawyers' Mat- classifier and linker.” In ters.” Lex Research Topics in Corporate Law & Economics 13. Chalkidis, I., Androutsopoulos, I., and Aletras, N. 2019. Neural Working Paper No, 2018-4, U of St. Thomas (Minnesota) Legal Judgement Prediction in English, Athens University of Economics and Business, https://arxiv.org/pdf/1906.02059 Legal Studies Research Paper No. 18-21. Pp. 0-31 SSRN: Accessed 26/11/2020 org/10.2139/ssrn.3227967 Accessed 26/11/2020 14. Conrad, J., and Al-Kofahi, K. Scenario analytics: Analyzing jury 25. Ferrucci,https://ssrn.com/abstract=3227967 D., Brown, E., Chu-Carroll, J., Fan, or http://dx.doi. J., Gondek, D., verdicts to evaluate legal case outcomes. 2017 In Proceedings Kalyanpur, A., Lally, A., Murdock, J., Nyberg, E., Prager, J., ICAIL-17. New York: ACM, pp. 29-37. Schlaefer, N. and Welty, C. 2010. “Building Watson: An 15. Contreras, A. and McGrath, J. 2020. “Law, Technology, and Overview of the DeepQA Project.” AI Magazine, Fall, 31: Pedagogy: Teaching Coding to Build a “Future-Proof” Law- 59-79. yer.” Minn. J.L. Sci. & Tech. 21: 2 297-332. 26. Grabmair, M., Ashley, K., Chen, R., Sureshkumar, P., Wang, 16. Corbett-Davies, S., Pierson, E., Feller, A., and Goel, S. 2016. “A C., Nyberg, E., and Walker, V. 2015. “Introducing LUIMA: computer program used for bail and sentencing decisions was

172 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

an experiment in legal conceptual retrieval of vaccine injury their compositionality. In Advances in Neural Information decisions using a UIMA type system and tools.” In Proceed- Processing Systems, 3111-3119. ings ICAIL-15. New York: ACM. pp. 69-78. 40. Colorado 27. Gretok, E., Langerman, D. and Oliver, W. 2020. “Transformers Law for Classifying Fourth Amendment Elements and Factors intelligence-and-lawMiller, S. 2019. “Artificial Accessed Intelligence 3/10/2019 and Law.” Tests.” 33d Int’l Conf. on Legal Knowledge and Information 41. Mochales,, https://www.colorado.edu/law/2019/05/03/artificial- R. and Moens, M. 2011. “Argumentation mining.” Systems, Jurix-2020. Amsterdam: IOS Press pp. 63-72. Artificial Intelligence and Law, 19(1), 1–22. 28. Halevy, A., Norvig, P., and Pereira, F. 2009. “The unreasonable 42. Murphy, M. and Pearce, R. “Algorithms, Ethics, and Legal effectiveness of data.” IEEE Intelligent Systems, 24 (2): 8-12. 29. Halterman. R. 2018 Fundamentals of Python Programming. and Professional Responsibility.” Unpublished manuscript Southern Adventist University (2018). https://archive.org/ Services: How Artificial Intelligence Will Disrupt Legal Ethics details/2018Fundamentals.ofPython. Accessed 27/11/2020. 43. Nooteboom, L. 2017. “Child-Friendly Autonomous Vehi- 30. Haselager, P. 2019. “Mediated action and the risk of entrap- cles;on file Designing with author. Autonomy with all road users in mind.” In ment”, invited speech at ICAIL-2019, Montreal, June 18. Developing Human Interactions with Autonomous Systems. 31. Hudgins, V. 2020. “Casetext Launches New Brief-Writing Nov. 13. https://medium.com/@HumanisingAutonomy/ Automation Platform Compose.” LegalTech News. Feb. 25. child-friendly-autonomous-vehicles-2880ca74165f Ac- https://www.law.com/legaltechnews/2020/02/25/casetext- cessed 29/3/2020. launches-new-brief-writing-automation-platform-compose/ 44. Norvig, P. 2007. “How to Write a Spelling Corrector.” https:// Accessed 26/11/2020 norvig.com/spell-correct.html Accessed 7/7/2020. 32. Katz, D., Bommarito, I., and Blackman, J. 2017. “A general 45. O’Connor, N. 2018. “Reforming the U.S. Approach to Data approach predicting the behavior of the Supreme Court of Protection and Privacy.” Council of Foreign Relations, Digi- the United States.” PLOS One. April 12. https://journals.plos. tal and Cyberspace Policy Program. https://www.cfr.org/ report/reforming-us-approach-data-protection Accessed Accessed 27/11/2020 26/11/2020 org/plosone/article?id=10.1371/journal.pone.0174698 33. Kohavi, R., and Provost, F. 1998. “Glossary of Terms.” Machine 46. O’Grady, J. 2018. “Suffolk Law School: Leading Transforma- Learning, 30 (2-3): 271–274. tion of Legal Education.” Practice Innovations 14. http:// 34. Larson, J., Mattu, S., Kirchner, L., and Angwin, J. 2016. “How static.legalsolutions.thomsonreuters.com/static/images/ We Analyzed the COMPAS Recidivism Algorithm.” ProPublica May 23. https://www.propublica.org/article/how-we-ana- Accessed 3/10/ 2019 lyzed-the-compas-recidivism-algorithm Accessed 7/7/2020. 47. Pavlus,newsletters/pracinno/Mar18_PracticeInnovations.pdf J. 2019. “Machines Beat Humans on a Reading Test. 35. Lauderdale, B., and Clark, T. 2012. “The Supreme Court's Quanta Magazine. Oct. 17. https:// many median justices.” American Political Science Review, www.quantamagazine.org/machines-beat-humans-on-a- 106 (4): 847-866. reading-test-but-do-they-understand-20191017/But Do They Understand?” Accessed 36. - 7/7/2020. ligence and Computational Technologies.” LegalTech Lever 1 48. Pennington, J., Socher, R., and Manning, C. 2014. “Glove: https://www.legaltechlever.com/2018/09/training-lawyers-Linna, Jr., D. 2018. “Training Lawyers to Assess Artificial Intel Global vectors for word representation.” In Proceedings of the 2014 Conf. on Empirical Methods in Natural Language Accessed 3/10/2019. Processing. EMNLP. pp. 1532-1543. 37. Liu,assess-artificial-intelligence-computational-technologies/ Y., Ott, M., Goyal, N., Du, J., Joshi, M., Chen, D., Levy, O., 49. Lewis, M., Zettlemoyer, L., and Stoyanov, V. 2019. “Roberta: Services.” Suffolk University Law School Research Paper Perlman, A. 2017. “Reflections on the Future of Legal A robustly optimized BERT pretraining approach.” ARXIV Preprint arXiv:1907.11692. 3/10/2019. pp. 1-11. No. 17-10. https://ssrn.com/abstract=2965592 Accessed 38. Medvedeva, M., Vols, M., and Wieling, M. 2020. “Using machine 50. Pivovarov, V. 2019. “Future Law School. What Does It Look learning to predict decisions of the European Court of Hu- Forbes 5 https://www.forbes.com/sites/valentin- man Rights.” Artificial Intelligence and Law, 28(2): 237-266. pivovarov/2019/02/12/futurelawschool/#67a0dc2f6a84 Like?” 39. Mikolov, T., Sutskever, I., Chen, K., Corrado, G., and Dean, J. Accessed 3/10/2019 2013. Distributed representations of words and phrases and

Law in Context, Vol 37, Issue 1, 2020 173 ISSN: 1839-4183

51. Reed, C., Kennedy, E., and Silva, S. 2016. “Responsibility, 59. Shulayeva, O., Siddharthan, A., and Wyner, A. 2017. “Recogniz- Autonomy and Accountability: Legal Liability for Machine ing cited facts and principles in legal judgements.” Artificial Learning.” Queen Mary School of Law Legal Studies Research Intelligence and Law 25 1: 107–126. Paper No. 243/2016: 1-17, 26-31. October 17. https:// 60. Simon, M. Lindsay, A., Sosa, L., and Comparato, P. 2018. “Lola v. Skadden and the Automation of the Legal Profession.” Yale Accessed 26/11/2020 J.L. & Tech. 20: 234-310. papers.ssrn.com/sol3/papers.cfm?abstract_id=2853462 52. Reid, M. 2018. “A Call to Arms: Why and How Lawyers and 61. Surdeanu, M., Nallapati, R., Gregory, G. Walker, J. and Manning, U. Tol. C. 2011. “Risk Analysis for Intellectual Property Litigation.” L. Rev. 50: 477-489. In Proceedings ICAIL-11. p. 116-120. Law Schools Should Embrace Artificial Intelligence.” 53. 62. Surden, H. 2014. “Machine Learning and Law”, Wash. L. Rev. rhetorical roles for segmentation and summarization of a 89: 87-116. Saravanan, M. and Ravindran, B. 2010. “Identification of legal judgment.” Artificial Intelligence and Law 18, 1: 45–76. 63. Walker, V. 2007. “A default-logic paradigm for legal fact- 54. 64. Zentgraf, D. 2015. “What Every Programmer Absolutely, Savelka, J. 2019. Statutory_Interpretation, https://github. 55. Savelka, J. and Ashley, K. 2018. “Segmenting US Court Deci- Positivelyfinding.” Jurimetrics Needs to Know 47: 193-244. About Encodings and Character com/jsavelka/statutory_interpretation Accessed 7/7/2020 Proceedings Sets to Work with Text.” Kunstube, https://kunststube.net/ of the 31st Int’l Conf. on Legal Knowledge and Information encoding/ Accessed 7/7/2020 sions into Functional and Issue Specific Parts.” In Systems. Jurix-2018. Amsterdam: IOS Press pp. 111-120. 65. Zhang, P. and Koppaka, L. 2007. “Semantics-based legal ci- 56. Savelka, J., Walker, V., Grabmair, M., and Ashley, K. 2017. tation network.” In Proceedings ICAIL-07). New York: ACM. “Sentence Boundary Detection in Adjudicatory Decisions pp. 123–130. in the United States.” Traitement Automatique des Langues. 58: 21-45. CASE LAW 57. United States: intelligenceSavkar, V. 2019. and machine“How Will learning.” Artificial Above Intelligence the Law. https://Change Dynamo Holdings et al. vs. Commissioner of Internal Revenue, abovethelaw.com/legal-innovation-center/2019/06/20/Law Schools? How law schools can evolve using artificial U.S. Tax Court, Docket No. 2685-11, 8393-12; July 13, 2016. - cessed 3/10/2019 Lola v. Skadden, Arps, Slate, Meagher & Flom LLP, 620 F. App'x 58. Sergot,how-will-artificial-intelligence-change-law-schools/ M., Sadri, F., Kowalski, R., Kriwaczek, F., Hammond, Ac 37, 620 Fed. Appx. 37 (2d Cir., 2015). P. and Cory, H., 1986. “The British Nationality Act as a logic program.” Communications of the ACM, 29(5), pp. 370-386.

