COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	mental[s87,jmc]		Mental situations
C00008 ENDMK
C⊗;
mental[s87,jmc]		Mental situations

One form of mental situation consists of a physical situation plus a
partial goal tree.  It is simplest if the goal tree consists entirely
of AND goals.  It is an improvement of the situation to add a goal
to the tree, i.e. by finding another necessary subgoal for the
achievement of a goal in the tree.  It is also an improvement to
achieve a goal.  Under some conditions improvements are always
possible and sufficient improvement will achieve the main goal.
Presumably the initial mental situation consists of a description
of the physical situation coupled with the main goal.

We need to find the simplest example of the use of mental situations
that goes beyond using physical situations and actually helps.

Try it in the simple blocks world.  Towers of Hanoi provides a
more sophisticated example.

More generally, a mental situation includes a description of a
physical situation plus additional data structure.  It is always
just a partial picture of the real mental situation and subject
to amplification by reification of some of the assertions into
objects by means of introspection.  In the simplest case, the
description of the physical situation is complete, i.e. the space
of possible descriptions is isomorphic to the space of physical
situations.  At the next level the space of descriptions of
physical situations is a homomorph (homomorphic image) of the space
of physical situations where the homomorphism also maps physical
actions onto actions in description space and there is also a mapping
backwards that is 1-1 but not usually onto.

One kind of improvement is to simplify the goal structure by contraction,
e.g. to do things that have to be done or might as well be done.

So what are the mental moves?  Does one compare various mental
moves to see what is the best or does one simply choose one that
makes an improvement?  The latter is presumably simpler and
perhaps computable by logic programming.

What about advice to simplify the goal structure before proceeding
to physical action?  How is that expressed and when does it take
effect?  In general, how is heuristic advice expressed and how does
it take effect?  Perhaps there is built in to the algorithm a step
that says ``apply applicable advice''.

To what extent can the mental situation consist of sentences?  What
is the relation of mental moves to deduction or circumscription.

For the time being we'll consider that we just want a better situation ---
not the best immdediately realizable situation.  So we have the
Prolog goal  $better(result(a,s),s)$, where  $s$ is known and $a$
is to be found.

	We need to disentangle the following idea.  A mental situation
is better if the $better$ predicate is improved.  It seems likely that
we will want to look for a better $better$ in a mental space meta to the
space in which $better$ is used and therefore, the new mental situation
uses a predicate $meta\-better$.

better(s1,s2) ≡ physically\-better(phys s1,phys s2) ∨ ∃g g1.s2=adj(g,s1)
∧ isgoals(g1,s1) ∧ newsupport(g,g1,s1)

Suppose we try to use an idea similar to my Prolog introspection.
We have a goal and a predicate $better$ and a situation.  We try to
find an action that gives a better situation.  If we succeed fine, and we
continue the process.  If not we introspect and try to improve in
the mental situation that results from introspection.  The question
is what does the introspection give us.  In some sense it attaches
know predicates to the known facts and generates the goal of knowing
a way to achieve the original goal.