Modeling cognition

cybernetics and information at the origins of cognitive science and in the on the pathway to artificial intelligence

Abstract

This article proposes a complementary approach to Jean-Pierre Dupuy’s study in On the Origins of Cognitive Science, in which the author analyzes how cybernetics provided fundamental models for the construction of a general science of the cognitive processes. Starting from Dupuy’s premise that to know is to produce models capable of structuring operations on phenomena and functions, we examine the main cybernetic models that underpinned the consolidation of the cognitive sciences: the Turing machine, the feedback model of Rosenblueth, Wiener, and Bigelow, the neural networks of McCulloch and Pitts, and the computational architecture of von Neumann. We show how these models made it possible to conceive cognitive operations as analogous or functionally equivalent to the logical-symbolic operations performed by artificial and natural machines. We then analyze the reasons for and consequences of the absence, in Dupuy’s investigation, of a systematic consideration of the digital information model formulated by Claude Shannon. We argue that Mathematical Information Theory was fundamental not only to problems of system organization, but above all to problems of communication among brains, machines, and cognitive systems, playing a decisive role in the development of the cognitive sciences and artificial intelligence. Finally, we indicate that the incorporation of this informational model would complement Dupuy’s account and allow us to understand more precisely the kind of cognition currently being promoted for contemporary artificial intelligence systems.