EconCS: " Interdependent Values and Robustness in Optimal and Near-Optimal Mechanism Design"

Seminar | December 3 | 12:30-1:30 p.m. | 531 Cory Hall

 Inbal Talgam Cohen, Stanford (CS)

 Department of Economics, Electrical Engineering and Computer Sciences (EECS)

The algorithmic game theory community has largely focused on independent and private values (IPV) as a model for bidders' values in an auction. A more general model is the classic model of interdependent values [Milgrom and Weber, '82], which better captures many real-life, high-stake auctions. We make first steps towards extending the rich IPV-based theory to interdependent values. In the context of revenue-optimal mechanism design, we show conditions under which Myerson's optimal mechanism can be applied to interdependent values. One of these conditions is robustness of the truthfulness and individual rationality guarantees, in the sense that they are required to hold ex post. We then consider an even more robust class of mechanisms called "prior independent" (a.k.a. "detail free"), and show that by simply using one of the bidders to set a reserve, it's possible to extract near-optimal revenue in an interdependent values setting. This shows that a considerable level of robustness is achievable for general values in single-parameter settings.

 Faculty, Students - Graduate

 lunch will be served!, 510-517-8185