ITR: Non-Cooperative Computing: Foundational Problems at the Interface of Computer Science and Game Theory

Stanford University, Stanford CA

Investigators

Abstract

This project brings together two disciplines, computer science and artificial intelligence on the one hand, and economics and game theory on the other, and by doing so addresses some fundamental problems of computing in the Internet era. Some of the problems addressed are rooted directly in today's Internet applications: How do you charge for network usage so as to smooth out peaks in the demand? How do you incent people to contribute personal information and recommendations in recommender systems and discretionary databases? While this project is motivated by its potential application to electronic commerce, the focus is on foundational matters. Computing in the Internet era means computing in the context of multiple self-interested entities, which in turn means that reasoning about computing must take the incentives of these entities into account. This, in turn, calls for a fundamental integration of ideas from computer science (such as fault tolerance, fairness, verification, algorithms, complexity, graphical models of uncertainty, machine learning) with elements of game theory (such as mechanism design, equilibrium analysis, game representations, learning, and agency theory). The outcome is expected to benefit electronic commerce, but also more fundamentally to help lay the foundations for a new area of research called "non-cooperative computing."