Date of Original Version




Rights Management

All Rights Reserved

Abstract or Description

The goal of our ongoing project is to develop and test various methods of modeling one's opponent in adversarial domains. The texas hold'em domain will be used because it has the key features of imperfect information, stochasticity, and the ability to deceive. These three components allow for a very wide range of human behavior making it difficult to create feasible models using traditional artificial intelligence techniques. It is precisely in these situations where models based on human cognition are most helpful. Their adaptive and flexible methods can overcome the uncertainty and deceptiveness intrinsic to human behavior aiding with naturally difficult domains. In other words, to understand a human it is helpful to think like one.


Advisor: John Anderson (Christian Lebiere)
Degree: B.S. Computer Science