Path: utzoo!attcan!uunet!husc6!bloom-beacon!bu-cs!mirror!rayssd!raybed2!linus!mbunix!bwk
From: bwk@mitre-bedford.ARPA (Barry W. Kort)
Newsgroups: comp.ai
Subject: Re: Feeling and thought: which comes first?
Message-ID: <42902@linus.UUCP>
Date: 11 Dec 88 16:35:17 GMT
References: <2609@datapg.MN.ORG> <4524@homxc.UUCP>
Sender: news@linus.UUCP
Reply-To: bwk@mbunix (Barry Kort)
Organization: Electronic Peaceware Project
Lines: 31

I enjoyed Marty Brilliant's model of feelings vis-a-vis goals
and reasoning.  I think he is on the right track.

I like to draw a technical distinction between "emotions" and
"feelings".  To me, emotions are the names of mental states,
while "feelings" refer to sensory-neural information.  The two
are frequently related so closely that it is hard to tell the
difference.  The mental state of fear is often accompanied by
reduced body temperature.  Shake hands with someone who is afraid
and you will feel a cold limp hand.  Anger is often accompanied
by secretion of adrenalin, and an increase in metabolic rate.
Curiosity, fascination and puzzlement are emotional states
frequently experienced by scientists.  Anxiety is an emotional
state experienced by students and novices in a public learning
situation.  I distinguish the mental state of anxiety with the
sensory-neural response of tightness in the stomach (butterflies).

A computer will probably not experience human-like somatic stress
responses to emotional states.  But a learning machine will likely
have emotional states corresponding to curiosity, puzzlement,
frustration, satisfaction, etc.  Frustration will increase until
a computer abandons an unachievable goal.  Satisfaction will
trigger a decision to accept a solution and to install it 
permanently into the compiled knowledge base.

To me, the above model seems logical, and it matches my own direct
experience, as well as successful approaches to AI.  Certainly the
ambulatory robots of MIT exhibit behavior consistent with the
above model.

--Barry Kort