Learning to Vote

Our ultimate goal is to design or employing existing machine learning methods to train voting rules with certain axiomatic properties. The learnability for voting rules satisfying some desirable fairness axioms is very useful. Once a new voting rule is proposed, it may have some special fairness properties. It’s pretty hard to handcraft another voting rule with the same axiom. However, we could learn a similar voting rule from instances, or even a simpler rule with the exactly same property. It’s promising to learn all those existing voting rules efficiently and have their good genes to design a single meta-voting rule, that satisfies more fairness criteria, or at least with a higher satisfiability to all fairness criteria than any existing voting rule.

The very first step to learn a voting rule or mechanism from instances is to construct a training set. The input profiles and the corresponding winner(s) are provided by an expert or oracle voting rule, which we are attempting to learn. We query experts or oracles with preference profiles, they reveal us their favorite candidate(s). We extract important feature from these query-answer pairs and create the training set. Here we make a brief note on the features we extracted from preference profile (a comprehensive survey on existing voting rules again is required).