Newsgroups: comp.ai.philosophy
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!gatech!swrinde!pipex!uunet!swissbank!root
From: gerryg@il.us.swissbank.com (Gerald Gleason)
Subject: Re: Thought Question
Message-ID: <1995Jan24.210647.4914@il.us.swissbank.com>
Sender: root@il.us.swissbank.com (Operator)
Nntp-Posting-Host: ch1d264nwk
Organization: Swiss Bank Corporation CM&T Division
References: <3fpdjn$9pj@mp.cs.niu.edu>
Date: Tue, 24 Jan 1995 21:06:47 GMT
Lines: 28

Neil Rickert writes
> In <1995Jan20.204411.19044@galileo.cc.rochester.edu>  
stevens@prodigal.psych.rochester.edu (Greg Stevens) writes:

> >Neural networks would account for learned (non-innate) heuristics by
> >associative mechanisms that are not propositional.  "Most of the time
> >X is linked with Y, so the next time X happens, let's link it with Y!"

> Then I guess we are using the word "heuristic" in different ways.  If
> neural networks, operating as you suggest, are responsible for much
> of our decision making, then we would be talking about a general
> procedure used throughout.  It seems strange to apply the word
> "heuristic" in that case.  I tend to think of "heuristic" as
> referring to an ad hoc procedure which is used for a limited range of
> cases.

I guess I made this same point earlier.  On the other hand, you could say  
that there is a general strategy to find heuristics that work.  Extending  
this further, you could say that the system is using heuristic processes  
in the general strategy as well.  So, there could be a general process in  
which every particular instantiation of the process exibits ad hoc  
behavior because of its unique experiences during developemnt.  The  
existence of widespread regularities between many similarly structured  
systems doesn't necessarily mean that the basic process is not ad hoc,  
just that the system has stable atractors that produce similar features in  
the behavior of systems with that type of structure.

Gerry Gleason
