Algorithmic Mechanism Design for Social Good

AI MATTERS, VOLUME 4, ISSUE 3 4(3) 2018

Mechanism Design for Social Good Rediet Abebe (Cornell University; [email protected]) Kira Goldner (University of Washington; [email protected]) DOI: 10.1145/3284751.3284761

Introduction fecting many individuals’ livelihoods. This research area falls at the interface of artifi- Across various domains—such as health, ed- cial intelligence, theoretical computer science, ucation, and housing—improving societal wel- and the social sciences. Since the fall of fare involves allocating resources, setting poli- 2016, the authors of this piece have been co- cies, targeting interventions, and regulating organizing the Mechanism Design for Social activities. These solutions have an immense Good research group, workshop series, and impact on the day-to-day lives of individuals, colloquium series Abebe and Goldner[2016, whether in the form of access to quality health- 2018]. The group comprises a large net- care, labor market outcomes, or how votes are work of researchers from various disciplines, accounted for in a democratic society. Prob- including computer science, economics, soci- lems that can have an outsized impact on ology, operations research, and public policy. individuals whose opportunities have histori- Members of the group partner with domain ex- cally been limited often pose conceptual and perts in non-government organizations, think technical challenges, requiring insights from tanks, companies, and other entities with a many disciplines. Conversely, the lack of inter- shared mission. The mission is to explore new disciplinary approach can leave these urgent frontiers, garner interest in directions in which needs unaddressed and can even exacerbate algorithmic and mechanism design insights underlying socioeconomic inequalities. have been under-utilized but have the po- To realize the opportunities in these domains, tential to inform innovative interventions, and we need to correctly set objectives and reason highlight exemplary work. about human behavior and actions. Doing so In this piece, we discuss three exciting re- requires a deep grounding in the field of in- search avenues within MD4SG. For each of terest and collaboration with domain experts these, we showcase ongoing work, underline who understand the societal implications and new directions, and discuss potential for im- feasibility of proposed solutions. These in- plementing existing work in practice. sights can play an instrumental role in propos- ing algorithmically-informed policies. In many cases, the input data for our algorithms may Access to Opportunity in the be generated by strategic and self-interested Developing World individuals who have a stake in the outcome of the algorithm. To get around this issue, we New technologies and data sources are fre- can deploy techniques from mechanism de- quently leveraged to understand, evaluate, sign, which uses game theory to align incen- and address societal concerns across the tives or analyze the strategic behavior of indi- world. In many developing nations, however, viduals who interact with the algorithms. there is a lack of information regarding underlying matters—whether that be the preva- The Mechanism Design for Social Good lence of diseases or accurate measurements (MD4SG) research agenda is to address prob- of economic welfare and poverty—due to the lems for which insights from algorithms, op- unavailability of high-quality, comprehensive, timization, and mechanism design have the and reliable dataUN[2014]. This limits the im- potential to improve access to opportunity. plementation of effective policies and interven- These include allocating affordable housing tions. An emerging solution, which has been services, designing efficient health insurance successfully demonstrated by the Information markets, setting subsidies to alleviate eco- Communication Technology for Development nomic inequality, and several other issues af- (ICT4D) research community, has been to Copyright c 2018 by the author(s). take advantage of high phone and Internet

