Game Theory, Computational Limitations

Game theory has provided useful predictions in many situations. In numerous games of interest, however, game theory’s predictions have missed the mark. Joseph Y. Halpern, Computer Science, and Rafael N. Pass, Cornell Tech, are focusing on one potential explanation: players’ computational limitations.

Traditionally, game theory assumes that all players are rational, computing their beliefs and a best response to what they believe other players are doing. These computations may not be so easy for players to perform. Halpern and Pass still want to assume players are rational but are working to model the fact that people are computationally limited and are affected by how games are framed or described in language terms.

A particular area of focus is how this approach can be used to explain irrational-seeming human behavior. Two examples are rational inattention, where people ignore variables that seem relevant, and anomalies in human behavior in real-world applications of game theory, such as in the security of wildlife and fisheries, forest protection, and drug interdiction. The team is also looking at computational and language issues in dynamic games, where agents’ understanding of the game improves as they play, affecting their judgments and actions. Halpern and Pass are searching for ways to model this learning process and the effects of different reasoning strategies. They are also applying their approach to cryptography, which can be viewed as a game between resource-bounded agents. The research will lead to a better understanding of the effect of computational limitations on people’s behavior and could enable better mechanisms for preventing undesirable actions.

Cornell Researchers

Funding Received

$1.2 Million spanning 4 years

Sponsored by

Other Research Sponsored by National Science Foundation