Human reasoners can reason from problems they have not encountered before and their creation of new chains of inferences is informative for the construction of artificially intelligent theorem provers that learn to reason in novel ways. We argue that human reasoners generate strategies to guide their sequences of inferences to the solution of novel or complex problems. We propose that human reasoners develop strategies by constructing a new strategic component based on features of a problem, and they assemble a new strategy by combining new and existing strategic components in novel ways. A new strategy can emerge, provoked by a new sort of problem that the existing strategies cannot solve. According to this view, reasoners can spontaneously improve their reasoning skills with practise because they gain expertise not only at implementing the individual parts of a strategy, but also at assembling the parts together efficiently. We report experimental results that support these proposals.
|Workshops in computing
|Irish Conference on Artificial Intelligence and Cognitive Science (5th : 1992)
|10/09/92 → 11/09/92