174 Law in Context, Vol 37, Issue 1, 2020 Received: November 28, Date of acceptance: December 12, 2020, Date of publication: December 20 2020, DOI: https://doi.org/10.26826/law-in-context.v37i1.129

Research Note LYNX: Towards a Legal Knowledge Graph for Multilingual Europe

By Víctor Rodríguez-Doncel, Orcid: https://orcid.org/0000-0003-1076-2511 Associate Professor at the Artificial Intelligence Department of Universidad Politécnica de Madrid, and Elena Montiel-Ponsoda, Associate Professor at the Applied Linguistics Department of Universidad Politécnica de Madrid, Orcid: https://orcid.org/0000-0003-3263-3403

Universidad Politécnica de Madrid, Madrid, Spain

ABSTRACT Lynx is an innovation project in Europe whose objective is to develop services for legal compliance. A legal knowledge graph is built over multilingual, multijurisdictional documents using semantic web technologies. A collection of services implementing natural language techniques enables better legal information retrieval, cross-lingual answering of questions and information discovery. Three use cases are discussed, as well as the overall impact of the project.

Keywords – Lynx Legal Knowledge Graph, compliance services, European legislation, multilingualism

Acknowledgements: This work has been funded by the project Lynx, which has received funding from the European Union’s Horizon 2020 research and innovation program under grant agreement no. 780602. For more information: http://www.lynx-project.eu. Disclosure statement – No potential conflict of interest was reported by the authors. License – This work is under Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) https://creativecommons.org/licenses/by-nc-sa/4.0/ Suggested citation: Rodríguez-Doncel, V. and Montiel-Ponsoda, E. 2021. “Lynx:Towards a Legal Knowledge Graph for Multilingual Europe”, Law in Context, 37(1): 175-178 , DOI: https://doi.org/10.26826/law-in-context.v37i1.129

Summary 1. Introduction 2. Results of Lynx project 2.1 A multilingual legal knowledge graph 2.2 Software 2.3 Demonstration in diverse scenarios 3. Impact of the project 4. Conclusions 5. References

Law in Context, Vol 37, Issue 1, 2020 175 ISSN: 1839-4183

6. INTRODUCTION natural language processing techniques had experienced The European Union (EU), post-Brexit, is comprised a decisive leap forward that made their joint exploitation of 27 member states populated by approximately 450 timely (Montiel-Ponsoda and Rodríguez-Doncel 2018). million people, who speak over 60 indigenous languages Technology would provide affordable services for compliance to be consumed by small and medium enterprises languages, and every EU national has the right to use any (SMEs). Funded with 3M€ until April 2021, the Lynx pro- ofwith these different 24 languages legal status. to contact The EU the has EU adopted institutions, 24 official and ject was run by a consortium of 11 partners integrating institutions are obliged to reply in the same language. universities (doing fundamental research), IT companies Although the European Commission favors three of these (providing real-scale infrastructure), language experts languages as ‘procedural languages’ (English, French and (machine translation companies, editors of dictionaries) German), EU law and many other legislative documents - - tions in their digitization processes). The main results of ever, member states do not translate their legislation into theand projectlaw firms are (which the following: are experiencing heavy transforma are published in all official languages except Irish. How location in one’s own language is not easy. 2.1. A MULTILINGUAL LEGAL KNOWLEDGE GRAPH foreignThe languagesdouble fragmentation and finding the (of applicable Law and norms languages) in any A knowledge graph is a collection of interconnected commercial exchanges in Europe. This has been recognized entities with semantic types and properties, represented byhampers the EU theauthorities, development who set of thea unified completion market of thewith Digital fluid in a machine-readable form (Kroetsch and Weikum 2016). Single Market as one of their 10 political priorities for the Much of the progress made by knowledge-related informa- term 2014-20191. The actions required to implement this tion systems in the last few years is owed to this technique, goal include disparate measures such as the abolishment applied by large companies in private knowledge graphs of roaming charges for mobile telephones across Europe or uploaded into the commons as linked open data by a and the funding of applied research towards lowering myriad of parties (Bizer 2009). The cornerstone of the barriers. Thus, the EU funded the now expiring Horizon Lynx project is a knowledge graph of legal and regulatory 2020 Framework Program (H2020)2 with 77,000 M€, information that brings together legislation and other and many of the projects included in the program have legal resources from several jurisdictions in Europe. The focused on language technologies. technology that enables this integration of multiple sources The H2020 Lynx project (‘Building the Legal Knowl- is Resource Description Framework (RDF), a framework edge Graph for Smart Compliance Services in Multilingual Europe’) is one of these efforts. This article describes how the combination of Semantic Web technologies ofspecified the W3C by forthe theWorld Semantic Wide Web Web, Consortium data becomes (W3C) more to and natural language processing techniques applied in connectedrepresent data and in is graph more structures. easily interoperable. Using the specifications Application the legal domain by the Lynx project opens the door to to the legal domain has been described by Casanovas et a new breed of legal information systems, cross-lingual - and cross-jurisdictional, whose opportunities are yet to jor legal resources emerging from the European Union, be fully explored. Spain,al. (2016). Austria, So far, Germany the Lynx and project the Netherlands, has identified necessary the ma to provide exploitable compliance services for Lynx business cases (Conejero et al. 2018). An ontology or data 7. RESULTS OF LYNX PROJECT model has been created to identify and describe legal The Lynx project began in December 2017 with the documents with metadata, structuring the content in as objective of developing services to facilitate compliance in minimal way as possible. The so-called ‘Legal Knowledge multilingual, multijurisdictional scenarios. The hypotheses were that semantic web technologies were mature and that Graph Ontology’ defined in Lynx is based on the European Legislation Identifier (ELI) and NLP Interchange Format 1 A The end-of-term assessment of these priorities has been evaluated in a report: European Commission (2019). The Juncker Commission's ten priorities. Doi: 10.2861/618373 2 http://ec.europa.eu/newsroom/horizon2020/document.cfm?doc_id=17607 3 Council conclusions inviting the introduction of the European Legislation Identifier (ELI). DO C 325 de 26.10.2012

176 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

(NIF). The ELI initiative,3 adopted by the EU and several the applicable legislation and technical standards that may apply in a certain location at a certain moment in available online in a standardized format’ and includes time. These three heterogenous commercialized cases, anEU ELImember ontology states,4 which specified has ‘aproven system useful to make in harmonizlegislation- prove the versatility of the Lynx solution. ing the metadata of norms. The NIF ontology is aimed at providing text annotations with linguistic information 8. IMPACT OF THE PROJECT (Hellman et al. 2014). The Lynx project was modest in its objectives but has had some impact, both from the technological perspective 2.2. SOFTWARE and from the socio-legal perspective. 1) Lynx has demonstrated that there is a gap between knowledge graphs, the Lynx project has also developed software.In addition This softwareto having has defined taken data the formmodels of aand service- legal and the legal information provided by commercial oriented architecture and a collection of interoperable systems,the public and legal has information proved that offered the gap by canofficial be covered sources compliance-related services. These services are capable fairly cheaply. The maturity of semantic web technologies and the considerable improvement in the quality trained for the legal domain; (ii) annotation of documents, of open natural language techniques have created an identifyingof (i) neural namedmachine entities translation, such usingas companies, models specifically temporal opportunity for leveraging enriched public legal in- entities or persons; (iii) summarization of documents; (iv) formation to develop compliance by design solutions. question answering and cross-lingual search (Khvalchik 2) Lynx has made use of recently created resources in et al. 2019) and (v) terminologically-related services. the linguistic linked open data cloud (Cimiano et al. 2020), demonstrating their value in a new application have been put in place (Schneider et al. 2018 and 2020): domain. Legal thesauri, terminologies and ontologies originalPopulation documents workflows have orchestratingbeen annotated these with services these help to improve cross information retrieval tasks in NLP-based services, uniformly structured and linked to different subdomains, such as data protection (Pandit internal and external references. A major role in provid- et al. 2018), intellectual property (Rodríguez-Doncel ing intelligence to these services is provided by the use et al. 2015), economics (Neubert 2009), and criminol- - ogy (Schmidt et al. 2020). formation retrieval operations in cross lingual scenarios 3) Lynx has demonstrated the need to further develop (Martín-Chozasof domain-specific et al.terminologies, 2019). which support the in uniform standards across Europe. Much as the ELI - scribed in the EU and some of its member states, there 2.3. DEMONSTRATION IN DIVERSE SCENARIOS isharmonized a need for the institutions way legislation to adopt is identified uniform contentand de Three different application scenarios have been consid- structuring for legislative documents, using standards such as CEN Metalex5 and OASIS LegalDocML.6 The to answer labor law-related questions posed by lawyers in legislator is thereby encouraged to issue legislation plainered by language, the Lynx considering project. The both aim legislation of the first and use collective case was not in traditional formats such as PDF, but also in bargaining agreements. The second use case, focused on machine-readable forms. contract analysis, aimed at extracting the key information 4) Access to legal information is democratized, and citi- in text contracts of any kind, making them available for zens have access to information previously available further processing in a structured form, thus reducing costs, and corporate and personal risks. The third use case, some commercial software systems (such as PoolParty7 related to geothermal energy projects, aimed at retrieving onlyor Tilde’s to law translation firms paying services fees. Although8), and some Lynx non-open has used 4 https://publications.europa.eu/en/web/eu-vocabularies/eli 5 http://www.metalex.eu/. 6 http://www.legalxml.org/ 7 http://poolparty.biz 8 http://tilde.com 9 https://www.lexicala.com/

