Abstract

Similarity permeates human cognition. There is evidence that objects are categorized based partly on similarity to previous category members and that the likelihood of transfer is governed by the similarity between the original and current situations. New problems are often solved by analogy to prior problems. Similarity is responsible for many human errors, such as perceptual confusions and many recall intrusions; but at the same time, analogy and similarity are important in scientific discovery. Consequently, we are developing a cognitive architecture in which similarity computations play a central role. This is unlike most architectural approaches, which either do not treat analogy and similarity computations play a central role. This is unlike most architectural approaches, which either do not treat analogy and similarity at all, or relegate them to a subsidiary role, to be called in sporadically when other mechanisms are stuck. We are using Gentner's Structure-Mapping theory [12] as our framework for defining similarity computations.The rest of this note addresses the list of issues suggested by the symposium organizers.