perm filename MENTAL[S85,JMC] blob
sn#807036 filedate 1985-09-11 generic text, type C, neo UTF8
COMMENT ⊗ VALID 00002 PAGES
C REC PAGE DESCRIPTION
C00001 00001
C00002 00002 Mental situation calculus
C00010 ENDMK
C⊗;
Mental situation calculus
The idea is to apply the situation calculus to mental situations.
In a mental situation the subject, man or machine, has certain beliefs,
desires and intentions. When a mental event takes place, a new
mental situation results with new beliefs desires and intentions.
An important class of mental events is the class of mental actions.
We can regard the subject's mental situation either as an autonomous
system or not. Regarding it as an autonomous system means giving
rules that determine its mental actions as a function of the mental
situation and also consider inputs as giving rise to mental events.
The other point of view considers the effects of mental actions
without considering how they arise.
What mental fluents are there? What mental events are there?
What is the effect of the mental events? What mental laws are
there determining what mental events will occur? What deontic
mental laws are there determining what mental events should occur.
Mental fluents.
goal(p,ss)
believes(p,ss)
db(p,ss)
canachult(p,ss)
should(a,ss) where a is an action
Mental events
learns-from-outside(p)
deduces(p)
discovers(λx.p(x))
Mental actions
tries-to-find(λx.p(x))
Mental objects
Just as there are towers in the blocks world, there should be complex
mental objects.
1985 sep 11 - Notes on mental situation calculus
1. A mental situation ss contains a database of beliefs p satisfying
believed(p,ss). However, this need not be all there is to a mental
situation. What else? goals, hopes, wishes.
Suppose even that one is merely attempting to stack some blocks. Besides
the facts about a particular situation there is also the goal, i.e. the
fact that it is the goal, the principle of rationality, execute a plan
that achieves a goal. Facts about plans and partial plans, facts like
the one about building towers from the bottom up.
2. Mental events occur. These include inferences, perhaps taking back,
observations. Perhaps physical events occur through willings. Also
physical events occur and give rise to new mental situations.
3. Exercise. Describe reason maintenance in the mental situation calculus.
4. A major objective of the mental situation calculus is to describe
control of reasoning. Mental goals are achieved by sequences of
mental actions. A planning program can plan them. One such goal
is determining a strategy for achieving a physical goal. Such a
strategy can be far more complex than a simple sequence of actions.
Moreover, even if it is a simple sequence of actions the process of
finding it can be more complex than straightforward reasoning in
about the effects of individual physical actions.
5. The mental actions performed are mainly hill climbing in
mental situation space. We have a predicate better(ss1,ss2), and
we make moves to better mental situations. The other thing that is
done is to carry out plans that we are committed to. Perhaps another kind of
component of a mental situation is a commitment. Making a commitment
is an elementary mental action, but it is then carried out subconsciously.
It would be well if mental actions were so elementary that only a small
number were possible in any situation. Then the algorithm could be
to generate them all and choose the best. This requires being able
to compare the alternatives. An even simpler mechanism involves
a predicate that only allows comparison with the present state. It
then does the first generated action that makes any improvement.
6. There is a component called NOW. Physical events change what
mental structure it is attached to. Any observed physical event
produces a corresponding mental event. Such events may make
the mental situation worse. (Perhaps lower level mental events, e.g.
carrying out a computational commitment, can make the higher level
mental situation worse).
However, suppose I am thinking about a chess move while physical
events occur in the background. They don't affect the situation
on the chess board or the mental situation of thinking about it.
It is an interesting question to ask what are the chess mental
situations.
7. Hill climbing in physical space has to be a subcase of hill-climbing
in mental space, i.e. when we go to a higher conceptual level we mustn't
lose our lower leve capability. Indeed, at the higher level, lower
level hill climbing should be the default case. If there is an
immediate lower level good move, we make it.