COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	textua[f86,jmc]		Textual circumscription
C00005 ENDMK
C⊗;
textua[f86,jmc]		Textual circumscription

	Consider the Yale shooting problem.  We propose now to solve
it in two stages.  Note that an event that might kill Fred is
mentioned --- namely the shooting.  Nothing that might unload the
gun is mentioned.  It seeems to me that, intuitively, this is the
essential difference between the two models that makes us prefer
the scenario in which Fred dies.  The present forms of minimization
don't capture the idea of what is mentioned.  Our two proposed
stages are as follows.

1. Stage 1 is an inference based on a new kind of reflexion principle.
It's result is a statement that the mentioned causes of events are
precisely as follows.  It is not conjectural, because it refers to a definite
text --- what has been stated about the incident.  However, it is
non-monotonic in that text.

2. Stage 2 is a minimization of unmentioned events.  It is conjectural
and also non-monotonic.  In this case the minimization results in
there being no unmentioned events.  Shooting is mentioned and unloading
is not, so we conclude that the gun was not unloaded.

The immediate problem is the formulation of the concept of mentioned
events.  It may be similar to keep the permanent information, i.e.
the axioms separate from the facts about the particular situation
and apply the concepts only to the particular facts.

Let's try

Yaledb = {Holds(Alive,S0),Holds(Loaded,S0),Occurs(Shoot,Result(Wait,S0))}

or

Yaledb = {Holds(Alive,S0),Holds(Loaded,S0)}

or maybe only

Yaledb = {Holds(Alive,S0)}.

Infer(Yaledb,Holds(Not Alive(Fred),Result(Shoot,Result(Wait,S0))))

How about the conclusion

Describe(C0,Yaledb) ⊃ 
	Holdsc(Holds(Not Fred,Result(Shoot,Result(Wait,S0))),C0)

or

Yaledb ⊃⊃ Holds(Not Alive,Result(Shoot,Result(Wait,S0))).