January 11, 2023. Wednesday. 3.30PM.
TG23 (Town Hall Building)
Speaker: Edith Elkind
Title: Learning axes from samples: single-peaked preferences and beyond
Abstract: We consider a society where voters, who participate in an election, report their preferences over candidates, and these preferences are one-dimensional, i.e., consistent with the candidates being ordered on a left-to-right axis. We would like to sample voters’ preferences to uniquely identify the underlying axis. We give bounds on the number of samples required, both for the case where we can sample entire votes and for the case where we only have access to pairwise comparisons, for two natural distributions of voters’ preferences. We also analyse a more general setting where voters’ preferences may be captured by two axes, and one or both of these axes are unknown.
Joint work with Sonja Kraiczy (U of Oxford), ICML’22
Bio: Edith Elkind is a Professor of Computer Science at University of Oxford. Her research is in the area of algorithmic game theory and preference aggregation. She has published over 150 papers in top conferences and journals in AI and game theory, and is serving or has served on editorial boards of AI Journal, Journal of AI Research, Mathematics of Operations Research, ACM Transactions on Economics and Computation, and Social Choice and Welfare. She is a EurAI Fellow and an ELLIS Fellow. Edith obtained her PhD from Princeton in 2005, and held postdoctoral and faculty positions in UK, Israel and Singapore before joining Oxford in 2013.