Ensuring Data and AI Work for People and Society Foreword

Home , Ada Lovelace, Alan Turing

i The Ada Lovelace Institute Ensuring data and AI work for people society Ensuring data and AI work for people and society Foreword

The Ada Lovelace Institute (Ada) has been Engine has no pretensions whatever established by the Nuffield Foundation with to originate anything. It can do whatever a £5m commitment, and with the support we know how to order it to perform.’ of founding partners. Much later, this influenced Alan Turing The Nuffield Foundation is an independent to intervene with the notion of a ‘Turing charitable trust with a mission to Test’ – the test of a machine’s ability to advance educational opportunity and exhibit intelligent behaviour indistinguishable ii social well-being across the UK. It has from that of a human. Ada Lovelace Institute long been at the forefront of addressing the ethical questions raised by scientific This debate between perspectives remains advancements. In 1991, the Foundation a live part of the discussion about the established the Nuffield Council on ethical and social implications of artificial Bioethics, which has been influential in intelligence – the extent to which technology establishing ethical frameworks for policy is able to, or should, replace or work in and regulation relating to innovations in partnership with, human intelligence. biology and medicine. We are very grateful to the Trustees and Ada has been created with the support staff of the Nuffield Foundation, particularly of founding partners: The Alan Turing Imogen Parker who has led the development Institute, the Royal Society, the British of the Institute, for their enthusiasm and Academy, the Royal Statistical Society, support, and to the colleagues from our Wellcome Trust, Luminate, techUK and the partner organisations who formed the initial Nuffield Council on Bioethics. Ada is named Advisory Board of the embryonic Institute: after the mathematician Ada Lovelace Dame Colette Bowe, Julie Maxton, Hetan (1815–1852), who wrote in visionary Shah, Genevra Richardson, Martin Tisne, terms about computing and its application Dan O’Connor, Pete Mills, Helen Margetts in the wider world. and Antony Walker. We would also like to acknowledge the large number of individuals Ada Lovelace is famous for some of the working in this field for their time and earliest ever written reflections on artificial generosity in contributing to this project. intelligence, arguing of the computer in a statement that was famously framed as Alan Wilson the ‘Lovelace objection’ that: ‘The Analytical Executive Chair, Ada Lovelace Institute

A new, a vast, and a powerful language is developed for the future use of analysis, in which to yield its truths so that these may become of more speedy and accurate practical application for the purposes of mankind than the means hitherto in our possession have rendered possible. Ada Lovelace Contents 2 The Ada Lovelace Institute 3

The context: the data-driven AI revolution 1 Ensuring data and AI work for people society 4 The agenda: impacts and ethics 6 Ways of working 8 Work with us The Ada Lovelace Institute

The Ada Lovelace Institute is an on different groups in society. It will guide independent research and deliberative body ethical practice in the development and with a mission to ensure data and AI work deployment of these technologies, and will for people and society. Ada will promote undertake research and long-term thinking informed public understanding of the to lay the foundations for a data-driven impact of AI and data-driven technologies society with well-being at its core.

2 Ada has three core aims: Ada Lovelace Institute

1 2 3

Build evidence Convene diverse Define and inform and foster rigorous voices to create good practice research and debate a shared understanding in the design and on how data and of the ethical issues deployment of AI affect people arising from data data and AI. and society. and AI.

In undertaking its work, Ada will:

• Be outward-facing and collaborative, acting independently of vested interests, and transparent about relationships and funding.

• Be at the forefront of developing change needed to improve people’s lives, through a focus on: rights-based approaches; establishing norms; influencing professional practice; technological innovation; regulation and the law; and public dialogue.

• Recognise the potential value of data, algorithms, and AI for individual and social well-being, taking account of human capacity to adapt and respond to new technological challenges.

• Combine reflective deliberation and rigorous research with the need to respond to a rapidly evolving social, technological and economic context.

