Computer Science Thesis Proposal
In Person and Virtual - ET - Gates Hillman 7501 and Zoom
BAILEY FLANIGAN , Ph.D. Student, Computer Science Department, Carnegie Mellon University
Accounting for Stakes in Democratic Decisions
At the backbone of any democratic society are collective decisions: society-level choices over, e.g., policies or candidates, on which a society's members weigh in and are in turn potentially affected by the outcome. We begin this thesis from the fundamental idea that a given collective decision can affect different members of society to differing degrees. For example, consider the decision of whether to instate a city-wide COVID masking requirement: immunocompromised people and/or essential workers can likely be much more dramatically affected by the outcome of this decision than people who work from home and can easily shelter in place. In mainstream language, we often talk about these disparities in terms of stakes, where someone with high stakes in a decision stands to gain or lose significantly depending on its outcome, while someone with low stakes is relatively unaffected by the decision.
It is intuitive not only that people's stakes may differ, but that accounting for these differing stakes is important to the quality of decisions' outcomes. The ubiquity of this idea is seen often, for example, in critiques of how processes fail to do so: for instance, many critique the fact that the global south, despite being disproportionately affected by climate change, has thus far been granted far less power in deciding global climate policy. Despite the omnipresent sense that stakes are an important factor in collective decisions, however, social choice theory – focused precisely on formally studying collective decision-making processes – offers no framework for even reasoning about people's stakes, let alone designing processes which account for them.
In this thesis, we aim to close these gaps by (1) building a theory of stakes in social choice, and then (2) using this theory to design new collective decision processes that account for stakes. The processes we propose achieve bounded or constant social welfare loss under no assumptions about voters' latent utilities or the size of the electorate – a previously unattainable goal in the social choice literature. Moreover, many of the processes we propose are highly practicable, as evidenced by support from interdisciplinary literatures, existing practical implementations, and in some cases, our own efforts to deploy our tools.
Ariel Procaccia (Chair)
Nika Haghtalab (University of California, Berkeley)
Ashish Goel (Stanford University)
In Person and Zoom Participation. See announcement.