A hybrid system of abductive tactical decision making

Abstract

One important aspect of intelligent decision making is to infer a best explanation for a set of observations, a process resembling abduction. In this paper, we report a hybrid system to simulate this process. Specifically, the system combines a symbolic Soar component for deliberate explanation generation and a connectionist UEcho component for implicit explanation evaluation. To evaluate the system, we trained the system to perform a tactical decision making task based on expert data in one scenario, and tested it on two novel scenarios. The implications of such hybrid systems are discussed.