Christopher J. MacLellan
Soar Technology, Inc.
SimStudent is a computation model of how humans learn given demonstrations. In particular, the system currently focuses on acquiring a procedural rule-based model given example demonstrations. I am interested in extending SimStudent in two directions. First, I am looking at methods to perform incremental concept learning (unsupervised and semi-supervised) while learning procedural rules (supervised). Second, I am interested in being able to teach SimStudent general strategies that it can apply regardless of the domain (i.e., guess and check, means-ends problem solving, etc.). Ideally, SimStudent would automatically acquire these strategies by studying examples.