Rules and Representations in the Classicism-Connectionism Debate
Dissertation, The University of Western Ontario (Canada) (
1998)
Copy
BIBTEX
Abstract
This dissertation is a work in the philosophical foundations of cognitive modelling. To a significant extent, it is presented as a response to a critique of connectionist modelling originated by Jerry Fodor and Zenon Pylyshyn. The essence of the critique is that either connectionist models implement classical models of cognition, or if connectionist models are not implementational, then they are incapable of modelling cognition. I argue that barring an implausible interpretation of "implementation," there exists a subset of connectionist models which cannot be implementations of classical models, and while no connectionist researcher has met all the demands Fodor and Pylyshyn placed on adequate models of cognition, the properties possessed by some non-implementational connectionist models provide reason to think that these demands can be met in principle, and that connectionist models are capable of kinds of processing which classical models are not capable of, processing which makes connectionist models worth developing in spite of some of their present limitations. ;Many have responded to the Fodor and Pylyshyn critique. However, the responses have not payed sufficient attention to key notions such as implementation and rule. One of the principal contributions of this dissertation will be an analysis of the notion of implementation. The notion of a tacit rule also will be defined. Many have suggested that the difference between connectionist and classical models of cognition is that connectionist nets do not have explicit rules. I argue that the preceding is incorrect; I also argue that there is a certain kind of rule which connectionist nets cannot be understood as implementing--a tacit rule. ;Chapters one through four respond to the Fodor and Pylyshyn argument against connectionism. Chapter five takes up a critique of connectionist models made by Andy Clark. Chapter six responds to an argument by Terrence Horgan and John Tienson which aims to show that folk psychological laws make use of a type of ceteris paribus clause not possessed by physical laws. Horgan and Tienson are shown not to have made a persuasive case