Learning Bayesian Game Families, with Application to Mechanism Design

Madelyn Gatchel (University of Michigan), Michael P. Wellman (University of Michigan)

Abstract

We demonstrate the advantages of learning an interim model for Bayesian game families through an in-depth study of empirical mechanism design for a dynamic sponsored search auction scenario. A full version of this paper, with additional background, methods, and results, is available at: https://arxiv.org/pdf/2502.14078.