Michael Kearns & Colin Camerer presents as part of the UBC Department of Computer Science’s Distinguished Lecture Series, March 13, 2014.
In brief: most of economic theory is based on the idea that people make decisions «rationally» to maximize their own gain. Practical experience—from reality shows to the stock market crash — make it painfully obvious that this idea often fails to describe reality. The question is what model of human behavior we can replace it with. Each of our two speakers bridges multiple disciplines himself: Camerer has joint appointments in Cognitive Psychology and Economics, while Kearns hails from Computer Science but has secondary appointments in Statistics and Operations and Information Management. Attendees should thus expect a truly interdisciplinary investigation into strategic decision-making from two of the world’s most influential thinkers on applied game theory.
Colin Camerer will lead off with his talk, Limits on strategic thinking: Evidence from brains to field data. This talk will describe where classical game theoretic equilibrium analysis (based on rationality assumptions) works well, where it doesn’t, and how to model human reasoning in the latter case. Specifically, he will describe «cognitive hierarchy» models and present evidence about how they work in a variety of lab and field settings. He will argue that in some cases equilibrium analysis predicts remarkably well either because it is approximated by CH (even with little learning) or the subjects have a special evolved advantage in particular types of interactions.
Michael Kearns will then describe his Experiments in Social Computation, considering the emerging phenomena of crowdsourcing and social computing. Most successful applications of crowdsourcing to date have been on problems we might consider «embarrassingly parallelizable» from a computational perspective. In part towards the goal of applying this technique to more challenging problems, for a number of years Kearns and his collaborators have been conducting controlled human-subject experiments in distributed social computation in networks with only limited and local communication. These experiments cast traditional computational problems as games of strategic interaction in which subjects have financial incentives to collectively «compute» global solutions. In summarizing his findings, Kearns will draw broad comparisons to predictions made both by the theory of computation and by microeconomics.