Sunday, July 14, 2013

Future of Game Theory: Stony Brook roundtable

This past week I had the opportunity to listen to a discussion of "The Present and Future of Game Theory". The highlight was a set of short talks by Aumann, Fudenberg, Kalai, and Maskin:

Robert Aumann* (Hebrew University and Stony Brook University) Overview
Drew Fudenberg (Harvard University) Predictive Game Theory
Ehud Kalai (Northwestern University) Game Theory and Computer Science
Eric Maskin* (Harvard University)  The Present and Future of Mechanism Design

Bob Aumann opens the session on the present and future of game theory
To my pleased surprise, they all seemed to see the future of game theory in applications of various kinds.








 Bob Aumann's first examples, of "hard" (as opposed to "soft") applications were from market design


Drew Fudenberg speaks about predictive game theory














 Drew focused on "predictive game theory," and the need for experiments and computational models.










Ehud Kalai speaks about game theory and computer science





Ehud Kalai spoke without slides, in front of photos of Lloyd Shapley and Jean Francois Mertens, who were honored in different ways at the conference. He spoke about how game theory and computer science are likely to become increasingly entangled in the years to come.

Eric Maskin speaks about the future of mechanism design








Eric Maskin included the fall of the planned economies of Eastern Europe among the applications of mechanism design, and also included the design of auction and matching markets.













Later in the same session I spoke about some of the differences between "game engineering" and game theory, and elaborated on the following points that I think are typical of the situation we often find ourselves in in market design:
1. We don’t know the whole game
2. We can’t compare designs just by their equilibria
3. Design solutions can’t always wait for reliable scientific knowledge
4. Market design solutions don’t last forever
5. We can’t just do the game theory, other problems have to be addressed to have a complete design

2 comments:

Note: Only a member of this blog may post a comment.