27 AI MATTERS, VOLUME 4, ISSUE 3 4(3) 2018 penetration rates across developing nations to been realized through this system. design new technologies which enable collec- tion and sharing of high-quality data. There Availability of new technologies also presents has also been recent work from within the AI opportunities to tackle fundamental problems community to use new data sources to close related to poverty. Advances in last-mile pay- this information gap Abebe et al.[2018], Jean ment technologies, for example, enable large- et al.[2016]. Such AI-driven approaches sur- scale, secure cash transfers. GiveDirectly face new algorithmic, modeling, and mecha- leverages this and the popularity of mobile nism design questions to improve the lives of money across the world to create a system many under-served individuals. where donors can directly transfer cash to re- cipientsGD, Blattman and Niehaus[2014]. A prominent example is in agriculture, which GiveDirectly moves the decision about how to accounts for a large portion of the econ- use aid from policy-makers to recipients, giv- omy in many developing nations. Here, vi- ing recipients maximum flexibility. Such aid ral disease attacks on crops is a leading generates heterogeneity in outcomes—e.g., cause of food insecurity and poverty. Tradi- families may use aid to start a business, pay tional disease surveillance methods fail to pro- rent, cover health costs, and so on. Policy- vide adequate information to curtail the im- makers used to prioritizing specific outcomes pact of diseases Mwebaze and Biehl[2016], may be uncomfortable by such a model. A re- Mwebaze and Owomugisha[2016], Quinn search question then is: can we predict how a et al.[2011]. The Cassava Adhoc Surveil- given population will use aid? Likewise, how lance Project from Makerere University imple- can we target people for whom the interven- ments crowd-sourcing surveillance using pic- tions will make the largest difference? Aid has tures taken by mobile phones in order to ad- historically been targeted on the basis of find- dress this gap Mutembesa et al.[2018]. The ing the most deprived people. The ability to tool is set up as a game between farmers and model heterogeneous treatment effects opens other collaborators, and aims to collect truth- the door for designing more nuanced mecha- ful, high-value data (e.g., data from hard-to- nisms that fairly and efficiently allocate subsi- reach locations). This approach underlines dies in order to maximize a desired outcome. interesting challenges, such as how to opti- mally incentivize individuals to collect high- Problems in the developing world surface quality information and how to augment this unique challenges at the intersection of AI, information with existing methods. Similar ICT4D, and development economics. So- issues arise in other domains—e.g., in citi- lutions often have to be implemented in zen science and in computational sustainabil- resource-constrained environments (e.g., over ity Xue et al.[2016a,b]. Finding solutions in feature phones or with low network connectiv- the context of the developing world may there- ity) Brunette et al.[2013], Patel et al.[2010]. fore have a broader global impact. Key populations of interest (e.g., women, people living in rural parts, individuals with dis- Lack of information also leads to inefficien- abilities) may not be easily accessible Sultana cies in existing systems, presenting a possi- et al.[2018], Vashistha et al.[2015b,a]. In- bility to introduce solutions that abide by exist- dividuals may have low-literacy Sambasivan ing cultural and technological constraints. For et al.[2010]. Lack of understanding of socio- instance, large price discrepancies and major cultural norms and politics, furthermore, may arbitrage opportunities present in markets for inhibit proposed interventions Vashistha et al. agricultural products in Uganda suggest large [2018]. All of these highlight the need for market inefficiencies Ssekibuule et al.[2013]. a multi-stakeholder approach that leverages To alleviate this, Newman et al.[2018] intro- technological advances, innovative technical duce Kudu—a mobile technology that func- solutions, and partnerships with individuals tions over feature phones via SMS service. and organizations that will be impacted by the Kudu facilitates transactions between farmers solutions. MD4SG fosters one such environ- in rural areas and buyers at markets in cities ment in which insights from across these dis- by allowing sellers and buyers to post their ciplines inform the design of algorithms and asks and offers. Kudu has been adapted by mechanisms to improve the lives of individu- users across Uganda and many trades have als across the world.

