Classical Questions, Radical Answers: Connectionism and the Structure of Mental Representations

Abstract

The many and obvious contrasts between Connectionism and (something like) “most other work in Cognitive Science” often give rise to rather grandiose talk of scientific upheaval. Thus it is common to hear the proposal that Connectionism represents (or is the beginning of, or at least points in the general direction of) a whole new approach and perhaps even a whole new paradigm in the study of cognition, an approach which directly contests and may even be displacing (something like) the dominant symbolic view. This is an exciting prospect, especially for those philosophers and others who prefer to view such topics as artificial intelligence and cognitive psychology in terms of the grand sweep of intellectual history. Unfortunately, despite the endless tabloid observations, nobody yet seems altogether sure just what Connectionism is, in any sense going beyond the standard fare of dull details and techno-truisms about the nature of typical Connectionist networks. Indeed, in discussion to date there is very little agreement even over how the phenomenon itself is best described: some regard Connectionism as nothing less than a whole new paradigm, while others-using progressively more vague terminology-have described it as a radically new architecture, theory, framework, model, approach, and simply as a new view.

Communication across the revolutionary divide is inevitably partial.

Kuhn (1962) p.149.