My main research interest is in how an agent can be interactively shaped by a human. In other (more technical) words, how should an agent learn online from signals of positive and negative reinforcement given by a human trainer who is observing the agent's behavior? Listed in my publications, the K-CAP 2009 paper on the TAMER framework and the AAAI Spring 2009 Symposium paper on designing human-shapable agents describe our early progress towards this goal. More broadly, I'm interested in learning agents, reinforcement learning, and general machine learning.

I am also currently leading a reading group called Agents that Learn from Humans.

   PhD student, Department of Computer Science

    University of Texas at Austin

    bradknox*AT*cs.utexas.edu