PhD Seminar • Artificial Intelligence • Learning Voting Rules Using Neural Networks

Please note: This PhD seminar will take place online.

Ben Armstrong, PhD candidate
David R. Cheriton School of Computer Science

Supervisor: Professor Kate Larson

We present our work using machine learning models to approximate social choice functions, a.k.a. methods of voting. Voting rules are functions that are given voter preferences and produce a winning candidate.

This talk will showcase the ability of a variety of machine learning models to accurately reproduce the output of several well-known voting rules. We show that the transformations applied to preference data before it is input to the ML model dramatically affect the ability of the model to learn. We find that the best transformation to perform upon data depends upon Fishburn’s classification of voting rules.