Law in Context, Vol 37, Issue 1, 2020 177 ISSN: 1839-4183

data (such as KDictionaries’ Lexicala9), a large part of 7. Khvalchik, M., Blaschke, C., and Revenko., A. 2019. “Question the source code and resulting data have been licensed Formulation and Question Answering for Knowledge Graph openly. Completion”. In Anderst-Kotsis G. et al. (eds), Database and Expert Systems Applications. DEXA 2019, Communications in Computer and Information Science, 1062, Cham: Springer, 9. CONCLUSIONS pp. 166-171. Europe has invested in technologies to build a digital 8. Kroetsch, M. and Weikum, G. 2016. “Special Issue on Knowl- single market that overcome language and legal hetero- edge Graphs”. Journal of Web Semantics, 37 (38): 53-54. geneities, and NLP and semantic web technologies have much to offer, as the Lynx project has shown –although 9. Martín-Chozas, P., Montiel-Ponsoda, E., and Rodríguez-Doncel, much effort is still required. The legal knowledge graph V. 2019. “Language resources as linked data for the legal domain”. In G. Peruginelli and S. Faro (eds.), Knowledge of the is extended to limited jurisdictions and gives coverage to Law in the Big Data Age. Amsterdam: IOS Press, pp. 170-180. justdeveloped a few inlanguages. this project Moreover, covers only it ignoresvery specific case domains, law (al- 10. Montiel-Ponsoda, E., and Rodríguez-Doncel, V. 2018. “Lynx: though other existing projects made great progress, like Building the legal knowledge graph for smart compliance Boella et al. 2015) and the accrual methods to update the services in multilingual Europe”. In G. Rehm, V. Rodríguez- graph are limited to the project term. However, the Lynx Doncel, and J. Moreno-Schneider (eds.) Proceedings of the project has set a precedent for others. The same needs 1st workshop on LREC (language resources and technologies covered by Lynx appear in other regions of the globe for the legal knowledge graph) workshop, Miyazaki, Japan, and the overall prospect of a new way of publishing and pp. 19-22. http://lrec-conf.org/workshops/lrec2018/W22/ citizens alike. 11. Neubert, J. 2009. “Bringing the "Thesaurus for Economics" pdf/book_of_proceedings.pdf#page=26 interlinking legislation will be beneficial for SMEs and on to the Web of Linked Data”. Linked Data On the Web, 10. REFERENCES 12. Pandit, H., Fatema, K., O’Sullivan, D., and Lewis, D. 2018. LDOW, http://ceur-ws.org/Vol-538/ldow2009_paper7.pdf 1. Bizer, C. 2009. "The Emerging Web of Linked Data." IEEE "GDPRtEXT-GDPR as a linked data resource." In European Intelligent Systems, 24 (5): 87-92. Semantic Web Conference (ESWC). Cham: Springer, pp. 481-495. 2. Boella, G., Di Caro, L., Graziadei, M., Cupi, L., Salaroglio, C. 13. Rodriguez-Doncel, V., Santos, C., Casanovas, P., Gómez-Pérez, E., Humphreys, L., and Simov, K. 2015. “Linking legal open A., and Gracia J. 2018. “A Linked Data Terminology for Copy- data: breaking the accessibility and language barrier in right Based on Ontolex-Lemon”. In U. Pagallo, M. Palmirani, European legislation and case law”. In Proc. of the 15th Int. P. Casanovas, G. Sartor, and S. Villata (eds.) AI Approaches Conf. on Artificial Intelligence and Law, ACM, pp. 171-175. to the Complexity of Legal Systems. AICOL, LNCS 10791, 3. Casanovas, P., Palmirani, M., Peroni, S., Van Engers, T., and Springer, Cham, pp. 410-423. Vitali, F. 2016. “Semantic web for the legal domain: the next 14. Schmidt, D., Dal Bosco, A., Trojahn, C., Vieira, R., and Quaresma, step”. Semantic Web, 7 (3): 213-227. P. 2020. “Aligning IATE Criminal Terminology to SUMO”. In 4. Cimiano, P., Chiarcos, C., McCrae, J. P., and Gracia, J. 2020. Int. Conf. on Computational Processing of the Portuguese “Linguistic linked open data cloud”. In Linguistic Linked Data Language. LNCS 12037, Cham: Springer, pp. 98-108. (pp. 29-41). Cham: Springer Nature, pp. 29-41. 15. 5. González-Conejero, J., Casanovas, P., and Teodoro, E. 2018. Manager for Curation Technologies in the Legal Domain”. “Business Requirements for Legal Knowledge Graph: the InSchneider, G. Rehm, J.M. V. Rodríguez-Doncel, and Rehm, G. 2018. and “Towards J. Moreno-Schneider a Workflow LYNX Platform”. In Technologies for Regulatory Compli- (eds.) Proceedings of the 1st workshop on LREC (language ance TERECOM@ JURIX, pp. 31-38. http://ceur-ws.org/ resources and technologies for the legal knowledge graph) Vol-2309/03.pdf workshop, Miyazaki, Japan, pp. 30-35. 6. Hellmann, S., Lehmann, J., Auer, S., and Brümmer, M. 2013. 16. Schneider, J. M., Rehm, G., Montiel-Ponsoda, E., Rodriguez- “Integrating NLP using linked data.” In H. Alani et al. (eds.), Doncel, V., Revenko, A., Karampatakis, S., and Maganza, F. International Semantic Web Conference: ISWC 2013, Part II, 2020. “Orchestrating NLP Services for the Legal Domain”. In LNCS 8219, Berlin, Heidelberg: Springer, pp. 98-113. Proceedings of the 12th Language Resources and Evaluation

178 Law in Context, Vol 37, Issue 1, 2020 Received: November 14 2020, Date of acceptance: December 12, 2020, Date of publication: January 3 2021, DOI: https://doi.org/10.26826/law-in-context.v37i1.134

Research Note Rules as Code

By Matthew Waddington, Orcid: https://orcid.org/0000-0002-2396-5417

Senior Legislative Drafter at the Legislative Drafting Office of Jersey,

Legislative Drafting Office, Jersey ABSTRACT “Rules as Code” is a label adopted by people in governments working on the idea of encoding (or just marking up) legislation while it is being drafted, so that the logic of the resulting legislation can be “read” (and checked) by a computer, to improve the manner in which legislation is produced and the way in which it is available digitally. The idea is currently attracting attention in the legal community, particularly on the question of whether it will lead to

by a legislative drafter. It argues that this conception of Rules as Code does not have the over-reaching ambitions that itautomation may sometimes replacing appear human to harbour, interpretation and in ofparticular the law. Thisdoes note not trespass briefly describes into removing Rules askey Code, interpretative at least as conceivedfunctions.

Keywords – Rules as Code, Legislation, coding, statutory interpretation

Acknowledgements: I thank Jeffrey Barnes and Pompeu Casanovas for their attentive reading and useful comments.

Disclosure statement – No potential conflict of interest was reported by the authors.

License – This work is under Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) https://creativecommons.org/licenses/by-nc-sa/4.0/

Suggested citation: Waddington, M. 2020. “Research Note. Rules as Code.”, Law in Context, 37 (1): 179-186 , DOI: https://doi.org/10.26826/law-in-context.v37i1.134

Summary 1. Introduction 2. What is distinctive about the Rules of Code approach? 2.1 “Coding” before enacting 2.2 Not executable programs ready to “implement” law, let alone automatically 3. Is Rules as Code a misconceived attempt to hand interpretation of law over from humans to computers? 3.1 A source of misunderstanding 3.2 Interpretation and imported material 3.3 A spectrum of interpretation 4. References

Law in Context, Vol 37, Issue 1, 2020 179 ISSN: 1839-4183

is not mainly driven by ambitions for automation, and may represent one of the best defences against the urge 5. INTRODUCTION in some quarters to automate legal decisions that should be made by humans. “Rules as Code” is currently a question, rather than a theory or a product. It is a label of convenience for a set But it is common for those coming upon the idea of of attempts to investigate the question of whether a new approach to coding (or at least marking up) legislation have the same ambitions and faults as some of the wilder fringesRules as of Code the computational for the first time law approach. to assume Given that Rulesit must as the way legislation is produced and the way it is made - mightavailable produce through a workablethe medium scheme of computer which programs. benefits both The ing that those of us working on it (and others using the idea started in New Zealand (GNZ 2018, 2019), before labelCode foris such other an purposes) open and sometimespractical field, use it generalisations is not surpris spreading to New South Wales (GNSW 2019), and then or shorthand that can be misconstrued as implying that to Canada1, Jersey2, the United Kingdom3 and elsewhere4, Rules as Code is about removing interpretation, discre- and then being reported on by OPSI5. tion and the human role from all of the law (or as if it is about AI, or as if no lawyers are working in it, or as if it is It is a practical movement led by developers and just hype and ignores the history of work on computing public sector digital innovators, whose preferences are and law, and so on). This note is an attempt to cast some for “show, don’t tell”, “sprints”, “wireframe models” and light on those assumptions and misleading impressions, “minimum viable products”. That approach is welcome and to invite those interested in the issues to engage positively without assuming that Rules as Code must be adopt any resulting scheme, and who tend to prefer up to no good. to thesee legislativea demonstration drafters of and something policy officials workable, who ratherwould than a theoretical explanation. They also mostly take a neutral stance on the question of what the best technical 6. WHAT IS DISTINCTIVE ABOUT THE RULES AS CODE APPROACH? one that is workable). approach might be (or even whether6 it is possible to find line on what “Rules 2.1 “CODING” BEFORE ENACTING as Code” means (or even whether it is the best label to A distinctive aspect of Rules as Code is that it builds use).This But means it is most that usefully there is noseen unified as separate from (though on the “Better Rules” work undertaken in New Zealand. related to) the idea of “computational law” in the sense of That means it is engaged with the fact that legislation is how law (of all kinds, not just legislation) might best be a conscious attempt to alter the law to give effect to some represented in code, and whether law can be adequately policy of the promoter of the legislation, and is interested represented in that way. Of course, computer programs in producing more coherent policy and legislation which have been made, sold and used for decades to imple- is more readily comprehensible and effective. So one key ment some legislation – nobody imagines that humans distinguishing feature of Rules as Code is that it is about are still sitting with quill pens in social security and tax using technology to help at the stage of forming policy, writing drafting instructions, consulting on legislative demanded. Equally, Rules as Code is separate from, but proposals and drafting legislation. That is a very different relatedoffices working to, the idea out ofwhat “automated benefits shoulddecision be making” paid, or –tax it

1 See the work of Scott McNaughton (2020) from the Government of Canada. 2 See our report to the UK and related jurisdictions’ knowledge-sharing event at the Bank of England on 28 February 2020. Cf. Waddington (2020). 3 See for example the work of Adam Wyner (2019), with the UK, Welsh & Scottish Parliamentary Counsel Offices, and on LegalRuleML, Wyner et al. (2017). 4 See for example the global 2019 Rules as Code Show and Tell Event at https://www.youtube.com/channel/UCjGU3qQWPMMuy_DSRABNTuw and the work of Sin- gapore Management University School of Law in partnership with Legalese on computational law - https://news.smu.edu.sg/news/2020/03/11/smu-awarded-15-million- grant-computational-law-research and https://cclaw.smu.edu.sg/ 5 The Observatory of Public Sector Innovation (an initiative of the OECD) produced a report in its “Innovation Primer” series, titled “Cracking the Code: Rulemaking for humans and machines”, Cf. Mohun and Roberts (2020). 6 This note therefore should not be seen as definitive. It is one participant’s contribution, from the perspective of a legislative drafter (rather than a coder or other digital innovator), reflecting perhaps a less ambitious view of Rules as Code than many. Twitter gives a good overview of diversity of work and views on Rules as Code, using #RulesAsCode. A key figure is Pia Andrews, who has worked in New Zealand, New South Wales and Canada.

