COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	.require "memo.pub[let,jmc]" source
C00007 ENDMK
C⊗;
.require "memo.pub[let,jmc]" source
.cb Bill wants to have an affair with Tom's wife

	We are surprised when the ambiguity of the above sentence
is pointed out.  Our common sense reaction to the ambiguity and
present philosophical reactions are quite different, and
I think the common sense reaction is better.  The common sense
reaction is localized to the sentence in question, we try to
determine which reading is appropriate if there is some reason
to suspect that Bill may not both know the woman by sight and
know that she is Tom's wife.  Having decided the case in point,
common sense is finished with the sentence.  The philosophical
reaction, which is also the reaction of AI researchers, is to
devise a formalism in which the two readings are expressed by
different formal sentences.

	The philosophical approach would be fine if it we could
solve the problem once and for all.  However, it may be that ever
more complicated examples of this kind could be devised.  The
approach of making basic revisions in the syntax, semantics and
ontology of the logic of intensions seems to become increasingly
pedantic, and we are led to spend our time on examples that
never arise in common sense contexts.

	Maybe circumscription or some other form of non-monotonic
reasoning can make the common sense approach work, but I don't
have a detailed proposal.  The idea is the following:

	Should it be the case that Bill wants to have an affair
with the woman and knows she is Tom's wife and the speaker does
not wish to commit himself on whether the fact that she is Tom's
wife contributes to Bill's wish, then the sentence is not
%2ambiguous in the context%1, both
readings are true in all possible worlds in which these other
facts are true.  Therefore, we try to say that %2the readings
are equivalent in the world unless there is a reason why they
aren't%1.

	Making this work requires solving two problems.  First, we
may need to apply circumscription to meaning itself, i.e. two
phrases have the same meaning unless there is some reason why
not, and I haven't seen how to write this formally.  Second,
if we are to follow common sense, the ambiguity shouldn't even
be noticed unless there is a reason to do so.  This suggests
a formalism in which the sentence implies the the consequences
of both readings unless there is a reason why not and is implied
by anything that will imply either reading unless there is a
reason why not.  A computer program that worked this way would
never produce two versions of the sentence unless it had to.

	The need for this approach would be obvious if we could
find examples in which the modal approach and introducing concepts
as objects both led to infinite regress.