Re: The Frame Problem

From: Harnad, Stevan (
Date: Sat Mar 08 1997 - 19:59:36 GMT

> From: Minnett-Westwood, Della <>
> As I understand it, the essence of the Frame problem is that if a
> symbolic computer had been programmed to recognise, for instance,
> different shapes composed of straight lines, and suddenly it was
> presented with a circle, it would "crash" because it could find no
> rules in its program for recognition of curves. A human being, on the
> other hand, would be able to infer from existing knowledge that the
> circle was still a shape, composed of a continuous line, and would
> expand their current rules of classification of shapes to encompass the
> circle. i.e. the "mind" is malleable by internal mechanisms - a
> symbolic computer is externally controlled and therefore incapable of
> self-change.

So far what you have described is just a system that "learns" (i.e.,
can change as a result of receiving certain data in the past in
such a way as to respond "correctly" with new data in the future).

Symbol Systems have great difficulty, when they "learn" of
a new action, to deduce what changes and what stays the same.

For a summary of how the discoverer of the Frame Problem,
Pat Hayes, describes it, see:

and find the frame problem using the index or a keyword search.

This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:51 GMT