Modelling Typical Alphabetic Analogical Reasoning

Grob, Judith D M and Wood, Sharon (2003) Modelling Typical Alphabetic Analogical Reasoning. In: The European Cognitive Science Conference 2003, Osnabruck, German.

Full text not available from this repository.

Abstract

A study investigating the way in which people solve alphabetic analogical reasoning tasks (cf Copycat; Hofstadter & Mitchell, 1995), revealed that participants tend to use the same basic strategy, which was modelled in the cognitive architecture ACT-R. Performance evaluations indicate an average goodness of fit of 66.75% and a 100% goodness of fit on a subset of problems for which participants were significantly more likely to produce a single 'typical' response (p<0.05). The model is discussed in the context of various features of human analogical reasoning which were observed in the study, and in relation to Hoftstadter and Mitchell's (1995) discussion of 'elegant' solutions to problems.

Item Type: Conference or Workshop Item (Paper)
Additional Information: Originality: This work is a novel design and implementation of a computational cognitive model emulating human performance for a well-known analogical reasoning task. The model is implemented in a well-known and understood cognitive architecture, providing insight into the mechanisms of, and knowledge required for, problem solving. Rigour: Studies with human subjects and detailed analysis of problem solving protocols provide the basis for a plausible model of human problem solving for this task, with systematic discussion and analysis of correspondences and deviations from human performance studies. Significance: Contributes to a body of knowledge on the underlying mechanisms of human problem solving in general, and for this range of problems in particular.
Schools and Departments: School of Engineering and Informatics > Informatics
Depositing User: Sharon Wood
Date Deposited: 06 Feb 2012 21:26
Last Modified: 13 Apr 2012 12:42
URI: http://sro.sussex.ac.uk/id/eprint/31234
📧 Request an update