Introduction

This is a report on our development of analogy within a production system architecture. A couple of years ago we became persuaded that analogical problem solving was fundamental in the skill acquisition domains we were studying: programming and mathematical problem solving. Students were always resorting to examples from their textbooks or to examples from more familiar domains for solving these problems. We did a number of simulations of this problem solving within the ACT* architecture (Anderson, Farrell, & Sauers, 1984; Anderson, Pirolli, & Farrell, 1988; Pirolli & Anderson, 1985). However, the process of simulating this was awkward within that architecture. It was also ad hoc in the sense that the architecture offered no explanation as to why students were using analogy as their primary problem-solving method. This led us down the path of thinking about how analogy takes place and how it fits into a production system architecture. This is the first report of the new system that we have developed in response to these ruminations.

The term analogy is used in multiple senses. We are concerned with how analogy is involved in problem solving and in skill acquisition. That is, how do people call up an analogous experience to help them solve new problems? This experience can come from their own past (when analogy is sometimes called repetition), it might come from looking at the behavior of another (in which case it is sometimes called imitation), or it might come from adapting an example given in a textbook or some other expository medium (in which case it is sometimes called copying).