Abstract: 

SimStudent is a machine-learning agent that inductively learns cognitive skills from examples. The underlying learning technique is called programming by demonstration in the form of inductive logic programming. SimStudent is also capable of learning cognitive skills by tutored-problem solving where a tutor interactively provides feedback and hint when SimStudent is solving problems. In this talk, I'll present three major applications of SimStudent -- intelligent authoring, computational model of learning, and teachable agent for learning by teaching. In the context of intelligent authoring, SimStudent is used as an automated cognitive modeler to generate a cognitive model (i.e., a set of problem-solving skills) by demonstration. Such cognitive model is then used as a domain expert model for a Cognitive Tutor. As for the computational model of learning, we use SimStudent as a crash test dummy by controlling various learning parameters to see how such manipulation affect learning outcome. For the teachable agent project, we use SimStudent as a synthetic peer learner so that (human) students learn by teaching SimStudent