Restructuring, Reminding and Repair: What's Missing from Models of Anology

Article type: Research Article

Authors: Wolstencroft, John^;

Affiliations: Department of Computer Science, University College London, London, England, (+44)-1-387-7050 Fax. (+44)-1-387-1397, [email protected]

Note: [] A earlier version of this paper was presented at SCAI'89 – the 2nd. Scandinavian Conference on Artificial Intelligence, Tampere, Finland, June 1989, and was awarded a prize for best student paper. This work was carried out under a SERC studentship award, the financial assistance of which is gratefully acknowledged.

Abstract: Models of analogy developed to date fail to consider all the stages of analogical reasoning required for an effective model. Further, much emphasis is placed on the structural, syntactic nature of analogy. We describe a seven stage model of analogy, and discuss four means of determining analogical similarity. In this context we review two prevalent and contrasting analogical approaches, and conclude by offering suggestions that alleviate some of the problems currently faced, in the form of a new model for analogy which takes into account all seven stages and four similarity constraints. An example (Duncker's radiation problem) is used to highlight deficiences with current models and describe the new model.

DOI: 10.3233/AIC-1989-2202

Journal: AI Communications, vol. 2, no. 2, pp. 58-71, 1989