Preference Engines and Fair Elections

Abstract: These days large language models are transforming many fields with their ability to generate rich, human-like text.  In this talk I’ll introduce some generative models for rankings, seeking to create an engine for simulating ranked elections.  In other words, faced with a large number of candidates that break into several types, how are people likely to rank their preferences?  This is a key step in a research program to better understand how different voting rules lead to qualitatively different representation, according to various possible norms of representative democracy.  I’ll particularly pay attention to the question of whether STV, a popular multi-winner ranked choice election method, tends to produce proportional representation.