We brought some excitement to the Wall Street neighborhood on Friday, as researchers from CS, Economics, Business, and other areas stormed the offices of the NY Academy of Sciences
--- not far from Wall Street --- for the NYCE
went first and spoke about the complexity of finding (approx) equilibria. He started off our lethargic minds with a cute proof of Sperner's lemma about multicolored triangles on colored grids, and via Brouwer, established hardness of finding exact equilibria. I wonder what non-theoreticians thought of it when he constructed gadgets such that their equilibrium will perform arithmetic operations, like multiplication and addition, but for a theory CS person it is exciting when the structure clicks in to place.
from Columbia U. was the next speaker. As you know, computer scientists go ahead and build things -- search engines, social networks etc., and modelers have to understand and explain their impact later. Asim, a Marketing professor who is an expert on Bayesian modeling, described in detail not only the basics of statistical modeling and the complexities of huge number of covariates even in static cases, but also its empirical application to a data set of artists sharing music.Susan Athey
went next. She is an Economist, and gave a really insightful overview of the complex games behind internet ads. Her talk was a hit with the audience. Some CS engineers built a platform for showing ads with search queries via an auction some 4+ years ago, and Economists now study the system, indeed find some resonance with principles they know, but discover nuances and novelties, and ultimately are excited to work on the system and change pieces! I asked Susan if she felt fundamentally new Economic theory would emerge, and she said yes, in particular since she is trying to model the search user's utilities as well. This has been the one line summary I give people for sometime now on why we in CS have been developing new methods for search ads for the past several years when Economists have studied auctions for 50+ years, so it is reassuring.
ps: Btw, there is a 4-party game (Problem 3
) we CS people in ad auction systems have been working on for a few years now, perhaps now it will get more attention of the Economists. Also, Joan
asked why a boutique shoe store and a mega shoe outlet will both compete for the same search query. As Susan said it is because users don't specify exactly what they are looking for, and just provide one or two search terms. I did not get to discuss with Joan that what this means is that the offline and online notions of competitors are very different.