Professor Judea Pearl
Cognitive Systems Laboratory
Computer Science Department
UCLA
Los Angeles, CA 90024-1596

Dear Judea

	The phrase ``the logical formulation of the episode'' on page 3
of your ``A Probabilistic Treatment of the Yale Shooting Problem''
wrongly supposes that using logic determines a specific logical formulation.
In fact a logical language is determined by the predicate and function
symbols in it, and even when a language is chosen, there may be more than
one formulation of the facts about something, e.g. about an episode.  It
would have been ok to refer to ``several logical formulations'' or
to ``the most commonly used logical formulations''.  Indeed Lifschitz
has given at least two reformulations that avoid the ambiguity and
so have others.

	I am protesting this mistake, because it is so common, and it has
the effect of discouraging students and other naive people from inventing
new logical formulations when the ones they have turn out to be unsatisfactory
for one reason or another.

	Of course, a major goal of research on the use of logic in AI is
to find a universal language for expressing common sense knowledge that
would not be subject to such unpleasant surprises when a new problem
comes along, and it is important to point out that this goal has not
been achieved.

	If I have comments on the rest of the paper I'll write.

Sincerely,