• Work with other organisations to situate its work in a global context. The context: the data-driven AI revolution

We are living through the fourth industrial the internet – can read, see, hear, translate revolution, driven by computing power languages and can use data to make and information flows across the internet. decisions, often better than humans. This generates and stores huge volumes These decisions are made on rules-based of data, both on and for, individuals and processes that can be automated. organisations. The analytics’ tools of data science and AI transform both the When human responsibility for 3 way we live and business processes in all decisions is still needed, AI can be Ensuring data and AI work for people society kinds of organisations – public and private. thought of as ‘augmented intelligence’. Some organisations are leaders in these Machines can’t yet pass the Turing developments, some have as yet barely test and ‘think’ but they can ‘learn’. engaged. Ultimately, these transformations Health care and many other areas of will be all-embracing. No social policy or service delivery will be revolutionised analysis of the conditions for future social by learning machines: receiving and well-being can be put forward without linking data, and providing inputs now acknowledging the fundamental for machine learning algorithms. and inevitably disruptive impacts that For example, machines can generate AI will bring. precision medicine and rapid medical diagnosis. Treatment plans can then The tools of AI generate new data be evaluated and ‘fed back’ through and deploy existing data in a number the system thus creating the learning of different ways. The data provides machine. This kind of application the basis for new analytics and computer of learning machines is likely modelling. ‘Machines’ – computers to become more prevalent or networks of computers linked by in a range of different contexts. The agenda: impacts and ethics

Ada’s first objective is to chart the impacts and the sharing economy, and security. of these technological developments on Combined, these benefits create smarter people and society. Since these are rooted urban (and rural) environments. in the data-driven AI economy, analysis will begin there. We will seek to understand the Individual benefits sit alongside social full capabilities of the technologies – for consequences. The impact of digital the present and, through horizon scanning, reforms in the justice system, for example, for their future potential. There are benefits raises concerns about the rights of the 4 to individuals directly – for example individual in relation to the state; and Uber, Ada Lovelace Institute through phone apps such as City Mapper though beneficial to customers, has led or direct connections to taxi services such to controversy relating to the workforce. as Uber – and in the delivery of a wide Individual autonomy is simultaneously range of services, for example in health, enhanced and challenged. education, justice, financial services, retail

Ethical challenges These developments therefore inherently or, because of their training data, can be generate ethical challenges. The data and biased. Are the outcomes fair, and how can the algorithms have to be accurate and we test that? Are lines of what is acceptable reliable and we have to be aware of the being crossed? For example, while consequences of failings in this respect. marketing and electioneering were relatively Many benefits can only be delivered open, their methods, however contested, effectively through use of personal data, were not seen as a threat to the democratic with analytical power enhanced by linking process itself. But some contemporary of data from a variety of sources. This methods are now clearly unacceptable – raises issues of privacy and consent. because of their deliberately covert nature, Machine learning-constructed decisions, directing ‘fake’ news and information especially those based on ‘deep learning’ at targeted audiences to undermine the algorithms, are typically not transparent; basis of shared public debate. Understanding the impacts of innovation The speed at which these technologies data. In bringing together the disciplines change society, and the dynamism of the of social science and the humanities with markets they fuel, mean that neither the those of data and medical science and research communities, nor the designers technical innovators, Ada’s work will offer or developers have the space clearly to a deeper understanding of the impacts of articulate the conditions for a successful technological innovation as they play out AI-driven society. There is no shared across the very different strata and cultures discourse on how we can begin to measure of our increasingly complex, connected, and 5 or even account for the social value of fragmented society. Ensuring data and AI work for people society

Articulating the costs and benefits We need to identify which of these fairness and accountability all potentially issues are capable of technical have different meanings for different resolution – transparency for example – groups of people and different cultures; and which raise social and political they all have value connotations. issues that are more challenging. Indeed, insofar as politics is about Understanding and contributing the relativities of power, we need to the ethics of data and AI involves to explore the networks of control unpicking these issues and sharpening and influence in the collection and our presentation of the vocabulary use of data, and in the exploitation in multiple dimensions. Whose values? of data in AI. We need to be able Can we set rules through establishing to articulate the costs and benefits certain norms and principles, or are of these developments for different we forced into thinking through groups – the distributive consequences – trade-offs in formulating best notably taking into account issues of practice? All of these questions gender and diversity. Concepts such are priorities for Ada – through as explainability, privacy, consent, bias, research and informed debate. Ways of working