180 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 question from using AI or NLP to crawl back over existing duty to make a non-delegable decision). Those purposes legislation and try to interpret it to encode the results. might be to explain the legislation to particular audiences When lawmakers pass legislation they are acting on (like an advice agency for young people), or to form part their understanding of the draft, which is normally aided of the internal procedures of a particular business (like by policy explanations and legal notes from civil servants a bank’s procedures for reporting to a regulator), or as and legislative drafters, which are generally published part of a government department’s implementation of alongside the legislation (but do not have legal force, their duties under the legislation. unlike the legislation itself). A coded version of the draft That means the version published by government would need to be program-neutral as far as possible. are understanding each other, that consultees can more Many of those working on the technical side are looking easilycould helpgrasp to theensure ideas that and the suggest drafter changes,and policy and officers that at “declarative” logic programming languages—taking the lawmakers can see an illustration of what the legislation lead from the work in 1986 using Prolog on the British is intended to achieve. It could also help the drafter, and Nationality Act (Sergot et al. 1986a, 1986b)— instead the policy maker to spot inconsistencies or unintended of the usual “imperative” or “procedural” programming consequences in their drafts, improving the rigour that languages (such as Python, Java and others). Others are legislative drafters already apply to their work. looking at extending mark-up languages such as Legal- RuleML7 and Akoma Ntoso8 —including using tools like the “Rules Advanced Web Editor” or RAWE (Palmirani 2.2 NOT EXECUTABLE PROGRAMS READY TO et al. 2013a, 2013b). Others favour using more tradi- “IMPLEMENT” LAW, LET ALONE AUTOMATICALLY tional programming, but in a form such as OpenFisca9. The other main distinctive aspect is that Rules as Code is not about deciding what is the best way to turn legislation and legislative drafters should learn to encode the rules into code as such. The idea is to publish something when themselves,Similarly, there or whetheris a question computing of whether experts policy should officials join the legislation is enacted (and before if there is consul- the team at the drafting stage. Jason Morris10, creator of tation) which is then kept up to date on an “application Blawx11, argues that new rule-making software should programming interface” (an API), so that others can use be as easy for lawyers to use as spreadsheets now are it to make programs. The “coding” that the government publishes would not be an executable program that imple- to Rules as Code, Professor Robert Kowalski is arguing ments the law automatically in some way. Instead others thatfor accountants. computers should In a related be taught field, to possiblyunderstand applicable natural would be producing executable programs for their own languages, or at least controlled natural languages that particular purposes, and there is no reason to assume are easy for non-programmers to use, and his “Logical those would be primarily to implement the law, let alone English”12 is inspired13 partly by the way that legisla- to do so automatically (and of course a statute cannot tion uses a natural language (in our case English) in a be implemented automatically when it gives a human a fashion that has become increasingly disciplined and is

7 https://www.oasis-open.org/committees/tc_home.php?wg_abbrev=legalruleml 8 http://www.akomantoso.org/ 9 https://openfisca.org/en/ 10 Jason also recently started working with the Centre for Computational Law (https://cclaw.smu.edu.sg/) at Singapore Management University, in partnership with Le- galese (https://legalese.com/). Their work is much wider than Rules as Code (looking at commercial contracts and other legal texts, not just legislation), and does have much more ambitious goals (more akin to those that are often wrongly assumed to be behind all Rules as Code work). 11 https://www.blawx.com/ 12 See https://www.doc.ic.ac.uk/~rak/papers/LPOP.pdf 13 See https://www.doc.ic.ac.uk/~rak/ 14 It is interesting that Rules as Code started in Commonwealth countries and has spread in those countries. They have lawyers working full time for governments or parliaments as legislative drafters or parliamentary counsel, sharing information through the Commonwealth Association of Legislative Counsel. The modern style of legislative drafting in Commonwealth jurisdictions is much more disciplined than it was in the past, and much closer to a controlled natural language. In particular “and” and “or” are separated in lists where only one of those connectives can apply at any given level (paragraph, sub-paragraph and so on). Also the move from “shall” to “must”/“is” has shone more light on when the drafter is writing a constitutive rule (such as a definition, or a “counts as” rule) or a normative rule (which can be contravened, meaning the drafter should be satisfied that the legal consequences of contravention are clear even if they are not expressly set out) or a more troublesome rule on the borderline between the two (such as those related to validity or to whether a notice, application or appeal succeeds in having any effect in changing someone else’s legal status). For discussion of the ways these distinctions can be drawn, see for example Sartor (2006), particularly sections 11 to 13.

Law in Context, Vol 37, Issue 1, 2020 181 ISSN: 1839-4183 becoming much closer to a controlled natural language book titled Is law computable? but with the subtitle Criti- in the formal sense14. cal perspectives on law and artificial intelligence (Deakin The point is that Rules as Code is not wedded to any of and Markou 2020). Both of those are rightly concerned these potential technology solutions so much as to the idea with the dangers of digitisation in law, but the article also that the “coding” (or mark-up) of the legislation should warns that Rules as Code risks undermining “the rules be widely usable, traceable to the legislation, rather than themselves, human freedoms, and democracy” if it involves a regulatory computer system that implements a single, or implementation. government-determined interpretation of legal rules. adding material to reflect assumptions about procedures But what if, hype aside, Rules as Code is not really intended to be magically transformative, and is not really 7. IS RULES AS CODE A MISCONCEIVED ATTEMPT about automating legal decisions or about programs that TO HAND INTERPRETATION OF LAW OVER FROM HUMANS TO COMPUTERS? theimplement plumbing” law and themselves? making gradual In reality improvements Rules as Code often is 3.1. A SOURCE OF MISUNDERSTANDING associatedmore about with taking the further slogan steps “bring in backexisting boring”. work It on is “fixing about Rules as Code is a practical proposal which will go applying human intelligence, rather than AI, and about nowhere unless it sparks interest among governments, so the less glamorous ways in which computers are already there can be a temptation to exaggerate what it is meant handling law and could do better in aiding humans. It can to achieve. It also needs to bring together many people involve merely highlighting the logical structures that the from different professions who are not familiar with each drafter is trying to create in the legislation, so that any use other’s worlds, particularly lawyers with little experience of that logic should always be traceable, explainable and of computing and developers with little experience of law. open to correction or appeal in the same way as it is when a human follows the logic from the text. That could mean to interest people, or just using expressions (particularly “interpretation”As such it can sometimes and “decision-making”) end up using over-simplifications that mean dif- each other better during the drafting, that consultees ferent things to the different audiences. It is also easy canthat more legislative easily draftersgrasp what and it policyproposed officers and demonstrate understand to confuse Rules as Code with the wider “computational how it could be changed, that inconsistencies in drafts law” movement, and particularly with a false idea that can be spotted before they become problems, and that humans bring only bias and irrationality whereas com- those who need to read the legislation can be helped to puters bring only impartiality and consistency, or with navigate complex sets of cross-references, conditions and exceptions to other exceptions. Those would represent every human problem15. Lawyers coming to Rules as Code an idea that “Artificial Intelligence” will offer solutions to near automating the implementation or enforcement of to understand how law works and particularly how it is legislation.significant benefits in themselves, without going anywhere interpretedfor the first timeand thecan way assume in which that it that is based is different on a failure from the approach of a computer interpreting its programs. 3.2 INTERPRETATION AND IMPORTED MATERIAL Lawyers also rightly worry about “automated decision- In the early days there was discussion in Rules as Code making” amounting to unlawful sub-delegation or to an about the need to reassure people that the point is not to abdication of a decision-making duty. remove human discretion in legal decision-making func- The relatively small number of legislative drafters tions. That led to talk of Rules as Code being more suited involved in Rules as Code need to take on the concerns to something conceived as “prescriptive” legislation, rather of other lawyers and try to clear up misconceptions. This than supposedly “discretionary” or “outcome-based” note was prompted by a recent article headed “‘Rules legislation, and that was sometimes framed in terms of as Code’ will let computers apply laws and regulations. the computer being able to “interpret” the one but not But over-rigid interpretations would undermine our the other. But those ideas were never properly thought freedoms” (Governatori et al. 2020) as well as by a new

15 An excessive faith in technology can lead to the well-rehearsed problems of unaccountable “black box” AI. Equally problems such as the “robo-debt” issue in Aus- tralia can lead to excessive scepticism of all increased involvement of technology in the law.

182 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 through, and amount to inadvertently selling Rules as Code short. The effort to reassure people is now moving away lawyers are unaware of these rules, having been taught from the more straightforward concept of discretion16 to statutoryused (to give interpretation a more precise as if itfocus). was always We find a question that many of applying various rules to cases of ambiguity, rather than I mentioned above the idea that the “coding” (or mark- up)the ofmore the legislationdifficult concept should of be interpretation. traceable to what is already conventions and rules applicable in their jurisdiction. in the legislation, and should avoid adding material to of understandingWe also rely on how being definitions able to say, work in a and particular what are piece the of legislation, merely “A person who does X commits an There is a tendency among some critics to assume that thereflect published assumptions code must about itself procedures be executable or implementation. in a way that for Z years”, without rehearsing or explaining the general implements the legislation, and that the idea is therefore criminaloffence and law is(whether liable to statutory a fine of or£Y case and law).to imprisonment Similarly, we to hard-code in additional material (such as assumptions can say a public body “must” do something, and leave it at about steps being taken in a particular order, or informa- that, relying on our jurisdiction’s body of administrative tion being supplied electronically) to turn the legislation law to deal with what happens if the public body does not into a set of procedures that can be coded. As a legislative do what we said it must – we do not have to try to pin that drafter, I am on the wing of the Rules as Code movement down with a statutory formula that will inevitably be left that aims, as far as practicable, not even to include mate- behind by case-law and come adrift from later statutory rial derived from other legislation (as opposed to coding versions of the formula. But it is still worthwhile for us that links to the coding of another statute, where there to publish the text of the legislation without publishing is a cross-reference in the text or where an Interpreta- a compendium of criminal and administrative case-law tion Act applies) or from common law or other sources. cross-referenced to each published item of legislation. Of course, no item of legislation can be understood fully Equally, Rules as Code does not need to encode all of this without reference to material outside that item of leg- external material for it to be worthwhile creating and publishing a coded/marked-up version of an individual to publish the text of legislation as it stands17. Modern piece of legislation. So if the offence provision is “If person legislativeislation, but drafters that does (at not least stop the us full-time finding it professional worthwhile A does X or Y, but not Z, then A commits an offence [and is drafters found in Commonwealth countries) use a variety liable to …]”, then it is worthwhile to code this to capture of conventions, as well as Interpretation Acts, to achieve only the “if”, “or”, “but not” and “then”, plus the fact that A the desired result while avoiding cumbersome expression has to be the same person in both appearances (and, if the that would lose or confuse human readers. For example, there is the convention that, unless there is some other without trying to capture anything about any common indicator, the same word will mean the same thing inside lawstatute rules defines on the Z, necessary perhaps a mens reference rea or to the that presumption definition), one piece of legislation, while a different word will have of innocence, or any relevant criminal procedure rules been used deliberately to carry a different meaning. Juris- and so on. dictions often have rules such as that expressions used in Developers have a natural tendency to want to showcase primary legislation have the same meaning in secondary work where the computer produces results, so much of the legislation made under it, and conventions that although initial work has been on the readily computable areas of singular and plural imply each other the singular will be

social security benefits and tax. But work has now moved 16 As far as discretion is concerned, the idea is to use the computer to help guide the human reader to the correct point at which a discretionary decision has to be made by a human, not to have the computer make or suggest the “correct” decision. But see below as to the human interpretation needed for each step on the way to any of those decisions. 17 This applies even in an extreme case, as follows. Most jurisdictions publish the text of their legislation on their websites, separately from any publication of case-law (some try to annotate legislation with references to case-law, but that is very difficult to do, particularly if the jurisdiction publishes its legislation as amended). A court might declare a statutory provision ultra vires or say words should be read in or ignored, but the legislation website will not take down or alter the text of the legislation just because that has happened. If subsequent amending legislation is passed to react to the court decision then those amendments will be published on the legislation website in due course, but the legislation website will not jump the gun by rewriting the published legislation to reflect the court decision. So jurisdictions already operate on the basis that there is utility in publishing the text of legislation as it stands, even when that legislation fails to reflect the true state of the law, as long as it is clear that no more is being achieved. Similarly, there would be utility in publishing coding that reflects that legislation, and being clear that this is all that is being achieved. It would then be left to those who want to make programs from it to decide whether they offer something based just on the legislation (answering the question “what does the X Act say?” rather than “what are my rights?) or whether and how they try to go further by feeding in corrections from court cases or other external material.

