Diana F. Gordon, Devika Subramanian
Our goal is to develop a hybrid cognitive model of how humans acquire skills on complex cognitive tasks. We are pursuing this goal by designing hybrid computational architectures for the NRL Navigation task, which requires competent sensorimotor coordination. In this paper, we empirically compare two methods for control knowledge acquisition (reinforcement learning and a novel variant of action models), as well as a hybrid of these methods, with human learning on this task. Our results indicate that the performance of our action models approach more closely approximates the rate of human learning on the task than does reinforcement learning or the hybrid. We also experimentally explore the impact of background knowledge on system performance. By adding knowledge used by the action models system to the benchmark reinforcement learner, we elevate its performance above that of the action models system.