Chapter 5, "Emergent Properties and Connectionism", introduces a (then) novel way of thinking about modeling cognition by borrowing the basic architecture and functional principles from biological neural networks. This approach effectively overcomes a number of critical issues associated with the view of the mind as a substrate-independent process of symbolic manipulation, and is supported by developments in the field of the nonlinear dynamics of self-organizing systems. These have provided an apt theoretical framework for investigating the emergent properties that are a signature of densely connected networks, and that do not sit well with the symbolic paradigm of classic cognitivism. The chapter ends up considering the possibility that these alternative perspectives may actually represent two levels of description: sub-symbolic/connectionist for the dynamics of the lower levels, and symbolic for the higher levels (e.g., language). We will review these topics, discussing how recent theories about sentience and cognition appear to have incorporated not only the connectionist approach, but also some core ideas of autopoiesis and enaction.

свернуть