When MS&E PhD candidate Faidra Monachou set out to organize the fourth Workshop on Mechanism Design for Social Good (MD4SG '20), she built solutions from her field of research into the conference itself.
Monachou, together with co-program chair Francisco J. Marmolejo Cossío of the University of Oxford, and a Steering Committee including MS&E Prof. Irene Lo, designed the workshop to be multidisciplinary, focused on applicability, and inclusive of communities that are typically underserved in academia.
The result was a diverse group of attendees and ideas, presented for that group. Invited speaker Natalia Ariza Ramírez, Economist at National University of Colombia and former Vice Minister of Education in Colombia, gave her plenary talk on educational interventions in Colombia in Spanish with live interpretation to English. A discussion panel with experts from Latin American and Caribbean communities (LAC) was hosted in Spanish, again with live interpretation. And real-time interpretation to Spanish was provided for the rest plenary talks given in English.
To achieve this novel linguistic diversity for a conference of this type, the organizers designed for it from the beginning. Monachou and team knew that a significant barrier for the participation of Latin American communities in similar initiatives has been language—almost all US-based conferences are held in English only. To overcome this barrier, the organizers released the Call for Participation both in English and Spanish, and Spanish-speaking Program Committee members reviewed and shepherded several submissions written entirely in Spanish.
Additionally, the organizers coordinated outreach to under-represented communities and institutions, many of whom faced financial difficulties or were based in the Global South. To overcome financial barriers, organizers worked with the MD4SD '20 Diversity & Inclusion chairs, including MS&E PhD candidate Wanyi Li, to identify candidates for registration fee waivers. Altogether, waivers were granted to 190 participants. For participants without reliable internet access, MD4SD '20 provided 21 data plan scholarships, 18 of which were awarded to participants in countries across the African continent.
As a result of these and other efforts, more than 700 attendees registered from 59 countries around the world. Over 20% of registrants identified as Black, African American, or African, and 18% were of Hispanic, Latino, or Spanish origin.
MD4SG '20 highlighted research where techniques from algorithms, optimization, and mechanism design, along with insights from other disciplines, have the potential to improve access to opportunity for historically underserved and marginalized communities.
For example, Monachou and team explicitly opened MD4SG '20 to submissions from legal scholars. Legal experts joined other researchers, policymakers, and domain experts and professionals interested in improving equity in a variety of domains such as education, labor, environment, healthcare, algorithmic fairness, and digital platforms.
Participants had backgrounds in computer science, AI, operations research, economics, public policy and humanities, and many of the 138 submitted papers combined methodologies and insights from multiple fields.
For example, the paper "Feminicide and Machine Learning: Detecting Gender-based Violence to Strengthen Civil Sector Activism" by lead author Catherine D'Ignazio, Professor at MIT, highlighted a highly relevant topic to the Latin American region by adopting machine learning tools to understand the complex issue of feminicide and inform policy at all levels. This paper was a co-winner of the MD4SD New Horizons award.
At the intersection between mechanism design and human-computer interaction, the paper "Modeling Assumptions Clash with the Real World: Configuring Student Assignment Algorithms to Serve Community Needs" by UC Berkeley team Samantha Robertson, Tonya Nguyen, and Professor Niloufar Salehi studied how the theoretical guarantees of the San Francisco Unified School District's student assignment algorithm can differ from the practical behavior of parents using the algorithm. This paper also received the New Horizons award.
Beyond science, policy, and humanities, the workshop even explored the intersection between art and mechanism design for social good. Keynote speaker Stephanie Dinkins, Artist Fellow at the Berggruen Institute, talked about her practice as a transmedia artist working with AI, which "employs lens-based practices, emerging technologies and community engagement to confront questions of bias in AI, consciousness, data sovereignty and social equity," according to Dinkins.
Several of the award-winning papers at the workshop focused on policy design, on topics such as educational policies for admissions at University of California, school choice in Peru and San Francisco, HIV prevention methods for homeless youth, and discrimination in labor markets.
Related to the workshop theme of bridging research and policy, award-winning papers highlighted new research directions for policy-oriented work in the MD4SG community. More specifically, the winning papers that equally shared the Best Paper Award focused on the design of effective educational policies in the United States and LAC.
The paper "Top Percent Policies and the Return to Postsecondary Selectivity" by Zachary Bleemer (UC Berkeley) used novel data from the "Top Percent" admissions policy at University of California to analyze the impact on barely-eligible applicants (typically from disadvantaged backgrounds) on their university admission and future career outcomes.
The paper "Competition under Social Interactions and the Design of Education Policies" by incoming Stanford GSB Professor Claudia Allende studied the role of peer preferences in school choice and the design of optimal assignment policies at elementary schools in Peru.
In general, education in LAC was a central theme to the workshop; the keynote talk by Dr. Natalia Ariza Ramírez focused on the implementation of the "Ser Pilo Paga" program aimed to make education in Colombia more accessible and equitable, while two contributed talk sessions and a discussion panel with experts from LAC also focused on education policy in practice.
In the midst of the isolation caused by COVID-19, Monachou and team expressed how valuable the social angle of a workshop like MD4SG '20 can be. According to Monachou, a common remark on the post-workshop participant survey was that participants enjoyed the social nature of the gather.town platform, which offered an environment wherein participants could "walk" their avatar through virtual space to meet and chat with other participants in a more casual way than traditional video conferencing platforms like Zoom. MD4SG partnered with Virtual Chair, who designed the virtual space for the conference and facilitated its technical logistics.
In fact, the social aspect of the workshop was so successful that MD4SG is currently in the process of providing a similar social space on a regular basis for MD4SG members throughout the year.
MD4SG '20 was organized by program chairs Monachou and Marmolejo, Steering Committee Prof. Lo, Rediet Abebe (Harvard University), Ana Stoica (Columbia University), Kira Goldner (Columbia University), Jon Kleinberg (Cornell University), Illenin Kondo (University of Notre-Dame), and Sera Linardi (University of Pittsburgh).
MD4SG is a multidisciplinary and multi-institutional online research initiative, which promotes research at the intersection of computer science, operations, economics, humanities and other disciplines, with the mission of bringing together a range of expertise to tackle problems impacting disadvantaged communities around the world.
Since its foundation in 2016, MD4SG has grown to a community of more than 1,700 participants and organizes workshops, tutorials, colloquium series at events such as the Association for Computing Machinery (ACM) conference on Economics and Computation (EC '20), Data Science Africa Summer School, ACM COMPASS, and more. MD4SG has seven working groups covering topics such as bias, discrimination, and fairness, developing nations, environment and climate, data economies, inequality, and civic participation.
MD4SG hosts year-round events, colloquia, and domain-specific working groups that anyone can join from home. For example, Monachou and Li in MS&E lead working groups on Bias, Discrimination and Fairness; and Environment and Climate, respectively.