ORC Seminar: Internal Pluralism and the Limits of Pairwise Comparisons

About this event

Bailey Flanigan Assistant Professor MIT Abstract: Across AI alignment and participatory design, pairwise comparisons are a workhorse for learning how people want automated decision rules to behave. We propose a preference model that contains standard pairwise-choice models as a special case, but which allows an individual to evaluate a decision rule according to multiple distinct priorities, and to experience irreducible conflict between them. This added generality exposes two ways in which standard learning pipelines can fail. First, some natural priorities—such as proportionality, egalitarianism, and equal treatment—are inherently global: what they imply for one local comparison depends on how the rule behaves elsewhere; we find that such priorities cannot in general be recovered from isolated local comparisons. Second, even when priorities are local, they may strongly disagree, creating genuine conflict that forced binary choices can distort. We will conclude by turning the framework around: rather than viewing priorities only as a source of complications for pairwise learning, we can treat them as the object we want to learn. This suggests richer methods for eliciting input on choices over not only automated decision rules, but over combinatorial spaces in general. The talk is based on this paper https://arxiv.org/pdf/2607.02672, published at EC 2026. Bio: Bailey Flanigan is a Theodore T. Miller (1922) Career Development Assistant Professor at MIT, joint between Political Science and Computer Science (LIDS). Prior to joining MIT in 2025, Bailey received her PhD from CMU in Computer Science in 2024, and then she spent one year at Harvard as an HDSI Postdoctoral Fellow. Her work builds, deploys, and evaluates tools for supporting richer democratic participation by everyday people. Applications of her work include deliberative democracy, participatory budgeting, and ranked-choice voting. Her work uses methods from algorithms and optimization, game theory and social choice, learning theory and statistics, political theory, political behavior, and survey methods.