28 AI MATTERS, VOLUME 4, ISSUE 3 4(3) 2018

Labor, Platforms, and Discrimination The aim is to choose hiring policies that will mitigate discrimination against protected cat- Online platforms are ubiquitous, providing a egories. Hu and Chen[2017] highlight addi- vast playground for algorithm design and arti- tional complexity that arises in dynamic set- ficial intelligence. Every policy decision, how- tings where workers are hired based on invest- ever, impacts and interacts with the platform’s ment decisions (e.g. college GPA) in an ini- strategic users. In this section, we will focus tial temporary labor market (e.g. internships) on online labor markets and how discrimina- and this job creates a worker’s initial produc- tion effects stem from a platform’s decisions. tivity reputation that is then used in the per- Past work begins to investigate some aspects manent labor market. Many of these findings of platforms, of strategic agents, and of dis- also discuss “trade-offs” between group-blind crimination in labor markets, but there are still and group-aware policies. major opportunities for work at the intersection, and insights from mechanism design and Another aspect of labor markets is that a AI are ripe for the job. worker may have the ability to pay to change a feature of her application in some illegitimate One central issue surrounding labor markets or unfair way in order to improve her outcome is that of hiring, in which a firm takes infor- in the labor market. Hardt et al.[2016] exam- mation about a potential candidate and makes ine this problem from a robust machine learn- an employment decision. Firms act as classi- ing perspective. Under certain assumptions of fiers, labeling each applicant as “hire” or “not the cost required to change an applicant’s rep- hire” based on an applicant’s “features,” such utation, they characterize classifiers that opti- as educational investment or a worker’s pro- mally compare to the original reputation (be- ductivity reputation. In the process of making fore the applicant modified it). hiring decisions, however, the firm may poten- tially make discriminatory decisions—perhaps These are only two aspects at the interplay by using protected attributes, or by not cor- between hiring and strategic agents; hiring, recting for differences in applications that stem furthermore, is only one aspect of the labor from systemic discrimination Bertrand and market. Consider today’s popular online la- Mullainathan[2004], Marlowe et al.[1996]. bor markets, such as Mechanical Turk, Up- Bias in hiring decisions may arise due to im- work, Task Rabbit, and Lyft, in which the plat- plicit human bias or algorithmic bias, in which form’s goal is to match workers to employers algorithms replicate human and/or historic dis- or jobs. In these labor markets, the platform’s crimination that is reflected in the data on decisions, even at a granular level, have a which they are trained Broussard[2018], Eu- large impact on the workers and firms. Con- banks[2018], Noble[2018], O’Neil[2016]. sider the following platform decisions. Visi- bility: How many firms can workers see at a One recent line of work investigates hiring time? What capacity do they have to search policies that achieve diversity or statistical par- job offers? Can workers see jobs and jobs see ity (with respect to certain groups) among the workers? Initiation: Which side (or both) can hired workers, and how workers make their submit applications? Initiate messaging? Set investment decisions (e.g. whether to at- contract terms? Information: What informa- tend college) based on the hiring policies in tion is displayed about parties on the opposite place. Coate and Loury[1993], Fryer Jr and side? Name? Photo? Ethnicity? Wage his- Loury[2013], Hu and Chen[2017] study set- tory? Reputation? tings where there is some known underlying bias or historical discrimination against certain Each of these decisions impacts the groups; the aim is to characterize hiring poli- outcome—not only the quality of the match, cies that are optimal-subject-to-fair-hiring, and but also whether (and how much) discrim- to quantify any loss in efficiency compared ination occurs. In a recent paper, Levy to optimal-but-discriminatory policies. These and Barocas[2017] outline categories of works explore two settings: first, when hiring platform decisions which may mitigate or decisions must be “group-blind,” that is, they perpetuate discrimination in labor markets, cannot take group membership into account, including the high-level categories of setting and second, when they are “group-aware”. platform discrimination policies or norms,

