Computational explorations of perceptual symbol systems theory

Abstract

The aim of this paper is twofold. First, we provide a methodological pathway from theories of situated, embodied cognition to simulations with an eye to empirical evidence, and suggest a possible cross-fertilization between cognitive robotics and psychology. Psychological theories, in particular those formulated at an abstract level, include models which are often severely underspecified at the level of mechanisms. This is true in the synchronic, constructive perspective (how can the effects observed in experiments be concretely generated by the model's mechanisms?) and in the diachronic, developmental perspective (how can such mechanisms be learned and developed?). The synthetic method of artificial cognitive systems research, and in particular of cognitive robotics, can complement research in psychology (and neurosciences) by exploring the constructive and developmental aspects of theories. Our second aim is to provide an example of such a methodology by describing simulations aiming at developing a perceptual symbol system (PSS) ( Barsalou, 1999). We then describe the two main theoretical constructs of the PSS, perceptual symbols and simulators, illustrate their development in an artificial system, and test the system in prediction, categorization, and abstraction tasks.