Adding "believes" and propositions to KIF?

macgregor@ISI.EDU
Message-id: <199304131905.AA05398@quark.isi.edu>
Date: Tue, 13 Apr 1993 12:05:19 -0800
To: interlingua@ISI.EDU
From: macgregor@ISI.EDU
Subject: Adding "believes" and propositions to KIF?

In reviewing the KIF document, I find several examples of the use
of "believes", but nothing indicating that "believes" is an integral
part of the KIF language.  This raises several questions:

(1) Should "believes" be a built-in predicate in KIF?

(2) If the answer to (1) is yes, should the meaning of "believes" in
KIF be fixed (to whatever degree is feasible), or should "believes" be 
primitive?  

Note that even if we fix the meaning of "believes" in KIF, effective
communication requires that users of KIF are free to ignore the KIF
meaning when transmitting statements of belief derived from their own
KBs. Put another way, for purposes of communication, it may not matter
very much exactly what system of belief is built into KIF.

However, specifying the FORMAT of sentences passed as arguments to
"believes" may be important. If we review the example that started much
of the current discussions:

>    (forall ((?a block) (?b block))
>            (=> (on ?a ?b)
>                (believes Tom `(smaller_than ,?a ,?b)) ))
> 
> However, Mike said that a simple comma is not sufficient and
> that the last line of this statement should be
> 
>                (believes Tom `(smaller_than ,(name ?a) ,(name ?b))) ))

we can ask ourselves whether we should be discussing the semantics of
"believes" (possible worlds etc.) or the syntax of the arguments to
believes. Thus, another question is

(3) How sensitive should the KIF "believes" relation be to syntactic
variations in its arguments?

I suggest that before jumping into semantics, the ground rules for the
discussion (the answers to questions 1 and 2, and possibly 3) ought to
be clarified.

Some of the net discussions have moved from the question of belief to
the question of what is a proposition. I would like to point out that
at the moment the notion of a proposition is also absent in KIF. So we
have

(4) Should propositions be built in to KIF?

The answer I have received in the past to (4) is "No, sentences are
sufficient and there is no need for introducing propositions." Or
possibly, the answer I received was "Yes, KIF has sentences, and we
should not distinguish between sentences and propositions (so in
effect, KIF already has propositions)." There is a communication gap
here, because the receiver of the prior communications (me) believes
that sentences and propositions ought to be treated as distinct
entities, but its not clear to me that the senders of those
communications share my belief. 

I think that before embarking on extended discussions about
propositions, we should reach a consensus on whether or not they should
exist (in KIF) as entities distinguishable from sentences.

- Bob

Robert M. MacGregor                                     macgregor@isi.edu
USC/ISI, 4676 Admiralty Way, Marina del Rey, CA 90292      (310) 822-1511