29 AI MATTERS, VOLUME 4, ISSUE 3 4(3) 2018 structuring information and interactions, and umented. By compiling the first ever evictions monitoring/evaluating discriminatory conduct. database, Desmond reveals that there is an estimate of 2.3 million evictions in 2016 alone In offline labor markets, it may be challeng- and argues that eviction is a cause of poverty ing or infeasible to collect data to understand EL[2018]. Using this database, and other the nature and extent of discrimination. On- similar datasets, we may be able to employ a line labor markets, on the other hand, yield combination of machine learning and statisti- rich data about employer-employee interac- cal techniques to gain a better understanding tions and present the possibility of conduct- of what causes housing instability and home- ing experiments aimed at reducing bias and lessness. We can then build on this work to discrimination or other desired societal objec- design algorithms and mechanisms that can tives. For instance, Barach and Horton[2017] improve on allocation policies. For instance, look at the impact on hiring of hiding workers’ Kube et al.[2018] use counter-factual predic- wage history. Horton and Johari[2015], Hor- tions to improve homelessness service provi- ton[2018] look at the impact of trying to elicit sions. By doing so, they realize some gains on additional information (features) from workers reducing the number of families experiencing or firms, and the impact of this strategically- repeated episodes of homelessness. At the reported information on hiring. Horton[2017] same time, Eubanks[2018] emphasizes that examines who the hired worker population is caution must be taken when using automated when a minimum wage is imposed on one decision-making tools for allocating limited re- platform. Each of these provide insights into sources in such high-stakes scenarios. Such labor dynamics that may inform platform de- tools may be used to reduce failure rates by sign and interventions. caseworkers, but, if not approached with care, Online labor markets provide a rich play- can deepen already existing inequalities. Fur- ground for techniques from algorithms, AI, and thermore, the use of such tools alone is lim- mechanism design to study how each aspect ited; it does not address the lack of housing of platform design impacts discriminatory ef- and homelessness resources or eliminate hu- fects, workers’ actions, and the desired objec- man biases or discrimination. It is crucial to tive for the platform. take advantage of the confluence of insights from cross many disciplines in order to serve the needs of such vulnerable populations. Allocating Housing and Homelessness Resources An issue that is growing in prominence in housing contexts is that of information. Little Allocation of resources—such as public hous- is documented about how landlords or housing, housing vouchers, and homelessness ing authorities screen applications and make services—has a long history in the economics decisions. One exception is the work of Am- and computation literature. Even simple-to- brose and Diop[2016], which shows that there state problems here have given rise to chal- is increased restriction in access to rental lenging research questions, many of which housing since landlords mitigate information are still open. Increased scarcity of hous- asymmetry by investing in screening tenants. ing resources, growing need for services, and With the increased use and availability of data the use of algorithmic decision-making tools about individuals, it is of paramount impor- all open up several avenues with major op- tance to understand the role of information in portunities for reforming policies and regula- the decision-making process of entities, such tions. Here, we discuss some foundational as landlords or housing agencies, who have work, new challenges, and opportunities that enormous discretion in how and whether fam- emerge at the nexus of algorithm and mecha- ilies are housed. nism design, AI, and the social sciences. Although the introduction of automated tools Millions of individuals across the US have introduces acute challenges related to hous- been evicted or are at risk of experiencing ing, the use of algorithmic techniques dates eviction every year. In groundbreaking work, back several decades and there are many fun- Desmond[2012, 2016] shows that eviction is damental problems that remain unsolved. An much more common than was previously doc- early work here is that of Hylland and Zeck-

