Fair Algorithms for Selecting Citizens’ Assemblies

The Department of Political Science welcomes

Bailey Flanigan, PhD Candidate, Carnegie Mellon University

Job Candidate for the Shared Position in the Department of Political Science and the Schwarzman College of Computing

Presenting : Fair Algorithms for Selecting Citizens’ Assemblies

Monday, November 20th  – 1:00 pm to 2:30 pm

In the Millikan Room, E53-482

We hope to see you there!

======================

Abstract: Academics and political practitioners around the globe are experimenting with a wide variety of democratic innovations under the heading of “deliberative mini-publics (DMs).” In a DM, a panel of constituents convenes to deliberate about specific issues and make recommendations to traditional political decision-makers (e.g., legislators) about policies, priorities, or key considerations that should guide policy outcomes. Many proponents of DMs intend them as a way to augment, rather than replace, representative democracy, seeing them as a way to inject more robust deliberative and participatory elements into traditional politics.

Nearly all DMs rely on sortition – random selection – to choose the panelists. Sortition is often thought of as a simple lottery that chooses all constituents with equal probability. This approach has intuitive appeal and is argued for by many theorists; in practice, however, simple random selection rarely yields representative panels because of selection bias in who accepts invitations to participate. Many practitioners of DMs therefore sacrifice the pure equality embodied by a simple lottery, instead imposing quotas on socially salient groups and then “randomizing” within those constraints. While selecting a panel this way may seem straightforward, my work demonstrates that it can come with important unintended costs: if the randomization is not engineered carefully, groups unprotected by quotas – notably, latent intersectional groups that deserve a chance to be represented – may receive almost no probability of selection.

Engineering this randomization while satisfying user-specified quotas turns out to be technically demanding. My talk covers our algorithmic solution to this problem, called Leximin (Fair Algorithms for Selecting Citizens’ Assemblies, Nature, ‘21). Subject to such quotas, Leximin randomizes in a way that maximizes the minimum probability of selection attained by any willing participant, thereby maximizing latent intersectional groups’ chances at representation. This algorithm has been adopted widely in practice. It also allows us to explore the previously-obscure tradeoffs between quotas and qualitative desiderata conferred by how we randomize. In doing so, our algorithm illuminates new, impactful questions spanning computer science, political science, and democratic practice.

I conclude by discussing a main thread of my future work, which builds on the flexibility of our algorithmic approach. In the work above, we sought to promote the ideal of equality; however, our framework can actually ensure fairly generic representation desiderata. In my ongoing work, I am developing deployable democratic tools to ensure representation for groups subject to precarity, in accordance with philosophical accounts of justice prioritizing the claims of those most vulnerable to unfavorable political decisions. But that is only the beginning: we could ensure representation based on contextual imbalances of power, the distinctiveness of an intersectional group’s perspective, or other criteria of interest.