Fair and Fast Tie-Breaking for Voting

We introduce a notion of fairest tie-breaking for voting w.r.t. two widely-accepted fairness criteria: anonymity (all voters being treated equally) and neutrality (all alternatives being treated equally). We proposed a polynomial-time computable fairest tie-breaking mechanism, called most-favorable-permutation (MFP) breaking, for a wide range of decision spaces, including single winners, $k$-committees, $k$-lists, and full rankings. We characterize the semi-random fairness of commonly-studied voting rules with MFP breaking, showing that it is significantly better than existing tie-breaking mechanisms, including the commonly-used lexicographic and fixed-agent mechanisms.