Re: CCAT: EMOTIONS (Pat Hayes)
Message-id: <>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Date: Wed, 23 Nov 1994 15:29:00 -0600
To: (Fritz Lehmann),
From: (Pat Hayes)
Precedence: bulk
At  3:33 AM 11/23/94 -0600, Fritz Lehmann wrote:
>....... "minimal
>set of feelings" which would be common to human beings,
>animals, bugs and maybe even robots.  This set would
>be small and and would be supplied with very reliable
>inferences -- on the order of "If creature A desires
>thing B, it will tend (aside from other considerations)
>to seek B." 

I dont think this is reliable. A might have conflicting desires, for
example, which is impossible here. And in any case the nonmonotonic
qualifiacation here is so huge and all-encompassing that it makes the rule

Ive tried to create 'simple' theories of emotions like this, and its not at
all easy to do. Its not clear that there IS a simple 'core' theory.

>     The second level of practical EMOTION ontology
>would be used to make sense of stories of human events.
>My preference would be for a _conservative_ ontology,
>that is, one with which few people would argue.  The
>facts which would be inferred from a creature having
>a certain emotion would be uncontroversial. 

But consider anger and fear. Sufficient fear will prevent the display of
behavior otherwise appropriate to anger. So there isnt a simple prediction
possible from just the fact that someone (or some creature) is angry. This
is what I meant by asking whether emotions formed a vector space. If we
just say that people can be <list emotions here>, than the situation is at

>ambitious and controversial systems of emotions would
>be deferred until later.
>     Pat Hayes suggested that we consider whether
>emotions form a "vector space"; by this I assume he
>means that each species of emotion is an independent
>"dimension" which takes on a range of (numerical)
>values.   Although this sort of thing is popular
>in interactive adventure games, or games like 
>"Careers", I would not want to adopt that as the
>central ontology-schema for emotions.  For one thing,
>I don't think the "dimensions" (anger, love, fear, etc.)
>are independent, nor necessarily linearly structured.

BUt how can they not be? If someone can be more or less angry, and more or
less afraid, etc., then there is a linear vector space being defined right
there. Maybe only a subset of it is biologically possible, and maybe
behavior is best predicted by some complicated kind of surface or subspace
in this vector space, but what other kind of description COULD be given of
emotional state, if we decide to set out by listing the emotions? What
other ways to proceed are there?

<You know, email needs a new punctuation mark. I need to be able to
distinguish a rhetorical question from a sincere, genuine question. That
was a genuine question, not a rhetorical denial.>


Beckman Institute                                 (217)244 1616 office
405 North Mathews Avenue        	(217)328 3947 or (415)855 9043 home
Urbana, IL. 61801                                 (217)244 8371 fax