Optimality and Risk in Purchase from Multiple Auctions

Abstract

In the face of multiple electronic sales sites, buyers can benefit from considering multiple opportunities and devising their purchase strategy to reduce risk and increase expected utility. However, human users cannot approach and rapidly monitor multiple sites. In particular, sites where prices change dynamically such as auction sites pose a major difficulty for human concurrent activity. Even without concurrency, the majority of human users do not have the ability or the resources to compute optimal purchase decisions. Such activities can be performed by computational agents. In this paper we present mechanisms that allow agents to perform purchases on behalf of users. In particular, we devised methods that allow an agent that faces multiple dynamic sales sites to compute bids that optimized the expected utility of the user, or instead reduce and manage the risk of the purchase. The proposed mechanism are currently embedded in agents we develop in our lab.