COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	%business letter outline to use with buslet.tex macros
C00005 ENDMK
C⊗;
%business letter outline to use with buslet.tex macros
\magnification =\magstephalf
\input buslet[1,ra]
\def\disleft#1:#2:#3\par{\par\hangindent#1\noindent
			 \hbox to #1{#2 \hfill \hskip .1em}\ignorespaces#3\par}
\def\display#1:#2:#3\par{\par\hangindent #1 \noindent
			\hbox to #1{\hfill #2 \hskip .1em}\ignorespaces#3 \par}
\def\adx#1:#2\par{\par\halign{\hskip #1##\hfill\cr #2}\par}


\jmclet

\vskip 30pt
\address 
Dr. Maria Klawe
K53-801
IBM Research
650 Harry Road
San Jose, CA 95120-6099
\body
Dear Dr. Klawe:

	This is in support of the proposed IBM award for the work
of Ron Fagin, Joe Halpern and Moshe Vardi on reasoning about knowledge.
I worked off and on during the 60s and 70s developing formalisms
for representing information about who knows what that would be
useful in AI and database systems.  For example, it is necessary
to represent the fact that travel agents know about airline schedules.

	One of the problems I worked on the the middle 1970s was
developing a model theory of the knowledge of several agents based
on building up levels of knowledge, e.g. ``S1 knows that S2 does not
know p'' is one level higher than ``S2 knows p''.  I did not succeed
in developing such a model theory and was pleased when Sato proposed
using a system analogous to Kripke semantics.

	Fagin, Halpern and Vardi have developed a semantics for knowledge
based on building up levels.  Since I was unsuccessful at this, I
respect their accomplishment.

	I will further remark that the whole subject has considerable
importance both for databases and for artificial intelligence.

\closing
Sincerely,    

John McCarthy    
\annotations
\vskip 1in
JMC/ra 
%\smallskip
%Enclosure
%\smallskip
%cc: Matthew Kahn
%\smallskip
%\ps
%P.S.: whatever you wish to say here

\endletter

\makelabel
\end