Abstract (EN)

Since the seminal works of Condorcet and Borda, social choice theory has explored how to aggregate individual preferences into collective decisions. Yet, social choice theory has focused primarily on identifying winners in elections involving few candidates, leaving questions about direct participation on multiple issues relatively unexplored. Here we analyze data collected in a direct participation experiment where people built their own government programs using 120 proposals from the candidates of the 2022 French presidential. We find that in this setting it is useful to introduce a measure of "divisiveness," which can be constructed for any voting rule, is orthogonal to them, and helps identify polarizing proposals. We show that divisiveness captures fragmentation across multiple dimensions (sex, age, political orientation, and urban-rural divide) and explore some of its axiomatic properties. These results suggest divisiveness is a relevant aggregate in direct forms of participation.