Mechanism Design and Economic Inference

The promise of data science is that system data can be analyzed and its understanding can be used to improve the system (i.e., to obtain good outcomes). For this promise to be realized, the necessary understanding must be inferable from the data. Whether or not this understanding is inferable often depends on the system itself. Therefore, the system needs to be designed to both obtain good outcomes and to admit good inference. This talk will explore this issue in a mechanism design context where the designer would like use past bid data to adapt an auction mechanism to optimize revenue. Data analysis is necessary for revenue optimization in auctions, but revenue optimization is at odds with good inference. The revenue-optimal auction for selling an item is typically parameterized by a reserve price, and the appropriate reserve price depends on how much the bidders are willing to pay. This willingness to pay could be potentially be learned by inference, but a reserve price precludes learning anything about willingness-to-pay of bidders who are not willing to pay the reserve price. The auctioneer could never learn that lowering the reserve price would give a higher revenue (even if it would). To address this impossibility, the auctioneer could sacrifice revenue-optimality in the initial auction to obtain better inference properties so that the auction’s parameters can be adapted to changing preferences in the future. In this talk, I will develop a theory for optimal auction design subject to good inference.

Speaker Details

Prof. Hartline is on sabbatical at Harvard Economics and Computer Science Departments for the 2014 calendar year (January 2014-December 2014). Prof. Hartline’s current research interests lie in the intersection of the fields of theoretical computer science, game theory, and economics. With the Internet developing as the single most important arena for resource sharing among parties with diverse and selfish interests, traditional algorithmic and distributed systems approaches are insufficient. Instead, in protocols for the Internet, game-theoretic and economic issues must be considered. A fundamental research endeavor in this new field is the design and analysis of auction mechanisms and pricing algorithms. Dr. Hartline joined the EECS department (and MEDS, by courtesy) in January of 2008. He was a researcher at Microsoft Research, Silicon Valley from 2004 to 2007, where his research covered foundational topic of algorithmic mechanism design and applications to auctions for sponsored search. He was an active researcher in the San Francisco bay area algorithmic game theory community and was a founding organizer of the Bay Algorithmic Game Theory Symposium. In 2003, he held a postdoctoral research fellowship at the Aladdin Center at Carnegie Mellon University. He received his Ph.D. in Computer Science from the University of Washington in 2003 with advisor Anna Karlin and B.S.s in Computer Science and Electrical Engineering from Cornell University in 1997.

Date:
Speakers:
Jason Hartline
Affiliation:
Northwestern University