An Incremental Elicitation Approach to Limited-precision Auctions

An Incremental Elicitation Approach to Limited-Precision Auctions Alexander Kress Master of Science Graduate Department of Computer Science University of Toronto 2004 Auction-based mechanisms are increasingly being used for automating resource allocation among large numbers of agents. To make these sort of mechanisms viable one needs to consider the issues of communication and computation expenditure required by these protocols as well as their stability. In this thesis we study limited-precision, iterative mechanisms with dominant strategy equilibria designed for allocation of a single good. Our goal is to limit the communication between the players and the mechanism, reduce the amount of information revealed by the players, as well as minimize the players’ computational costs. We accomplish this by placing a number of operational constraints that permit the above objectives. We prove several necessary conditions that severely restrict the space of mechanisms satisfying our criteria. We develop a number of mechanisms and show that with a large and variable number of players, in the case of limited-precision, iterative mechanisms are superior to single-shot mechanisms.