30 AI MATTERS, VOLUME 4, ISSUE 3 4(3) 2018 hauser[1979], which considers the “house al- the social sciences, be used to improve ac- location problem” of assigning each individ- cess to opportunity, especially for communi- ual to one item, such as a house. They in- ties of individuals for whom opportunities have troduce a mechanism which satisfies natu- historically been limited. In this piece, we have ral efficiency and fairness notions but is not highlighted MD4SG research avenues related incentive-compatible. That is, individuals may to issues in developing nations, labor markets, be able to improve their outcome by misreport- and housing. For each of these, we have dis- ing their true preferences. Since then, several cussed the need to work in close partnership mechanisms have been proposed, including with a wide range of stakeholders to set objec- the popular Randomized Serial Dictatorship tives that best address the needs of individ- (RSD) mechanism, which uses a random lot- uals and propose feasible solutions with de- tery Abdulkadiroglu˘ and Sonmez¨ [1998]. This sired societal outcomes. There are numerous mechanism is used as a standard mechanism other domains in which this kind of interdisci- in many domains, including housing. While it plinary approach for designing algorithms and is incentive-compatible, it fails to satisfy the mechanisms can improve the lives of many fairness criteria of Hylland and Zeckhauser individuals; we invite readers to learn more [1979]. An important question is then the de- through our colloquium and workshop series. sign of incentive-compatible, fair, and efficient mechanisms for the house allocation problem. Acknowledgments Due to increased scarcity of resources, cur- rent allocation protocols often involves wait- We are indebted to members of the Mech- ing lists and priority groups. Policy constraints anism Design for Social Good research make wait-list design a dynamic rationing group–Ellora Derenoncourt, Alon Eden, Lily problem rather than the static assignment Hu, Manish Raghavan, Sam Taggart, Daniel problem discussed above. Dynamic mecha- Waldinger, and Matt Weinberg—our co- nisms present several technical and practical organizer Irene Lo, and our advisors Anna challenges; e.g., incentive-compatibility may Karlin and Jon Kleinberg for their generos- be infeasible in dynamic settings due to wait- ity in sharing their knowledge and support ing time trade-offs for applicants. There are throughout the past two years. We addition- consequential design decisions related to how ally thank Mutembesa Daniel, Paul Niehaus, to manage wait-lists and different metropolitan Fabian Okeke, and Aditya Vashistha for help- areas have different policies (e.g., setting pri- ful discussions and pointers. We are grate- ority groups, conditions under which individu- ful, as ever, for the numerous researchers als are removed from the waiting list, set of in the economics and computation research choices, and many others). Each of these poli- communities for their enthusiastic support and cies impacts the allocation dynamics, waiting guidance throughout the development of the time, and quality of matches. Recent work group and workshop series. has studied how to design mechanisms sat- isfying various desiderata and quantify differ- References ences in quality of matches across various mechanisms Arnosti and Shi[2018], Leshno Givedirectly: Send money directly to the ex- [2017], Thakral[2016], Waldinger[2017]. treme poor. https://givedirectly. org/. A world that counts: Mobilising the data rev- Conclusion olution for sustainable development. Tech- nical report, November 2014. Report pre- As the use of algorithmic and AI techniques pared at the request of the United Nations becomes more pervasive, there is a growing Secretary-General by the Independent Ex- appreciation of the fact that the most impactful pert Advisory Group on Data Revolution for solutions often fall at the interface of various Sustainable Development. disciplines. The Mechanism Design for Social Good research agenda is to foster an envi- The Eviction Lab. https://evictionlab. ronment in which insights from algorithms and org/, 2018. mechanism design can, in conjunction with Atila Abdulkadiroglu˘ and Tayfun Sonmez.¨

