There is a current societal interest in social choice theory (e.g. voting systems) motivated at least in part by the divisiveness apparent in many countries. Can computer science and AI bring any insight into some of the more prominent issues?
More specifically, recent elections in the US and Canada (and elsewhere) have brought to prominence a couple of issues, Gerrymamdering and the role of primaries in voting. (Voting goes beyond political elections but political elections are the most newsworthy.)
As one can imagine, there are complex modeling and computational issues when dealing with large social and political systems.
We have some preliminary but we think interesting insights into gerrymandering (strategic design of electoral districts) and primary systems. I will briefly talk about a work at IJCAI 2018 on gerrymandering and how the ``power of gerrymadering'' relates to the degree of urbanization. I will mainly talk about the issue of primaries vs direct elections as our work here is a blend of both theory and experimental work. Here is the basic question: What is the impact on the ``quality'' of our chosen leaders by having primaries where each party has its own election to choose their candidate for the general election? Does this tend to result in more extreme candidates? In a paper at AAAI 2019 paper on primaries, we conduct the first quantitative study of primary vs direct elections.
Joint work with Omer Lev, Nisarg Shah and Tyrone Strangway
12/12/2019