Collective coin flipping, robust voting schemes and minima of Banzhaf values

Abstract

The power of players in a collective decision process is a central issue in Mathematical Economics and Game Theory. Similar issues arise in Computer Science in the study of distributed, fault tolerant computations when several processes, some perhaps faulty, have to reach agreement. In the present article we study voting schemes which are relatively immune to the presence of unfair players. In particular, we discuss how to perform collective coin flipping which is only slightly biased despite the presence of unfair players. Mathematically this corresponds to problems concerning the minima of Banzhaf values in certain n -person games. These are measures of power studied in Game Theory. It is quite remarkable that while dictatorial voting games are, of course, the most sensitive to the presence of unfair players, some voting schemes that we propose here are significantly more robust than majority voting. Coin flipping was selected as a study case because of its simplicity and because collective coin flipping is widely used in randomized algorithms for distributed computations. It is our feeling that Game Theory has much to contribute to Computer Science and we are sure that further applications will be found.