Law in Context, Vol 37, Issue 1, 2020 183 ISSN: 1839-4183

- ardent AI evangelists have to recognise that AI can only gations of banks, and charity registration. It will become take us so far with that. In reality a computer does not on to fields such as gambling regulation, reporting obli know anything—it is a question of whether the inputs more about improving the way the rules are constructed come from other programs or from humans. So if there (toincreasingly avoid inconsistencies obvious that andthe benefits loopholes, of Rules rather as than Code to are be was a computerised register of dogs, and the legislation machine-executable) and about how they can be explained was to say that for all legal purposes something only and altered, as well as about assisting human readers to counts as a dog if it is on that register (and anything on follow the often long and twisting logical strings that the that register counts as a dog) and that a dog was to be drafter had to embed in the text. presumed/assumed to live at the address given for it on the register, then a program encoding an Act that mentions dogs could rely on computerised input from that register 3.3. A SPECTRUM OF INTERPRETATION to establish whether a dog lives at a particular address. The role of interpretation in law is obviously a vast If the Act required dog-owners to display a sign on their topic in academic law, and then there is an odd corner house to be visible from the road, then perhaps AI could be labelled “statutory interpretation” in which it seems all trained to use something like Google StreetView to check too often introductory courses encourage students to whether there was a sign on a house that appeared on believe in magical “golden rules” that have been debunked the register, and the computer could send an automated long ago by specialist texts such as Bennion (Bailey and email imposing a civil penalty on a registered owner who Norbury 2020). How judges make new law has also been did not have a sign. But of course this is nonsense—no a long-standing fascination for academic law in the com- sensible legislature is going to say that the computerized mon law world. The socio-legal studies movement has 19, and any led to a focus on social practices, not just in courts, but lawyer would expect the registered owner to be given a - rightdog register of appeal is legally against definitive any penalty of what imposed a dog isbased on AI ence what the legal system produces. But relatively little reading images of streets. Outside of the more fevered attentionin administrative is given officesin legal and education, in the way even citizens in socio-legal experi imaginations of some AI proponents, and the hype of courses, to how legislation is created and to how that some people with no qualms about selling fake miracles, makes it different from case-law and general principles. there is no sense in which anyone is really proposing that On the other hand, people with backgrounds in the dog registration and sign display rules should be fully technology rather than law tend to assume that there automated. are “government regulations” and that the government But what a computer can usefully do is take a human has the authority to dictate how those regulations are through a set of questions, where the human inputs an interpreted. One of the features of Rules as Code is the answer based on their interpretation of the words in the way that legislative drafters have been able to explain to question, and the computer uses that answer to take the their technical colleagues that once legislation is enacted human to the correct next question. In typical legislation it is for the courts, not the government or Parliament, to on dog registration, say, there would be a set of conditions interpret it. for dog registration to be mandatory, and a set of exceptions, some of which will have their own exceptions and interpretation involved in the course of applying legislation some of which will cross-refer to other rules and complex to facts.But this A computer leaves us does with not an over-simplifiedknow what “dog” idea means of the18, - because it does not know what a dog is. Even the most erences to the rule about displaying a sign, which will not definitions. There will then be exceptions and cross-ref 18 But what the legislative drafter (or accompanying coder) can do is to mark up the text, or have an element in the code, to tell the computer that every occurrence of those three letters (and the four letters for “dogs”, or any other grammatical variants) refers to the same concept (and that “the/that dog” refers back to an immediately preceding instance of “a dog”, and so on). See above on how these conventions are followed in modern legislative drafting in Commonwealth countries. This also reflects the difference between Rules as Code, which has humans apply the mark-up/coding during the drafting, and other approaches that useAI to try (after enactment) to fathom what “dog” was doing in the text. With the mark-up/coding published by government, as envisaged in Rules as Code, the point would be that a human still has to answer the question “is this a dog”. Meanwhile the mark-up/coding would enable others (whether in government or not) to create software that takes a human through the logic to the correct part of the statute to end up at that question, based on the human’s answers to other questions about the other English expressions used in that statute. 19 Equally, the published mark-up/coding would just be “official” in the sense that any government explanatory text is official. The coding is not intended to say, when the legislation does not, that something should not be recognised as a dog unless an electronic register confirms that it is a dog.That would be going beyond the legislation, and any “official” status could not be enough to give it any credence or usefulness, never mind give it any legal effect.

184 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 correlate completely with the registration rule, with the exception for dogs owned by farmers from an exceptional result that many readers of the legislation will become lost rule that farmers do not need to display the sign as long when tracking back and forth to work out their position, even though they may have no trouble (on the particular as people who work in any agricultural business. This is facts) in identifying whether the animal in question is a whatas they is goingregister to taketheir most dogs, people and it time might and define effort farmersto wade dog, whether the sign is visible from the road, and so on. through, and where most mistakes will be made by those That is where the computer can be used to capture the implementing the legislation or checking their rights, but “if-then-and-or-not” logic20 of the legislation, the use of we do not see this as sacred ground on which human input is essential. The human input is needed for whether the dog counts as a poodle, whether the owner counts as a ofdefined “dog” termsor “visible (and fromthe use the of road”. “means” If it oris an“includes” offence toin person who works in an agricultural business, and whether definitions),fail to display and the sign,so on, then rather the courtthan thewill actual have to meaning answer what the farmer did counts as displaying the sign or reg- yes or no to whether there was a dog registered at that istering the dog. Even those basic terms require human address and whether the sign was visible from the road. input, let alone a provision saying the farmer commits an Ultimately a court, using a natural language, will “inter- offence by failing to take “reasonable” steps to ensure the pret” all the relevant words of a statute when applying it to sign is not removed by anyone else, where it is obvious a set of facts – interpretation is not just about ambiguous that the court will have to apply its own human approach words. But there is a spectrum of elements of a statute, to working out what is reasonable. The human input is where the elements need a greater or lesser degree of essential for those points, rather than for unravelling the respect for the court’s function of interpretation where nested exceptions and their relationship to the main rule. there is no claim of ambiguity. So for example if there is The legislative drafter will try to make that logic as simple a Dogs Amendment Act, and it says that paragraph (1) of and transparent, and free of inconsistencies, as possible in section 1 of the Dogs Act (as it stands when the Amendment the time available and under whatever other constraints Act comes into force) is amended to substitute “poodle” might be involved (such as a need to mirror some other for “spaniel”, then, as long as there is a s1(1) which has provision). If the coding in Rules as Code merely picks the word “spaniel”, we are not really saying that identify- out the if-then-and-or-not logic of these exceptions and ing this provision and switching the words is at a level of interpretation at which we want the court to be able would produce an unintended result), then it will have to make its own mind up about the meaning (the court donedefinitions an extremely (and points valuable up a scenariojob. But itwhere will not the have first trieddraft might strike down the amendment as ultra vires, or hold to usurp the human interpretative role by trying to capture that the provision as amended should be read to cover the entire meaning of the Act’s provisions or automate both poodles and spaniels on constitutional grounds, the entire handling of the Act. This is much less exciting but that is a different story). If the coding captures the for those who want to see Rules as Code as just another substitution provision, nobody is going to object to that attempt to have computers take over the world of law, but as usurping the judicial or human function of interpreta- it is still exciting enough for legislative drafters. tion. In fact several jurisdictions already do produce and publish consolidated versions of their legislation, using 8. REFERENCES XML coding in the original legislation and the amending legislation to do the job (by identifying a paragraph as 1. Bailey, D., Norbury, L. 2020. Bennion, Bailey and Norbury on being s1(3)(d) of the X Act as amended up to a certain Statutory Interpretation. 8th Edition, LexisNexis Butterworths. date, for example), without causing an outcry about hav- 2. Deakin, S., Markou, C. (eds.). 2020. Is Law Computable? Criti- ing added in and hard-coded their own interpretation.21 cal Perspectives on Law and Artificial Intelligence, Oxford: Similarly the Dogs Act, before the amendment, might Hart Publishing. say that if the dog is a spaniel then it is not covered by an

20 The logic that captures the relations in “if X then Y”, “X and Y”, “X or Y” and “not X”. For instance “or” can be inclusive or exclusive (not both X and Y). Usually the intended meaning is obvious enough to a human, so we do not overburden the reader by spelling out in the English text. But when more is needed a drafter might write “X or Y, but not both”, “either or both of X and Y”, or “any one or more of the following”, and so on. 21 See for example New South Wales https://beta.legislation.nsw.gov.au/help/xml-pdf-files and the United Kingdom https://www.legislation.gov.uk/developer/formats/ xml