The agenda outlined here is substantial in we will seek to distinguish the challenges all its sensitivities and nuances: charting that are essentially technical from those the impact of data and AI on people and that need social science and humanities society; research and debate on ethics; lenses. The complexity of the research and informing best practice. Ada, both agenda implies that progress will be most of necessity and as a matter of principle, rapid through the exploration of use cases. will work collaboratively – with the range To develop our understanding of the almost- of relevant research communities along ubiquitous impact of data and AI on people 6 the spectrum from data science to social and society, we will review aspects of Ada Lovelace Institute science; with organisations deploying how people live – housing, work and data and AI in the public and private incomes, education, health, retail and sectors; and with different publics and a wide range of services – along with the their representatives in an increasingly organisations delivering the data-driven diverse society. AI innovations across these dimensions. At a broader scale, we will seek to articulate The range of research topics is extensive societal implications. and essentially multidisciplinary. As noted,

A collaborative approach We are conscious that Ada will function of the impact and ethics of data and AI within a complex landscape of exploration will be distinctive, and we will ensure that and research in these areas, including we collaborate widely, communicating the Centre for Data Ethics and Innovation and learning in both directions. We are established by the Department for Digital, independent, connected to the research Culture, Media and Sport. We believe our frontiers, and hope to provide thought focus of ‘people and society’ in the context leadership for the longer term. 7 Ensuring data and AI work for people society

These are large challenges that no We will work towards the articulation one organisation or approach can solve of best practice in all sectors through alone. Ada has been established to our research on a variety of use cases. research, to convene, and to collaborate. To take practical steps in establishing We will succeed if we can throw light its work programme Ada will work on some of the key questions generated in the following ways: by our preliminary analysis:

• Undertake, commission and catalyse 1. How should society equitably distribute research projects and methodological the power and benefits of data and AI innovation. We will produce working while mitigating harm? papers and contribute to journals and the media debate. We will work with 2. In whose interests, and to what purpose, different disciplines, and seek expertise should information be accessed or used? in their related methods and domains, to build a shared vocabulary of issues. 3. What rights and protections should citizens have and how can they • Provide synthesis, analysis and be realised? communications to inform and engage in public debate on data use and its 4. How should we protect vulnerable, impacts, and to make the uses of AI and marginalised or disadvantaged groups its impacts more visible. We will hold in society? events and support fellowships to bridge the gap between academic and technical 5. How can we ensure those who have expertise, public values and practical power to shape society act with public action. We will initiate working groups legitimacy? around ethical ‘test cases’. As these questions show, our agenda • Bring together the perspectives of is not only substantial but also urgent. industry, government and civil society The issues raised on impact, ethics and to understand the complex systems best practice are alive now, and connecting interacting with data and AI and identify research communities to the public incentives and barriers to ethical practice. interest is a high priority. Work with us

We are keen to hear from organisations and individuals who would like to engage with the work of the Ada Lovelace Institute.

You can contact us at [email protected], and follow us on twitter @AdaLovelaceInst. 8 Ada Lovelace Institute For more information on the development of Ada so far, visit our website and sign up to our mailing list at www.adalovelaceinstitute.org.

In addition to establishing the Ada Lovelace Institute, the Nuffield Foundation will continue to fund research, analysis and pilots in the field of digital society and data ethics. For more information about research funding from the Nuffield Foundation, go to www.nuffieldfoundation.org/research. About the Nuffield Foundation

The Nuffield Foundation funds research, Our student programmes – analysis, and student programmes Nuffield Research Placements and that advance educational opportunity Q-Step – provide opportunities for and social well-being across the young people, particularly those from United Kingdom. disadvantaged backgrounds, to develop their skills and confidence in quantitative We want to improve people’s lives, and and scientific methods. their ability to participate in society, by understanding the social and economic We are the founder and co-funder of factors that affect their chances in life. the Nuffield Council on Bioethics, which 9 The research we fund aims to improve examines and reports on ethical issues Ensuring data and AI work for people society the design and operation of social in biology and medicine. We have recently policy, particularly in Education, Welfare, established the Ada Lovelace Institute, and Justice. an independent research and deliberative body with a mission to ensure data and We have recently established the Nuffield AI work for people and society. Family Justice Observatory to support the best possible decisions for children by We are a financially and politically improving the use of data and research independent charitable trust established evidence in the family justice system in 1943 by William Morris, Lord Nuffield, in England and Wales. the founder of Morris Motors. 10 Ada Lovelace Institute

T: 020 7631 0566 Registered charity 206601

www.nuffieldfoundation.org www.adalovelaceInstitute.org

@NuffieldFound | @AdaLovelaceInst