COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	intend[w87,jmc]		x intends to do y - for w87 cs326 exam
C00006 ENDMK
C⊗;
intend[w87,jmc]		x intends to do y - for w87 cs326 exam

The first question is whether "x intends to do y", symbolized intends(x,y)
is to be defined or merely axiomatized.  I think it's better to do the latter,
especially since we aren't working from a mutually understood base
of more elementary terminology.

We need the relation between  intends(x,y)  and  willdo(x,y).  Neither
should imply the other.  Tom may intend to do  y,  but we may know
of a circumstance that will prevent his doing so.  Again Tom may
not intend to do  y,  but we may know of a circumstance that will
force him to.  How about the relation between  intends(x,y)  and
believes(x,willdo(x,y))?  This is closer, but  x  may intend to do  y
and still not assent to  willdo(x,y), because  x  may recognize that
circumstances may occur that will prevent his doing  y.  How about
something like  x  believes that  x  will do  y,  provided  x's  expectations
about what else happens are met?  Better yet, but  x  doesn't claim
infallibility and so doesn't expect that all his expectations will be
met.  We must consider only relevant expectations.

Also there are two reasons why  x  may not carry out his intentions.
First something may prevent it.  Second something may occur that causes
x  to change his mind about whether doing  x  is a good idea.

Here is a preliminary axiomatization leaving several things vague.

intends(x,y) ⊃ believes(x,"z ⊃ willdo(x,y)") ∧ expects(x,z)

believes(x,"z ⊃ willdo(x,y)") ∧ expects(x,z) ∧ ¬ab1(x,y,z) ⊃ intends(x,y)

Some of the vagueness is nicely handled by ¬ab1(x,y,z).  There could
be additional circumstances that make intending not quite equivalent
to believing that one will do something provided one's other expectations
are met.  However, leaving out anything like situations from the
axioms means that they will require large modifications before they
become usable in a common sense database.  The difficulty in including
these in the axioms for now arises from the lack of a good formalism
for putting sentences involving  result(e,s)  as an object of belief.