COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	.<<mcderm[s84,jmc]		McDermott example? done by circumscription>>
C00005 ENDMK
C⊗;
.<<mcderm[s84,jmc]		McDermott example? done by circumscription>>
.require "memo.pub[let,jmc]" source;
.cb McDermott's night watchman problem

	According to Drew McDermott, some monotonic formalisms
have trouble expressing the following rules.  The lights, once
turned on, are presumed to stay on unless there is information
that they are turned off.  Likewise when turned off they are
presumed to stay off until they are turned on.  Sometime after
midnight, the night watchman is presumed to turn them off.
Here is a formalization using the %2ab%1 predicate of my
1984 paper 
"Applications of Circumscription to Formalizing Common Sense Knowledge"
soon to be submitted for publication in %2Artificial
Intelligence%1.

	%2∀t0 t1.¬ab aspect1(t0,t1) ∧ t0 ≤ t1 ∧ light-on(t0) ⊃ light-on(t1)%1.

This expresses the presumption that the light stays on if on.

	%2∀t0 t1.¬ab aspect2(t0,t1) ∧ t0 ≤ t1 ∧ ¬light-on(t0) ⊃ ¬light-on(t1)%1.

This expresses the presumption that the light stays off if off.

	%2∀t.¬ab aspect3(t) ⊃ ∃t1.0 < t1 - t < 24 ∧ 0 ≤ hour(t1) ≤ 1 ∧
¬light-on(t1)%1.

This expresses the presumption that the light is turned off sometime
between every midnight and 1am.

	As expressed here, this requires prioritized circumscription in
order to avoid giving equal weight to the %2sneak%1 interpretation that 
the light stays on so that %2aspect1(t,t1)%1 is normal and