My research projects center around the architecture underlying intelligence. I see that as a necessary precursor to building general, autonomous intelligent agents, which is my ultimate goal. Although there are different themes and scientific questions being addressed by my research projects, most of them share the same underlying architecture: Soar. Moreover, my goal is to develop a general architecture that supports agents that can work in a variety of complex, dynamic environments. Humans are the best example we have to date of intelligent agents with the ability to be successful in many, different environments, and thus I continually try to learn as much as I can about the structure of human cognition. My research often deliberately conflates research in cognitive science and artificial intelligence, trying to integrate what we’ve learned from both into a single system. My strategy is to fold back into Soar all that we learn from our research projects. All of this research owes a tremendous debt to Allen Newell and Paul Rosenbloom, and other members of the Soar research group. Although there are significant interactions among my research projects, for presentation purposes, it is simpler to separate out my more AI work from the more cognitive modeling work, and some work I’ve done on AI in computer games.
- AI research: Soar architecture development, AI architecture evaluation, integrated intelligent agents in distributed simulations, learning within intelligent agents, …
- Cognitive modeling research: Concept acquisition, dual tasks, analogy, subtraction, high level cognitive models, … subtraction
- Computer games research: Using AI in computer games (Quake not Chess), intelligent opponents, opponent modeling, automatic creation of opponents through learning, …