Law in Context, Vol 37, Issue 1, 2020 185 ISSN: 1839-4183

3. GNZ. Government of New Zealand. 2018. Better Rules for Go- on Rules. In CEUR workshop proceedings, Vol. 1004. http:// vernment Discovery Report. March, https://www.digital.govt.nz/ ceur-ws.org/Vol-1004/ dmsdocument/95-better-rules-for-government-discovery-report/ 10. Palmirani, M., Cervone, L., Bujor, O., Chiapetta, M. 2013b. html “RAWE: An Editor for Rule Markup of Legal Texts”, OA- 4. GNZ. Government of New Zealand. 2019. What is Better SIS, https://2013.ruleml.org/presentations/RuleML- Rules? By H. Fraser, 20 December, https://www.digital.govt. palmirani2013v2.pdf nz/blog/what-is-better-rules/ 11. Sartor, G., 2006. “Fundamental legal concepts: A formal and 5. GNSW. Government of New South Wales. 2019. Rules as Code. teleological characterisation.” Artificial Intelligence and Law, https://www.digital.nsw.gov.au/digital-transformation/ 14 (1-2): 101-142. policy-lab/rules-code-0 12. Sergot, M.J., Sadri, F., Kowalski, R.A., Kriwaczek, F., Hammond, 6. Governatori, G., Barnes, J.,Zeleznikow, J., de Koker, L., Poblet, P. and Cory, H.T. 1986a. “The British Nationality Act as a M., Hashmi, M., Casanovas, P. 2020. “‘Rules as Code’ will logic program”, Communications of the ACM, 29(5): 370-386. let computers apply laws and regulations. But over-rigid 13. Sergot, M., Cory, T., Hammond, P., Kowalski, R., Kriwaczek, F. interpretations would undermine our freedoms”, The Con- and Sadri, F., 1986b. “Formalisation of the British nationality th versation, November 26 . https://theconversation.com/ act”, International Review of Law, Computers & Technology, rules-as-code-will-let-computers-apply-laws-and-regula- 2 (1): 40-52. tions-but-over-rigid-interpretations-would-undermine- 14. - our-freedoms-149992 man input, discretion & logic.” Rules as Code Knowl- 7. McNaughton, S. 2019. “Week 64 — The State of Rules as edge-SharingWaddington, M. Event, 2020. 28 “Non-financial February, Bank legislation: of England. hu Code in the Government of Canada”. https://medium.com/@ https://www.slideshare.net/MatthewWaddington3/ mcnaughton.sa/week-64-the-state-of-rules-as-code-in-the- government-of-canada-8f3cb327448d 15. Wyner, A. 2019. “Annotating and Querying Content within 8. Mohun, J. and Roberts, A. 2020. "Cracking the code: Rule- non-financial-legislation-human-input-discretion-and-logic-28220- making for humans and machines", OECD Working Papers on mentary Counsel, London May 21. https://azwyner.info/ Public Governance, No. 42, OECD Publishing, Paris, https:// Machine-readable Legal Instruments”, Office of the Parlia doi.org/10.1787/3afe6ba5-en. 16. Wyner, A., Gough, F., Levy, F., Lynch, M. and Nazarenko, A. 9. Palmirani, M., Cervone, L., Bujor, O. and Chiappetta, M., 2013a. 2017.wp-content/uploads/2019/05/WynerUKGovt_v3.pdf “On Annotation of the Textual Contents of Scottish “RAWE: a web editor for rule markup in LegalRuleML”. Pro- Legal Instruments”. A. Wyner and G. Casini (eds.) Legal ceedings of the 7th International Rule Challenge, the HLT and Knowledge and Information Systems, JURIX-2017, Amster- the DC at RuleML2013, the 8th International Symposium dam: IOS Press, 101-106.

186 Law in Context, Vol 37, Issue 1, 2020 Received: December 14 2020, Date of acceptance: December 15 2020, Date of publication: December 27 2020, DOI: https://doi.org/10.26826/law-in-context. v37i1.132

Book Review Advanced Introduction to Law and Artificial Intelligence

By Stephanie Falconer, Associate Lecturer of Law at La Trobe University, Orcid: https://orcid.org/0000-0002-5773-7375 La Trobe University, Melbourne, Australia

ABSTRACT

Advanced Introduction to Law and Artificial Intelligence. Cheltenham, UK: Edward Elgar Publishing, ISBN-10: 1839100311, ISBN-13: 978-1839100314, 208 pp. Book Review: Woodrow Barfield and Ugo Pagallo. 2020. Keywords – Book review, artificial intelligence, law, advanced introduction.

Acknowledgements: The Book was provided free of charge by Edward Elgar. Disclosure statement – No potential conflict of interest was reported by the author. License – This work is under Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) https://creativecommons.org/licenses/by-nc-sa/4.0/ Suggested citation: Falconer, S. 2020. “Book Review: Advanced Introduction to Law and Artificial Intelligence Law in Context,

37(1): 187- 189, DOI: https://doi.org/10.26826/law-in-context.v37i1.132 , by W. Barfield and U. Pagallo”,

Law in Context, Vol 37, Issue 1, 2020 187 ISSN: 1839-4183

The increasing sophistication and rapid development The middle section of the book deals thematically with AI’s disruptive impact on legal doctrine. Arguably the most to complex legal challenges. In Advanced Introduction to important substantive legal question for AI is the attribu- Lawof artificially and Artificial intelligent Intelligence technology (Advanced (AI) hasIntroduction) given rise, tion of liability for damage caused by smart technology, - and chapter four covers this issue. A potential point of ing primarily on examples from Europe and the US, give aDr broad Woodrow and accessible Barfield and overview Professor of AI Ugo and Pagallo, the law. draw personhood,frustration for but readers this is likelyseeking because definitive the technology,answers is thatand The Advanced Introduction thusthe authors the discussion, do not commit is still developing.to specific conclusions Nonetheless, about the text guides the reader through the various positions and the Research Handbook on the Lawis not of the Artificial first collaborative Intelligence (alsoresource published produced by Edwardby Barfield Elgar and in Pagallo, 2018). whoBoth coedited authors further exploration of this challenging and fundamental concept.provides sufficient information for those interested in in Engineering, a JD, and an LLM in intellectual property law,are clearly and has experts written in extensively their field. Dron BarfieldAI, robotics, holds and a PhD human enhancement. Professor Pagallo, a legal academic, EuropeanChapter General five examines Data dataProtection protection, Regulation another (GDPR) significant as a stream of monographs and other publications dealing concern‘an (almost) raised all-encompassing by the proliferation regulation’ of AI, and informing identifies the withhas made AI and a significantthe law, governance, contribution and to information scholarship techwith- standpoint of many legal systems around the world. Here nology law. the authors focus more on how AI may impact existing laws, identifying gaps and ambiguities to provide insight Unlike the Research Handbook, this book serves as into likely future legal development. a grounding resource for a wider audience including legislators and government agencies who seek to regu- Chapter six explores the intersection of AI with tort late AI, law students and students from other academic law, with an interesting examination of challenges that AI disciplines seeking an introduction to the role of law in poses to the duty of care. When a system can think and regulating smart technology, and practicing lawyers who - deal with AI issues in litigation. Though the authors deal plore potential solutions proffered by legal and computer with the content in a clear and careful manner, without scienceact for itself, experts, who such owes as the the duty conferral of care? of Thepersonhood authors exon a grounding in general legal principles and a degree of AI systems, establishing a ‘common enterprise’ theory of understanding about AI and its uses, readers may become overwhelmed with the extremely nuanced and complex of care owed. concepts discussed. liability, or the modification of duties and the standard Criminal law is the focus of chapter seven. Here the authors investigate the right to a fair trial for accused theories, and the essential concepts, such as algorithms, persons, the emergence of potential new AI-based expertThe booksystems, begins and by establishingmachine learning, the definitions, that ground actors, crimes, and whether AI can be considered a ‘new sort subsequent discussion. The authors refer to accessible of accountable criminal agent’. The primary takeaway examples such as the Turing Test and CAPTCHA, enabling from this chapter is that the limitations of working with the reader to grasp the thresholds of AI’s actual utility, AI, whether as a legal support system, as a new subset as well as its present limitations. Chapter two explores of criminal offence, or as a legal subject capable of com- the human rights considerations raised by AI technology mitting crimes independent of human involvement (and including a brief reference to Amnesty International’s pilot with the necessary actus reus and mens rea), are still too program using AI’s machine learning capabilities to assist much in their infancy to be able to clearly address. As the with investigations into alleged human rights violations. technology develops, so too will the response and the Chapter three then considers the issues arising from AI’s authors have outlined clear arguments to help guide the potential to interfere with constitutional rights, with a discussion when the time comes. particular focus on the US and European Union.

188 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183

Chapters eight, nine and ten examine the impacts of AI on the law of copyright, patent and business. Chapter and development of law in the areas of administrative law, eight analyses the challenge of determining authorship and The book finishes with a call for further consideration ownership where works are generated by AI with little- of war, workplace law, and human enhancement. As the to-no human involvement. Patent law and, in particular, authorsfinancial make regulations, clear, articulating health and “precision a coherent medicine”, legal frame law- when AI, or AI-created work, is eligible for patenting, is work is challenging because of the developing nature of the focus of chapter nine, with the authors taking the AI technology. As such, the conclusions are appropriately position that AI should not be awarded inventor status, tempered. Nonetheless, the book is full of helpful exam- and arguing that identifying a human author or owner ples, and the exploration of relevant treaties and legislation carefully situates the reader within the appropriate on the inadequacy of existing trade secrecy law, and the context. Ultimately, this book provides an authoritative potentiallymay become harmful increasingly effects difficult.of the lack Chapter of transparency ten focuses in AI-powered business transactions. springboard into further research, and will prove a useful resourceintroduction for itsinto intended the specific audience. legal topics covered, and a

Law in Context, Vol 37, Issue 1, 2020 189 Received: December 18 2020, Date of acceptance: December 19 2020, Date of publication: December 20 2020, DOI: https://doi.org/10.26826/law-in-context. v37i1.130

Book Review Anne Wesemann: Citizenship in the European Union Constitutionalism, Rights and Norms

By David Wishart, Adjunct Professor of Law the Law School, La Trobe University, Orcid: https://orcid.org/0000-0003-2281-4745

La Trobe University, Melbourne, Australia

ABSTRACT

Book Review: Anne Wesemann. 2020. Citizenship in the European Union. Constitutionalism, Rights and Norms. Cheltenham: Edward Elgar Publishing. ISBN: 9781839103162, 192pp.

Keywords – Constitutionalism, European Union, Citizenship, European Court of Justice, Robert Alexy

Acknowledgements: The Book was provided free of charge by Edward Elgar. Disclosure statement – No potential conflict of interest was reported by the authors. License – This work is under Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) https://creativecommons.org/licenses/by-nc-sa/4.0/ Suggested citation: David Wishart, 2020. “Book Review: Citizenship in the European Union Constitutionalism, Rights and Norms, by A. Wesemann”, Law in Context, 37(1): 190- 193, DOI: https://doi.org/10.26826/law-in-context.v37i1.130

