Hayesism vs. Lehmannism

fritz@rodin.wustl.edu (Fritz Lehmann)
Date: Fri, 11 Feb 94 03:06:50 CST
From: fritz@rodin.wustl.edu (Fritz Lehmann)
Message-id: <9402110906.AA05803@rodin.wustl.edu>
To: cg@cs.umn.edu, interlingua@ISI.EDU, martin@cs.ucla.edu
Subject: Hayesism vs. Lehmannism

Dear David Martin:

     You wrote:
>Date: Fri, 21 Jan 94 11:44:35 -0800
>From: martin@CS.UCLA.EDU (david l. martin)
>To: cg@cs.umn.edu, interlingua@ISI.EDU, phayes@cs.uiuc.edu
>Subject: Re:  Higher-order KIF & Conceptual Graphs
[quoting me]
>>     Also the committee might initially address our earlier
>>"Hayes/Lehmann dichotomy" between a system with only fully
>>logically interpreted predicates, for an entire agent's
>>"semantics", and a system combining logically interpreted
>>predicates with "primitive" predicates (like Hayes' "VERY-BIG")
>>interpretable only outside the system, for practical knowledge

>Fritz -
>Could I ask for a clarification of the above notion of a system
>"with only fully logically interpreted predicates"?  

     I failed to make it clear that I am the one who favors having some

uninterpreted predicates in a knowledge interchange language, which get their

"semantics" only by human understanding or maybe "operationally" in a robot.

Pat Hayes was the one who _complained_ about  a student saying that a 
predicate like VERY-BIG has a meaning without providing Tarskian model-
theoretic semantics for it (i.e. in terms of operations on sets of objects in

the universe of discourse).  That phrase you quote was my attempt to state 
his position in a short phrase.  We had a voluminous email exchange on this

subject and ultimately agreed about our difference.  I'd rather let Hayes 
speak for himself than try to explain his position.  It is much more 
ambitious than mine since it seeks to ground all conceptual meaning in logic.

I don't believe this is possible even for natural numbers without cheating

(stealing semantics from the descriptive syntax).

>Surely it's always necessary to have at least a few primitive
>predicates.  Even if everything is defined in terms of set theory,
>one begins with a few primitive predicates describing sets.

     Yes indeed, that's my view, and those "primitives" are intensional.

>What about a system which wants to use sense-description predicates
>such as "red" - is there any way in which these sorts of predicates
>can be fully logically interpreted predicates?

     Maybe Hayes can devise a way.  I can't.

>Probably I just don't understand the sense in which you are
>thinking of "fully logically interpreted predicates".  I am
>thinking of predicates which are defined in terms of other
>predicates using a "if and only if" connectives.

     I leave it to Pat Hayes to explain his position if he wants to.
I'll send you (by private email) copies of a couple messages left on my hard

disk from our earlier discusion of this topic.

>Thanks for your comments.
>- Dave Martin
>- martin@cs.ucla.edu

                             Yours truly,  Fritz Lehmann
4282 Sandburg Way, Irvine, CA 92715  714-733-0566  fritz@rodin.wustl.ed