Discrimination nets, production systems and semantic networks: Elements of a unified framework

A number of formalisms have been used in cognitive science to account for cognition in general and learning in particular. While this variety denotes a healthy state of theoretical development, it somewhat hampers communication between researchers championing different approaches and makes comparison between theories difficult. In addition, it has the consequence that researchers tend to study cognitive phenomena best suited to their favorite formalism. It is therefore desirable to propose frameworks which span traditional formalisms. In this paper, we pursue two goals: first, to show how three (symbolic) formalisms widely used in theorizing about and in simulating human cognition—discrimination nets, semantic networks and production systems—may be used in a single, conceptually unified framework; and second to show how this framework can be used to develop a comprehensive theory of learning. Within this theory, learning is construed as (a) developing perceptual and conceptual discrimination nets, (b) adding semantic links, and (c) creating productions. We start by giving a brief description of each of these formalisms; we then describe a theoretical framework that incorporates the three formalisms, and show how these may coexist. Throughout this description, examples from chess, a highly studied field of expertise and a classical object of study in cognitive science, will be provided. These examples will illustrate how the framework can be worked out into a more detailed cognitive theory. Finally, we draw some theoretical consequences of the framework proposed here