COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	\input let.tex
C00010 ENDMK
C⊗;
\input let.tex
\jmcletterhead
\vskip 12pt
\rm

\rjustline{August 10,1981}

\vskip 30pt

\adx 0pt:Professor Steven Stich\cr
Department of Philosophy\cr
University of Maryland\cr
College Park, Maryland 20742\cr

\parskip 11pt minus 6pt

Dear Professor Stich:

	It is somewhat past time for me to have sent you the paper
on which you are to comment at the conference on philosophy and artificial
intelligence.  Unfortunately, I have had problems fulfilling the
promises of the abstract, and it looks like the presentation will
have to be rather informal.  I suppose you have a copy of the abstract,
but I enclose another copy just in case.

	I heard that you had a fall back plan to suppose that my Vancouver
paper would be a continuation of my previous paper on circumscription in
{\it Artificial Intelligence}.  This is essentially correct, but I
will also use or at least mention the formalism of my `First Order
Theories of Individual Concepts and Propositions', a copy of which
is enclosed.

	My present intention is to base my presentation on the following
propositions; you will doubtless have your own opinion of their
correctness.

\display 40pt: 1.: The concepts used by humans are more often than not ambiguous.
Children do not begin by learning definitions or even sets of axioms.

\display 40pt: 2.: That there should be such ambiguities is an essential
part of the epistemological situation in the common sense world.
Computer programs will have to use ambiguous concepts even apart
from their need to communicate with humans, because they also
must use them before the information necessary for precise definition
is available.

\display 40pt: 3.: Artificial intelligence must accept modest goals for
its formalisms at least to begin with.  Thus we need formalisms
adequate to express and reason about what travel agents know
about how to travel by airplane.  Even if the formalisms give problems
when extended to other domains or even if they can't deal with
exotic questions about air travel, they may still be appropriate.

\display 40pt: 4.: Accepting the existence of ambiguity does not require
giving up formal methods.  In fact, the point of the paper is that
circumscription provides a clue on how to treat ambiguous concepts
formally.

\display 40pt: 5.: My `First Order ...' uses distinct expressions for
concepts and objects.  In the Vancouver paper I want to explore
the possibility of a language that withdraws the distinction, i.e.,
which is ambiguous between sense and denotation.  Perhaps it can
be interpreted as a return to Frege's idea that sometimes the
meaning of an expression is its sense and sometimes its denotation.
If you think I have misunderstood Frege, plese say so, but don't 
make too much of it, because the paper is not intended as an
exegisis of Frege).

\display 40pt: 6.: Some of the formal examples will be elaborations of the following
sentences:

\display 65pt:{ }: {\sl knows(pat,telephone mike)}

\display 65pt:{ }: {\sl dials(pat,telephone mike)}

\display 65pt:{ }: {\sl equals(telephone mike,telephone mary)}

\display 65pt:{ }: {\sl $∀x.(p(x) ∧$ equals $(x,y) ∧ ¬prevsub(p,x,y) ⊃ p(y))$,}

\display 65pt:{ }: {\sl$∀$ person $x$ $y.prevsub(λu.$ knows(person,u),$x,y)$.}

	Circumscribing the predicate {\sl prevsub} in the above
sentences will then allow one to conclude that Pat dials Mary's
telephone number but not that Pat knows Mary's telephone number.
The general facts about substitution prevention are given in a way independent
of what specific substitutions are forbidden, e.g., some people
might be unaware at first that the second argument of {\sl knows}
does not allow substitution.  In fact, arriving at a contradiction
the last sentence about substitution not being allowed in knowledge
statements may not be present.  The

\display 65pt:{ }: {\sl $prevsub(λu.$ knows (pat,u),telephone,mike,telephone,mary)}

may be be deduced by arriving at
a contradiction with {\sl ¬knows(pat,telephone mary)} obtained
in some other way.  However, because {\sl prevsub} has three
arguments, this will not prevent attempts to apply substitution
in other instances of knowledge.  This may be good or bad.

	The precise notation may change,
since I have a prejudice in favor of first order formalisms.

	This example illustrates the slogan {\it Ceteris paribus
de re = de dicto}.

	There may be other formal examples.

	I hope this will help you decide in general what comments
you want to make.

\vskip 11pt

\adx 3in: Sincerely,\cr

\vskip 30pt
\parskip 0pt

\adx 3in: John McCarthy\cr

\vskip 15pt

\vfill\eject\end