Summary

1. A book with Four Plots 2. The structure 3. The elements Spoiler Alert! 4. Conclusion

190 Law in Context, Vol 37, Issue 1; 2020 ISSN: 1839-4183

5. A BOOK WITH FOUR PLOTS any such question. She does, however, answer such ques- A good soap opera needs at least three plots to keep it question, ‘What is the law?’ Wesemann does not answer entertaining. One doubts that is true for academic books yet it is the case for Anne Wesemann, Citizenship in the dotions those as, ‘Howanswers are meandecision for as the to nature citizenship of the arrived European at?’ European Union: Constitutionalism, Right and Norms, ‘What is the role of the European Court of Justice?’ ‘What- Edward Elgar: Cheltenham, 2020. Indeed, its title is something of a misnomer: it is only in part about Citizenship withUnion?’ the And argumentative ‘Can German methodologies jurisprudence of bethe applied humanities out in the European Union. Its subtitle, Constitutionalism, ratherside Germany?’ than deploying She applies the legal herself reasoning to these of deductionquestions Rights and Norms, hints at more but still sells the book from rules or induction from cases. Rather, she tests her somewhat short. The stories to be found in the book are indeed fourfold: approach. While such methodological confusion might 1. Can the constitutional rights norms theory of Robert havehypotheses been fatal against to a lesscases comprehensive as data, a distinctly treatment scientific of its Alexy be usefully applied to certain decisions of the subject matter, Wesemann’s arguments survive simply because they lie well within the accepted conventions 2. and norms of a discourse, although it may not be the one she thinks it is. 3. IsEuropean German Court constitutionalism of Justice? useful outside its home What is EU citizenship? Each of the substantive chapters (episodes) provide 4. jurisdiction? This is a sensible strategy and makes the book exceed- inglymaterial useful of moreas a resource or less significancekit for the four to discretethe four subjects stories. TheyIs the are European interlinked Court stories, of Justice as they overreaching should be, although itself? (Alexy’s theory, European constitutionalism, Citizenship, yet all have a value, if not as entertainment then as intel- the European Court of Justice) comprising that material. lectualas we shall stimulus. see only the first is dealt with convincingly, Yet Wesemann does not draw the four plots together in her Conclusion, there focusing entirely to the applicability of Alexy’s theory to the jurisprudence of the European 2. THE STRUCTURE Before spoilers of those stories are ventured, the struc- Indeed, it appears at times she is not entirely sure as to ture of the book as a whole needs review. Traditionally whatCourt herof Justice book andis about, the consequences although she that sometimes flow from hauls that. thesis-like, it has an Introduction with corresponding herself up and explicitly states that it is about “norm- Conclusion, the former with methodology and outline. theoretical analysis of Articles 20 and 21 TFEU governing With creeping scientism a methodology is now required EU citizenship” (p 115). Nevertheless, the place of German of legal academic work even when that work is entirely constitutionalism as jurisprudence and the role of the and obviously within a prevailing discourse. Wesemann European Court of Justice in an inchoate institution are sets out her methodology thus: well and usefully discussed at appropriate points. These discussions seem missed opportunities for worthwhile … this book will follow a doctrinal methodology, analyz- originality claims. ing Treaty norms through the lens of CJEU jurisprudence in order to determine the relevance or norm theory for Alexy’s constitutional rights theory. German jurisprudence the EU framework in the particular given context. Primary is aWesemann’s peculiar, even first headache-inducing, substantive chapter thing critically to the analyses Antipo- and secondary legal sources will be considered alike and dean mind. There is a drive to deprive concepts of content, evaluated in the context of constitutionalism and con- to create a science of law in the way that logicians deal stitutional rights norm theory. Particular consideration with philosophy. That is no criticism, for there may be a will be given to the EU Treaties and the work of Robert point to the structuring of legal reasoning, it is just that it Alexy, as well as the critiques of it produced by Jakab, harks back to the jurisprudence of Julius Stone which dis- Raz and others. (pp 1-2) solves in a tangle of individual mental tropes without ever Doctrinal methodology is a matter practicing within the accepted norms and practices of the discourse of Nevertheless, the chapter convincingly argues that a norm dealing with the ultimate question of, ‘Why this decision?’ law, of working within the law to find out answers to the

Law in Context, Vol 37, Issue 1, 2020 191 ISSN: 1839-4183 can be either a rule or a principle and in the latter form The next chapter again steps back from the tram-tracks is subject to contestation by other norms. The resolution of such contestations should be viewed as measuring the the status of the Court of Justice of the European Union. principles against each other in the circumstances and Thisof the is argument a most illuminating to fill in background, chapter, quite in this intriguing case to explore for its determining which is more important. suggestion that a court may constitute a polity as much The second substantive chapter (Chapter 3) is about as a sovereign, act of revolution or some other congealing the nature of the European Union. A variety of concepts of law-making power. Autopoiesis leading to autochthony of EU constitutional law and legal order, such as par- has a variety of mechanisms in human society as much ticularism, holism and federalism, are discussed with as it has for plants. an ultimate focus on constitutional pluralism. Adopting These four chapters set up the argument with the ma- the pluralist analysis allows Wesemann to apply Alexy’s terial for Chapter 6. Wesemann takes a number of cases approach to EU constitutional norms. The discussion is decided by the Court of Justice of the European Union and a useful critical overview of EU constitutional theory and its development, although federalism is given a somewhat short shrift. Federalism comes in more varieties and with asks whether they fit the reasoning structure set out by more diverse underpinnings than allowed. After all, the outAlexy’s what constitutional that might mean norm for theory. the nature Of course, of citizenship she finds of States of Australia and of the United States in themselves thethat European they do and Union, in Chapter arguments 5 and about the Conclusion the overreach fleshes of the Court, and the applicability of Alexy’s theories outside power—Australia and the United States are confedera- Germany. Wesemann’s methodological confusions play tionsretain rather jurisdictional than federations. specificity and A closer sources analysis of law-making of such constitutional structures might have reduced the seeming that to test a hypothesis is merely to determine whether idiosyncrasy of the EU. itout is falsehere, rather as scientific than to methodology establish its truth. would The have cases it maythat Chapter 4 is about citizenship. To the legal scientist, the content of concepts is irrelevant yet Wesemann surveys the idea in order to make possible the test of Alexy’s theories proveswell fit towithin be a minorAlexy’s point approach because to norm Wesemann application does butnot against the jurisprudence of the European Court of Justice inmay the just end as claimwell fit truth, within merely another applicability approach. However,with what it in relation to the citizenship articles of the European Trea- may be found to follow. ties. Again, it is a useful discussion although somewhat limited by its refusal to engage with the concept as explic- 3. SPOILER ALERT! itly expressed in law. For example, the history of British citizenship is one of subjection rather than participation At the outset of this review, the four stories to be found yet this in itself provides the core formative institution of in this book were set out. Wesemann does bring the whole the British state, allegiance, through the formulation of around to the application of the constitutional rights Lord Coke’s constitutional vision in Calvin’s Case ((1608), norms theory of Robert Alexy to certain decisions of the 77 ER 377, (1608) Co Rep 1a). Subjection and citizen- European Court of Justice. Of course, were the book so ship are formed through allegiance and thus concepts of simple it would not have been necessary to take us through demos fall away. Similarly, in Australia, belonging without Chapters 3, 4 and 5. A theory can come from anywhere; it citizenship has been accepted in Love v Commonwealth, is tested against the data (the cases), and if it works what Thoms v Commonwealth ([2020] HCA 3). The shattering that means can then be deduced. Here it does work and of citizenship into a myriad part in that jurisdiction ren- indeed works well. It provides a useful way of thinking ders illusory the teleology for citizenship annunciated by about rights and makes clear distinctions between rule Wesemann in the chapter. On the other hand, the retreat to application and rights determination. a more doctrinal method, the interrogation of the notion Yet Wesemann has interwoven other stories, perhaps of European citizenship, in this chapter and in Chapter less convincingly but no less intriguing for that. One is 6 is most useful and illuminating, especially as it allows about citizenship of the European Union as something the constitutive effect of determinations of the meaning with a less than determinate content but with generative of citizenship on the EU as a polity. capacity compatible with sovereignty of European states.

Mind you, citizens of confederal nations would not find

192 Law in Context, Vol 37, Issue 1, 2020 ISSN: 1839-4183 this alien. The process of formation of a polity starting 4. CONCLUSION with a court and a jurisdiction rather than territory and Overall, Anne Wesemann, Citizenship in the European force is worth observing. Maybe there is overreach and Union: Constitutionalism, Right and Norms is useful both maybe Alexy’s theory allows a defense; all in all the book descriptively and argumentatively. Descriptively we learn provides us with insight. We are given the arguments, a lot about the European Union and its processes and in- although not the solution as one cannot claim a theory stitutions. Argumentatively, the conclusions are more or to be the law. less relevant outside the European Union. Alexy’s theories Referring to Wesemann’s third theme or story, taking are a useful tool of analysis and European citizenship is as a jurisprudence out of its home jurisdiction and applying it elsewhere is an exercise in comparative law. The whole the argument. I would highly recommend this book for book becomes an exercise of doing just that and as such thecontext-bound non-EU reader as any. as it Comparative introduces us law to awill way benefit of thinking from is excellent. We see the utility of the contentless concept (not just Alexy’s) which is dense, well-read, tightly argued at the same time as exploring the limitations of this form and somewhat myopic. For the EU reader, I dare say it is a of reasoning. Beyond the mere intellectual curiosity as useful contribution to an important debate. I doubt that to what Europeans are up to, it is this exercise which the British will want to know. provides the attraction of the book to the world outside.

Law in Context, Vol 37, Issue 1, 2020 193 Received: December 14 2020, Date of acceptance: December 15 2020, Date of publication: December 23 2020, DOI: https://doi.org/10.26826/law-in-context. v37i1.131

Book Review Can We Still Afford Human Rights? Critical Reflections on Universality, Costs and Proliferation

By Emma Henderson, Director of Graduate Research in the Law School at La Trobe University, Orcid: http://orcid.org/0000-0002-4501-1147 La Trobe University, Melbourne, Australia

ABSTRACT

Book Review: Jan Wouters, Koen Lemmens, Thomas van Poecke and Marie Bourguignon (Eds.) [Leuven Centre for Global Governance Studies]. 2020. Can We Still Afford Human Rights? Critical Reflections on Universality, Costs and Proliferation. Cheltenham, UK: Edward Elgar Publishing, ISBN-13: 978-1789905120, ISBN-10: 1789905125, 369 pp.

Keywords – Book review, human rights, universality, costs, proliferation

Acknowledgements: The Book was provided free of charge by Edward Elgar. Disclosure statement – No potential conflict of interest was reported by the author. License – This work is under Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) https://creativecommons.org/licenses/by-nc-sa/4.0/ Suggested citation: Henderson, E. 2021. “Book Review: Can We Still Afford Human Rights? Critical Reflections on Universality, Costs and Proliferation, edited by J. Wouters, K. Lemmens, T. v. Poecke, M. Bourguingon”, Law in Context, 37(1): 194- 196, DOI: https://doi.org/10.26826/law-in-context.v37i1.131

