Melanie Mitchell

Emergent Computation and Representation
in Dynamical Systems

In recent years, a number of cognitive science researchers have questioned the relevance of notions such as "information processing", "computation", and "representation" for understanding cognition, and instead propose the language of dynamical systems theory as a more useful foundation for cognitive science. In this talk I will review arguments on both sides of this ongoing debate, and will also review some results on new notions of "emergent" computation in dynamical systems. These results will be used to present a simple example in which computation and representation -- defined in a nonstandard sense -- are seen to be key in understanding the emergent adaptive behavior of a particular class of dynamical systems that were evolved by a genetic algorithm to perform specific tasks. It is hoped that the analysis of such examples will help forge a new, unifying framework for cognitive science, encompassing both dynamical and (albeit nonstandard) computational approaches.

Return to Representation Day Page