20 May 2022, Computational Social Choice Seminar, Carlos Navarrete

Abstract
Social choice theory studies how to aggregate individual preferences into those of groups. Yet, most work on social choice theory focuses on elections involving a few candidates, leaving relatively unexplored the question of how to aggregate preferences in direct participation involving multiple policy proposals. Here we study people’s preferences over 90 policy proposals using data collected in digital platforms deployed in Chile and Lebanon during massive mobilizations. Our results show that, unlike in elections involving a few candidates, in elections involving multiple proposals, measures of disagreement—which are orthogonal to traditional social choice theory aggregates—work better at identifying issues at the core of the mobilizations. We explore this data further using the axiomatic properties of social choice theory, finding that aggregate preferences are 90% Condorcet efficient, that measures of divisiveness are more robust to tyrannies of the majorities than measures of agreement, and that the top-ranked proposals are robust to the removal of any single alternative. These results suggest that divisiveness is a relevant form of aggregation that should be considered in electoral systems involving multiple policy proposals.

For more information on the Computational Social Choice Seminar, please consult https://staff.science.uva.nl/u.endriss/seminar/.