Auctions with Artificial Adaptive Agents
James Andreoni and John H. Miller
Key Words:
Auctions, Artificial Adaptive Agents, Genetic Algorithm, Bidding Behavior, Learning
Experiments on auctions find that subjects make systematic bidding errors that cannot be explained within the context of Nash equilibrium bidding models. Experimenters and others have conjectured that learning by subjects could lead to errors consistent with those observed. Here, we create and analyze a model of adaptive learning and demonstrate that such a model can capture the bidding patterns evident among human subjects in experimental auctions. Moreover, our model provides a variety of insights into the nature of learning across different auction institutions.