tjhorton@ai.toronto.edu ("Timothy J. Horton") (03/18/88)
-------------------------------------------------------------------------- From gpu.utcs.toronto.edu!yorkvm1.bitnet!COGSCI-L Fri Mar 18 08:58:01 1988 Sender: Cognitive Science Discussion Group <COGSCI-L@yorkvm1.bitnet> From: Michael Friendly <FRIENDLY@yorkvm1.bitnet> Subject: March 25th meeting ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: | | | Cognitive Science Discussion Group | | | | Speaker : Peter Roosen-Runge (Computer Science, York) | | Title : Forward-chained vs. Backward-chained Rules: | | A Crucial Polarity in Cognitive Models | | Date : Friday, Mar. 25, 1988 -- 1pm | | Location: Rm 207 Behavioural Science Bldg., York University | | | ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: Abstract The directionality of rules in a program seems to be related to deep issues in epistemology. I will briefly sketch how the con- trast between forward- and backward-chained rules is manifested in AI software, and how it corresponds to well-known oppositions in the structure of scientific theories, in the structure of com- puter programs, and in problem-solving strategies. More speculatively, I will suggest that backward-chaining is a necessary and almost sufficient requirement for a symbol- processing system to exhibit a minimal form of "intentionality". (As an argumentative corollary: Prolog is thus much more rele- vant to cognitive science than is Lisp). If correct, this result refutes a basic premise of the highly influential Newell/Anderson approach to cognitive modelling which has been entirely based on explicitly forward-chained architec- tures.