Concept formation over explanations and problem-solving experience

Abstract

Recent research suggests the utility of performing induction over explanations. This process identifies commonalities across explanations that cannot be extracted solely by explanation-based techniques. This has important implications for the 'correctness' of learned knowledge [Flann and Dietterich, 1989] and, as we show, on the efficiency with which learned knowledge can be reused. Specifically, we illustrate that inductive concept formation can abstract and organize explanatory knowledge for efficient reuse in a domain of algebra story problems.