31 AI MATTERS, VOLUME 4, ISSUE 3 4(3) 2018

Random serial dictatorship and the core Matthew Desmond. Eviction and the repro- from random endowments in house alloca- duction of urban poverty. American Journal tion problems. Econometrica, 66(3):689– of Sociology, 118(1):88–133, 2012. 701, 1998. Matthew Desmond. Evicted: Poverty and Rediet Abebe and Kira Goldner. Mechanism profit in the American city. Broadway Books, Design for Social Good. http://md4sg. 2016. com, 2016. Virginia Eubanks. Automating Inequality: How High-tech Tools Profile, Police, and Punish Rediet Abebe and Kira Goldner. A report on the Poor. St. Martin’s Press, 2018. the workshop on mechanism design for social good. ACM SIGecom Exchanges, 16 Roland G Fryer Jr and Glenn C Loury. Valuing (2):2–11, 2018. diversity. Journal of Political Economy, 121 (4):747–774, 2013. Rediet Abebe, Shawndra Hill, Jennifer Wort- man Vaughan, Peter M Small, and H An- Moritz Hardt, Nimrod Megiddo, Christos Pa- drew Schwartz. Using search queries to un- padimitriou, and Mary Wootters. Strate- derstand health information needs in africa. gic classification. In Proceedings of the arXiv preprint arXiv:1806.05740, 2018. 2016 ACM conference on innovations in theoretical computer science, pages 111–122. Brent W Ambrose and Moussa Diop. Informa- ACM, 2016. tion asymmetry, regulations, and equilibrium John J Horton. Price floors and employer pref- outcomes: Theory and evidence from the erences: Evidence from a minimum wage housing rental market. 2016. experiment. 2017. Nick Arnosti and Peng Shi. Design of lotteries John J Horton. Buyer uncertainty about seller and waitlists for affordable housing alloca- capacity: Causes, consequences, and a tion. 2018. partial solution. 2018. Moshe Barach and John J Horton. How do John J Horton and Ramesh Johari. At what employers use compensation history?: Evi- quality and what price?: Eliciting buyer pref- dence from a field experiment. 2017. erences as a market design problem. In Proceedings of the Sixteenth ACM Con- Marianne Bertrand and Sendhil Mullainathan. ference on Economics and Computation, Are Emily and Greg more employable than pages 507–507. ACM, 2015. Lakisha and Jamal? a field experiment on labor market discrimination. American eco- Lily Hu and Yiling Chen. Fairness at equi- nomic review, 94(4):991–1013, 2004. librium in the labor market. arXiv preprint arXiv:1707.01590, 2017. Christopher Blattman and Paul Niehaus. Aanund Hylland and Richard Zeckhauser. The Show them the money: Why giving cash efficient allocation of individuals to posi- helps alleviate poverty. Foreign Affairs, 93 tions. Journal of Political economy, 87(2): (3):117–126, 2014. 293–314, 1979. Meredith Broussard. Artificial Unintelligence: Neal Jean, Marshall Burke, Michael Xie, How Computers Misunderstand the World. W Matthew Davis, David B Lobell, and Ste- MIT Press, 2018. fano Ermon. Combining satellite imagery Waylon Brunette, Mitchell Sundt, Nicola Dell, and machine learning to predict poverty. Rohit Chaudhri, Nathan Breit, and Gaetano Science, 353(6301):790–794, 2016. Borriello. Open data kit 2.0: expanding and Amanda R Kube, Sanmay Das, and Patrick J. refining information services for developing Fowler. Allocating interventions based on regions. In Proceedings of the 14th Work- counterfactual predictions: A case study on shop on Mobile Computing Systems and homelessness services. 2018. Applications, page 10. ACM, 2013. Jacob Leshno. Dynamic matching in over- Stephen Coate and Glenn C Loury. Will loaded waiting lists. 2017. affirmative-action policies eliminate nega- Karen Levy and Solon Barocas. Design- tive stereotypes? The American Economic ing against discrimination in online markets. Review, pages 1220–1240, 1993. Berkeley Tech. LJ, 32:1183, 2017.

