Applying Multi-Objective Evolutionary Computing to Auction Mechanism Design

The mechanism design problem in economics is about designing rules of interaction for market games which aim to yield a globally desirable result in the face of self-interested agents who may attempt to take advantage of the mechanism in order to maximize their own individual outcomes. This problem can be extremely complex. Traditionally, economists have used game theory and other formal methods to construct mechanism rules. In this paper, we report on an alternative approach which we hope will eventually yield more robust solutions than the present analytical counterparts. Our methodology views mechanism design as a multiobjective optimisation problem and addresses the problem using genetic programming. This paper reports on preliminary work in this direction where we evolve an auction-pricing rule for a continuous double auction using a multi-objective fitness function.