194 Law in Context, Vol 37, Issue 1; 2020 ISSN: 1839-4183

At the end of a year in which the loudest screams of The drafting of the UDHR in the absence of a common ‘freedom’ and ‘rights’ have seemed increasingly unteth- philosophical framework for the existence of human ered from reality, comes a collection of essays which pose rights, while at the time a triumph of pragmatism and the question: has the mid-20th century experiment in intent, has over the decades led to serious compromises result of a conference held in 2018 to celebrate the 70th that universal acceptance was only possible by omitting birthdayuniversalism of the outlived Universal its usefulness? Declaration This of Human collection, Rights the anand enforcement complications. mechanism. The foundational The schism flaw ofhighlighted the UDHR inis (‘UDHR’), poses thoughtful, albeit not completely reas- the 1940s, between societies keen to guarantee individual suring, answers, to this frightening question. civil and political rights but wary of economic and social rights, and vice versa, has become both more entrenched The editors, all from KU Leuven University in Belgium, have brought together a range of European legal academ- ‘thin’ and ‘thick’ societies (and communities within societies)and complicated. to explain Thethe humaneditors rights use Kaplan’s dilemma. definition Dramatic of variations in implementation pose an existential threat focusesics and practitionerson the compromises to reflect and on complicationsthe central pillars of the of to the sustainability of the universal human rights pro- universalismthe human rights established project. in The the UDHRfirst section and asks of thewhether book ject. This has led to calls for a return to basics – a new the time has come for a new global human rights treaty. minimalist global treaty focused only on the fundamental The middle section considers the proliferation of human core rights to which all societies, both thick and thin, can rights in the 21st century, circling around the issue of a agree, but this time attached to a robust enforcement hierarchy of rights. In a world in which new rights are mechanism. Nick Goetschalckx explores the ‘mythic’ endlessly contested, is there a danger that more ‘impor- qualities of universalism in his framing chapter. In chapter 3, Konstantinos Magliveras takes up this call, arguing for focuses on the costs of human rights—both the dangers a new global treaty overseen by a single treaty body. In inherenttant’ rights in arefailing devalued? to realise The human final rightssection fully, of the and book the the following chapters, Paul Lemmens, Johan Lievens and literal costs of development. The introduction persuasively Nele Verbrugghe complicate the possibilities of a singular makes the case for the interdependence of universalism, treaty as they explore the operation of the ‘margin of proliferation and costs as issues the human rights project appreciation’ in the European context. Could a single, global, treaty allowing for ideological differences between promises, these main themes crop up throughout the book thick and thin societies/communities through a margin rathermust resolve than being if it rigidlyis to flourish. segregated Just and as the this introduction strengthens of appreciation, overseen by a new assembly of State Par- the overall argument. For instance, the main take-away ties, protect the universalism of human rights, or would from the section on universality, that it is time to replace it simply reduce the rights protected while transposing human rights treaty, cannot be divorced from the prob- lemsthe system raised of by issue-specific the proliferation treaties of humanwith a comprehensive rights nor the existingThe second problems section onto of thethe newbook framework? focuses on the problems costs of denying them. Read as a whole, the collection caused by the proliferation of human rights. The UDHR, provides a layered, nuanced and complicated picture of an aspirational statement with no enforcement mecha- the human rights project in the 21st century. nism, has now led to a huge number of separate human rights treaties and optional protocols. While states have Each section within the book follows the same struc- - tion of the treaties has not led to uniformity in terms of theoretical imperatives of the topic, followed by three compliance.adopted these Indeed, treaties this according is one of tothe preference, central arguments ratifica ture; the first chapter sets out an overview and the central in favour of a new singular global treaty with an enforce- problems within that topic. It’s a useful structure, mak- ment mechanism. And in addition to the system of UN ingchapters the content which graduallyaccessible ‘zoom to a broad in’ on audience, specific issues but it oris rights, rights have also proliferated in regional systems undeniably a European-centric book with only two of such as in Europe, Africa and the Americas, meaning that the more narrowly-focused chapters examining issues there are now multiple levels of rights recognition and outside of European jurisprudence. enforcement. As Kasey McCall-Smith notes in chapter 6,

Law in Context, Vol 37, Issue 1, 2020 195 ISSN: 1839-4183 the world is awash with human rights ‘law’. More than this, the world has developed a vernacular to speak about for development. It is now clearly recognised by the UN human rights that transcends the legal forum. Unfor- feature, to being a necessary but not sufficient condition tunately, as McCall-Smith argues, common use has led to delivering basic human rights; in an era of so-called to slippage in the understanding and attempts to apply austerity,that a significant when the part legitimacy of public of raisingexpenditure taxation is devoted is being rights. Rights, intended to act as a protective framework constantly challenged by conservative and populist govern- for state-individual interactions, are now proclaimed ments, the relationship between economics and human rights is ever more central. Isa considers the theories of context. Even worse, the language of rights has become Sen, Nussbaum and UN experts while looking for criteria acasually, weapon frequently of disinformation, and even used flippantly to support outside violence this that would enable a rights-forwarding process of distri- and war. Dalia Palombo demonstrates another form of slippage in chapter 7, describing the resultant adoption none of these theories offer unambiguous solutions to the of isolationism in protests against the transformative bution of scarce resources, only to find that not only do impulses of adjudicative bodies. Treaty bodies which each of these theories give different solutions. Fleshing have taken conservative treaties and, through expansive outconflicts the impact inherent of economicsin development in the projects, human butrights also arena that interpretation, given birth to second and third genera- are chapters by Dina Townsend and Nicky Broeckhoven tion rights, have overseen, Palombo argues, a noticeable (focused on development projects in Ethiopia) and Hillel reduction in compliance in protest. This middle section Sommer (the impact of judicial activism on the right to of the book offers a thought-provoking analysis of the problems posed by the creative adaptation and expansion Yared focuses on gender equality and climate change of rights discourse in a system ultimately enforced purely tohealth argue in thatIsrael). the Incosts the offinal human chapter rights of the are book, massively Dima by state consent. Michelle Meulebrouck, focusing on trade outweighed by the transformative reality of the progress agreements in the EU, and Louise Reyntjesns, examining realised by their realisation. the right to nationality in the EU (and its deprivation in In this way, the book comes to what I would describe the name of national security), flesh out these problems the current articulation of the human rights project, for in specific European contexts. allas aof Churchillian its limitations end and– for failures, all of the what difficulties other optionposed by is of the economic, social and political costs of human rights, or moreThe final accurately, third of the bookcosts is associated devoted to with a consideration a failure to humanity, climate change, racism, technology and cor- realise the human rights project. Felipe Gómez Isa explores ruptionthere? In to thename context but a few,of the the immense idea of the challenges universality facing of the historical shift from human rights development as an as it did at the birth of the United Nations. To the ques- the primary driver of economic growth and more impor- equality first expressed in the UDHR remains as essential tantly,obstacle from to economic economic efficiency, growth being to human the most development important as tion ‘can we still afford human rights?’ the only answer is the rhetorical: How can we not?

196 Law in Context, Vol 37, Issue 1, 2020 Received: July 2 2020, Date of acceptance: October 30 2020, Date of publication: January 4 2021, DOI: https://doi.org/10.26826/law-in-context.v37i1.135

Book Review A Lesser Species of Homicide. Death, Drivers and the Law

By Robert P. Brown, Orcid: https://orcid.org/0000-0002-5773-7375 La Trobe University, Melbourne, Australia

ABSTRACT

Book Review: Kerry King. 2020. A Lesser Species of Homicide. Death, Drivers and the Law. (1st ed), Perth: University of Western Australia Publishing, ISBN: 978-1-76080-002-4 2020, 464pp.

Keywords – Book review, criminology, homicide, drivers.

Acknowledgements: The Book was provided free of charge by University of Western Australia Publishing,. Disclosure statement – No potential conflict of interest was reported by the author. License – This work is under Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) https://creativecommons.org/licenses/by-nc-sa/4.0/ Suggested citation: Brown, R.P. 2020. “Book Review: A Lesser Species of Homicide. Death, Drivers and the Law, by Kerry King”, Law in Context, 37(1): 197- 198, DOI: https://doi.org/10.26826/law-in-context.v37i1.135

Law in Context, Vol 37, Issue 1, 2020 197 ISSN: 1839-4183

This work boldly examines decades of socio-legal reduced sanctions that might otherwise be applied. She history and re-situates developments in charging and sentencing practices in the broader societal context. Dr conduct as deserving a more severe punishment when Kerry King systematically unpacks the trends around certainidentifies aggravating as particularly factors problematic are present: the stealing tendency a car, to find for road deaths in Western Australia from 1946-2018, acting - as a witness to evolutions in Australian car culture and able, these distinctions are used to create dichotomies socio-legal norms of driver responsibility. The central thatinstance, result or in drink similar driving. conduct While being on first treated glance differently reason question the work seeks to answer is this: when death is by an order of magnitude. King points to a concern that occasioned by the use of a motor vehicle, is it treated as a lesser species of homicide injured by this framework that gives special protection friends and family overlooked in charging and sentenc- tooften-marginalised the ‘momentarily members careless’ of‘ordinary society findperson’ themselves but de- ing decisions, or are such ?concerns Are the concerns driven by of a misplacedaggrieved monises a criminal ‘other’ whose behaviour is deemed worthy of extraordinary punishment while both groups the loss of human life, rather than the culpability of the cause a similar amount of harm to society. sense “that the punishment should reflect the tragedy of There are not many criticisms to be made of this A Lesser Species of Homicide challenges the orthodox superb work. Firstly, a more in-depth consideration of narrativesconduct”? (p.of road 7) safety propounded by some members arguments of opponents to legal reform may have been of the public, of the legal profession and of Parliament in order to shine a light on the conduct of drivers and their and at a certain length, her dismissal of their arguments legal reverberations. Discourses around road safety often maywarranted. have been While further their words reinforced are reported by a more with in-depth fidelity focus in on the idea of “vision zero” – that is, designing consideration of their points. Secondly, A Lesser Species of Homicide safe to fail (crash) by changing dangerous network design the, admittedly mostly recent, developments in the domain (e.g.the road blackspot network intersections), through scientific modifying principles road rulesto make (e.g. it of roadside drugmay alsotesting have and benefited the controversy from a discussion over extent of reducing speed limits) or through targeted behavioural 1 interventions (typically advertising campaigns). More King’s insight into the contested policy area would have generally, our conceptions of driving and the open road beento which highly such illuminating. tests truly reflect drivers’ impairment. are often romanticised and linked to broader ideals of Overall, in the opinion of the reviewer, A Lesser Species freedom and liberty. King begins with an introspective of Homicide is an extensive and detailed work that provides great insight into the social, cultural and legal history of her vision obscured by cultural touchstones and the no- road deaths and should be read. King’s arrangement and tionanalysis of an of ‘accident.’ her own relationshipAccidents and to roadthe road, deaths and always finds presentation of history is orderly and interesting – keep involve conscious human decisions and conceiving of in mind, however, that there is an extensive and detailed them of accidents is, in King’s view, an often incorrect judgement as to the quality of that decision-making and from the work in any way, for it is essential to its purpose, the conduct that results. butcoverage it can of have many an horrific emotional crashes. impact. This Finally, does not you subtract should What sets this book apart is the extensive empirical A Lesser Species of Homicide, research into legislative developments and judicial deci- and to immediately have at hand a program for reform. sions. In chronologically organised chapters, she eluci- not expect to finish reading dates her perspective on the culpability of drivers and the from many reforms attempted in the past – instead only deaths that happen in their wake through evaluating and byKing “cementing identifies the understandings missing or counterproductive that the mismanagement outcomes contextualising decades of cases. Her conclusions are so of a vehicle is as dangerous or negligent as discharging incisive that they seem inescapable: how could drivers, prosecutors, judges and legislators act in the way they of aggravating factors and by completing a “wholesale did, seemingly disregarding the reckless or dangerous revisiona firearm of in penalties a public…” across place the even entire without spectrum the presence of driv- nature of conduct behind the wheel and applying much ing conduct…” can we envisage real change (pp. 272-3).

1 Australian Broadcasting Corporation, ‘Retired Magistrate David Heilpern critical of NSW drug driving laws’, The Law Report (podcast) (accessed 26 June 2020); Aidan Ricketts, ‘Roadside drug testing: Incoher- ent policy or uncertainty by design?’, (2018) 25 Alternative Law Review 1: 30-34.

198 Law in Context, Vol 37, Issue 1, 2020