32 AI MATTERS, VOLUME 4, ISSUE 3 4(3) 2018

Cynthia M Marlowe, Sandra L Schneider, and Nithya Sambasivan, Ed Cutrell, Kentaro Carnot E Nelson. Gender and attractive- Toyama, and Bonnie Nardi. Intermediated ness biases in hiring decisions: Are more technology use in developing communities. experienced managers less biased? Jour- In Proceedings of the SIGCHI Conference nal of applied psychology, 81(1):11, 1996. on Human Factors in Computing Systems, Daniel Mutembesa, Christopher Omongo, and CHI ’10, pages 2583–2592, New York, NY, Ernest Mwebaze. Crowdsourcing real-time USA, 2010. ACM. viral disease and pest information: A case Richard Ssekibuule, John A Quinn, and Kevin of nation-wide cassava disease surveillance Leyton-Brown. A mobile market for agricul- in a developing country. In HCOMP, pages tural trade in uganda. In Proceedings of the 117–125, 2018. 4th Annual Symposium on Computing for Ernest Mwebaze and Michael Biehl. Development, page 9. ACM, 2013. Prototype-based classification for image Sharifa Sultana, François Guimbretiere,` analysis and its application to crop disease Phoebe Sengers, and Nicola Dell. Design diagnosis. In Advances in Self-Organizing within a patriarchal society: Opportuni- Maps and Learning Vector Quantization, ties and challenges in designing for rural pages 329–339. Springer, 2016. women in bangladesh. In Proceedings Ernest Mwebaze and Godliver Owomugisha. of the 2018 CHI Conference on Human Machine learning for plant disease inci- Factors in Computing Systems, CHI ’18, dence and severity measurements from leaf pages 536:1–536:13. ACM, 2018. images. In Machine Learning and Ap- Neil Thakral. The public-housing allocation plications (ICMLA), 2016 15th IEEE Inter- problem. Technical report, Technical report, national Conference on, pages 158–163. Harvard University, 2016. IEEE, 2016. Aditya Vashistha, Edward Cutrell, Gaetano Neil Newman, Lauren Falcao Bergquist, Borriello, and William Thies. Sangeet swara: Nicole Immorlica, Kevin Leyton-Brown, A community-moderated voice forum in ru- Brendan Lucier, Craig McIntosh, John ral india. In Proceedings of the 33rd An- Quinn, and Richard Ssekibuule. Design- nual ACM Conference on Human Factors in ing and evolving an electronic agricultural Computing Systems, CHI ’15, pages 417– marketplace in Uganda. In Proceedings of 426. ACM, 2015a. the 1st ACM SIGCAS Conference on Com- Aditya Vashistha, Edward Cutrell, Nicola Dell, puting and Sustainable Societies, page 14. and Richard Anderson. Social media plat- ACM, 2018. forms for low-income blind people in in- Safiya Umoja Noble. Algorithms of Op- dia. In Proceedings of the 17th International pression: How Search Engines Reinforce ACM SIGACCESS Conference on Comput- Racism. NYU Press, 2018. ers and Accessibility, ASSETS ’15, pages Cathy O’Neil. Weapons of Math Destruc- 259–272. ACM, 2015b. tion: How Big Data Increases Inequality and Aditya Vashistha, Fabian Okeke, Richard An- Threatens Democracy. Broadway Books, derson, and Nicola Dell. ’you can always do 2016. better!: The impact of social proof on partic- Neil Patel, Deepti Chittamuru, Anupam Jain, ipant response bias. In Proceedings of the Paresh Dave, and Tapan S Parikh. Avaaj 2018 CHI Conference on Human Factors in otalo: a field study of an interactive voice fo- Computing Systems, page 552. ACM, 2018. rum for small farmers in rural india. In Pro- Daniel Waldinger. Targeting in-kind transfers ceedings of the SIGCHI Conference on Hu- through market design: A revealed prefer- man Factors in Computing Systems, pages ence analysis of public housing allocation. 733–742. ACM, 2010. 2017. John Alexander Quinn, Kevin Leyton-Brown, Yexiang Xue, Ian Davies, Daniel Fink, Christo- and Ernest Mwebaze. Modeling and moni- pher Wood, and Carla P. Gomes. Avi- toring crop disease in developing countries. caching: A two stage game for bias reduc- In AAAI, 2011. tion in citizen science. In Proceedings of

33 AI MATTERS, VOLUME 4, ISSUE 3 4(3) 2018

the 2016 International Conference on Au- tonomous Agents and Multiagent Systems, AAMAS ’16, pages 776–785, 2016a. Yexiang Xue, Ian Davies, Daniel Fink, Christo- pher Wood, and Carla P. Gomes. Behav- ior identiﬁcation in two-stage games for in- centivizing citizen science exploration. In Michel Rueher, editor, Principles and Prac- tice of Constraint Programming, pages 701– 717, Cham, 2016b. Springer International Publishing.

Rediet Abebe is a PhD candidate in computer science, advised by Jon Kleinberg. Her research focuses on algorithms, artiﬁcial intelligence, and their applications to social good. In particular, she uses algorithmic and computational insights to better understand socioeconomic inequality and inform interventions for improving access to opportunity. Her work is generously supported by fellowships and scholarships through Facebook, Google, and the Cornell Graduate School. Prior to Cornell, she completed an M.S. and a B.A. in Applied Mathematics and Mathematics from Harvard University and an M.A. in Mathemat- ics from the University of Cambridge. She was born and raised in Addis Ababa, Ethiopia. Kira Goldner is a ﬁfth- year PhD student at the University of Washington in the Department of Computer Science and Engineering, advised by Anna Karlin. Her research focuses on problems in mechanism design, particularly on (1) maximizing rev- enue in settings that are motivated by practice and (2) on mechanism design within health insurance and online labor markets. She is a 2017-19 recipient of the Microsoft Research PhD Fellowship and was a 2016 recipient of a Google Anita Borg Scholarship. Kira received her B.A. in Mathematics from Oberlin College and also studied at Budapest Semesters in Mathematics.