This paper presents work on a computer model of how humans learn to solve problems with diagrams. Although earlier work has discussed the kinds of representations which experts use when solving problems, not much work has been done on how these representations may be learned. We focus on a particular problem solving domain and describe a specific diagrammatic representation for it. We define a novel computer model which is able to use the perceptual information available in typical problems in this domain to learn appropriate representations to assist the problem solving process. Specifically, we consider the different strategies used by subjects in constructing a diagrammatic representation of an electric circuit known as an AVOW diagram. Experiments offer evidence that subjects learn perceptual chunks about the domain, and also show how subjects exploit the external representation. Of primary importance for a more comprehensive model is how the problem solver creates an appropriate internal representation of the domain. We use a model for Long-Term Memory based upon an extended version of EPAM, which can handle multiple representations. The model as currently implemented retrieves visuo-spatial information using a directable eye and obeys the constraints of Short-Term Memory to learn perceptual chunks about circuits and their associated AVOW diagrams.
Further details of this project may be found at: http://www.psychology.nottingham.ac.uk/research/credit/projects/problem_solving, including a postscript version of this paper.To receive an off print, please email Irene Jackson.