Year: 2004
Type: article
Abstract: A Model of Analogical Retrieval Using Intermediate Features Mark Alan Finlayson ([email protected]) Patrick Henry Winston ([email protected]) Computer Science and Artificial Intelligence Laboratory, MIT 32 Vassar St., Cambridge, MA 02139 USA We present a model of analogical recall in people which draws inspiration from recent work in visual classifica- tion by Ullman (2002). Our model is intended to unify two bodies of evidence regarding recall in people: on the one hand, we seek to cover a body of evidence that in- dicates people drawn from a population without regard to task expertise are heavily influenced by surface sim- ilarity during retrieval; on the other hand, we also seek to account for the fact that experts are able to achieve analogical recall on a consistent basis. Our model works by breaking the symbolic graph representation of an in- put situation into sub-graph structures (structures we call features), and looking for these features in other sit- uations. By varying the informativeness of the features we use to retrieve situations, we are able to promote or suppress analogical retrieval. Our model is consistent with previous models of recall (Thagard, Holyoak, Nelson, & Gochfeld, 1990; Forbus, Genter, & Law, 1994) which indicate object similarity, first-order relations, and some small amount of struc- ture dominate recall in normal subjects. These models were primarily intended to account for evidence of the predominance of so-called “mere-appearance” matches in normal recall (Gick & Holyoak, 1980; Rattermann & Gentner, 1987), while still acknowledging some struc- tural effects (Holyoak & Koh, 1987; Wharton et al., In contrast to these previous models, however, our model indicates an explanation for certain results in the field of expert problem-solving and retrieval, which has received less attention to date. Evidence drawn from this literature (Chi, Feltovich, & Glaser, 1981; Schoenfeld, 1982; Shneiderman, 1977) indicates that certain sorts of people do consistently achieve analogical recall in par- ticular domains: while these people often fall under the heading “expert,” non-experts are also able to attain structural reminding under particular circumstances. We run our model on a dataset of descriptions of com- plex political scenarios, and show the predicted switch- ing of preference from mere-appearance to analogical matches when moving from low average feature infor- mativeness to high average feature informativeness. Fur- thermore our results indicate, as Ullman’s did, that fea- tures of an intermediate size and complexity provide the most robust recall within analogical category. Acknowledgments This work is supported in part by the National Science Foundation under Grant Number 0218861. References Chi, M. T. H., Feltovich, P. J., & Glaser, R. (1981). Categorization and representation of physics prob- lems by experts and novices. Cognitive Science, 5, Forbus, K. D., Genter, D., & Law, K. MAC/FAC: A model of similarity-based retrieval. Cognitive Science, 19, 141-205. Gick, M. L., & Holyoak, K. J. (1980). Analogical prob- lem solving. Cognitive Psychology, 12, 306-355. Holyoak, K. J., & Koh, K. (1987). Surface and struc- tural similarity in analogical transfer. Memory and Cognition, 15, 332-340. Rattermann, M. J., & Gentner, D. (1987). Analogy and similarity: Determinants of accessibility and inferential soundness. In The Ninth Annual Con- ference of the Cognitive Science Society (p. 23-35). Lawrence Erlbaum Associates. Schoenfeld, A. H. (1982). Problem perception and knowledge structure in expert and novice math- ematical problem solvers. Journal of Experimental Psychology: Learning, Memory, and Cognition, 8, Shneiderman, B. (1977). Measuring computer program quality and comprehension. International Journal of Man-Machine Studies, 9, 465-478. Thagard, P., Holyoak, K. J., Nelson, G., & Gochfeld, D. (1990). Analog retrieval by constraint satisfaction. Artificial Intelligence, 46, 259-310. Ullman, S., Vidal-Naquet, M., & Sali, E. (2002). Visual features of intermediate complexity and their use in classification. Nature Neuroscience, 5, 682-687. Wharton, C. M., Holyoak, K. J., Downing, P. E., Lange, T. E., Wickens, T. D., & Melz, E. R. (1994). Be- low the surface: Analogical similarity and retrieval competition in reminding. Cognitive Psychology,
Authors: Mark A. Finlayson, Patrick Henry Winston
Institution: Massachusetts Institute of Technology
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Funder: National Science Foundation
OpenAlex ID: W2580641315