Saturday, September 05, 2009

CS and Econ at Cornell: Day II

  • Avrim Blum spoke about potential games, and cost of uncertainty. His talk reminded you the CS art of problem formulation which I think of as one of our contributions to the world. Yoav Shahom argued against addiction to equilibrium. He seemed like one who can research or play or combine both and he did. He spoke about equilibrium of strategies in computational games and demo-ed his winning software for pool! Bill Sandholm spoke about population games where the number of players is large and evolutionary game theory kicks in. David Parkes spoke about cases like dynamic auctions where you can't go for strategyproof and optimality directly, but proposed starting with heuristic mechanisms and tweaking them. Dirk Bergemann spoke about modeling and tradeoff of targeting. Michael Schwarz spoke about finding stable matches in which people are matched to the median partners in utility. Fun stuff with the stable matching lattice.
  • Lance Fortnow gave a quintessential theory talk, formulating computational time creatively as discounted utility in two party games and other similar problems. Costas Daskalakis gave a great talk, neatly laying out the hardness of general games against zero-sum games and sketched out some special cases which may be easier. His talk drew a lot of discussion from Economics, mainly wondering what the hardness of these models meant for real behavior of agents. PPAD met milk price at Walmart.
David Easley, Joan Feigenbaum, Joe Halpern and H. Peyton Young led a panel at the end. Joan formulated a cogent argument for studying game theory on graphs, that IP networks at infrastructure level comprise real economic agents that strategize. Joe Halpern did the much needed job of extended the scope of the discussion to include logic and languages, and to real systems issues like asynchrony, fault tolerance, verifying correctness, etc. Kindred spirit there. David Easley emphasized that we need to understand market design better, with the example of his experience with the Spanish Option Exchange design. It is hard to experiment with market design in some worlds like financial exchanges, seemingly less so in Internet ads world. Peyton Young too emphasized how we need to understand designing markets, in particular, with limited information.



Post a Comment

<< Home