[Heartlogic-dev] Re: goals (was Re: caught up except for?)

[William L. Jarrold]

Sorry for the delay in responding.

I have an idea.  What about with each discussion thread
(this email is part of a discussion thread) we have a draft
document being considered.  In every email that is part of
the thread there will be a pointer (a URL perhaps?) to the
document in question.  Such a reference point will be
helpful when we're having highly asynchronous discussions.

On Mon, 1 Dec 2003 address@hidden wrote:

> > On Sat, 29 Nov 2003 address@hidden wrote:
> > > Here is a brief errata:  I claim that your desirability
> > > is the same thing as my "individual intention".  This
> > > is not correct.  What you mean by desirability involves
> > > two things:
> > >
> > > 1. I want something (Sally wants orange juice.)
> > > 2. I got what a wanted (Mommy gives Sally orange juice.)
> > > or
> > > 2. I didn't get what I wanted (Mommy puts on her glasses.)
> > > or
> > > 2. <whatever>
> > >
> >
> > Right.  Desirabliity according to me (and by the way OCC) is a property
> > ascribed to an event and depends on one's goal structure.  As an
> > extremely crude approximation, we might say it is dependent on how
> > congruent a single goal is to a given event.
> >
> > > In all these examples, there are two steps.  What I mean
> > > by "individual-intention" is only step 1.  For example:
> > >
> > > 1. Billy wants to go on the train.
> >
> > hrmm, maybe "individual-intention" is the same thing as "goal" or "focal
> > goal" or "goal structure"?
>
> How about this, "individual-intention" is one of:
>
>   goal           "I want a banana"
>   anti-goal      "I don't want ..."
>   lack of goal   "I am not hungry."

Okay.  I would consider what you are doing here to be developing
the ontology of goals.  You might say you are developing the
ontology of individual intentions.  You might say that goals
are subsumed under individual intentions.  That's fine with
me at the present level of discussion.  I might become
more picky about names as we leave the brainstorming phase
and move closer to doing kr....Other lingo that is near
my "Goal" notion and near your "Individual-Intention" notion
includes "Propositional Attitude" (official jargon from academe),
and "Hedonic Disposition" (just made up by me).

What you refer to as "goal" I will refer to via "Goal-Pritikin", likewise
 "anti-goal" will be referred to via "AntiGoal-Pritikin".
So, here is a question, Does every instance of "Goal-Pritikin" have
an object?  Likewise does every instance of "AntiGoal-Pritikin"
have an object?

I am interested in the slots that instances of Goal-Jarrold might
have.  My KM model, offers some glimmers in this regard.  Much
much more work needs to be done in this area.  Tested working rules
which determine the degree to which a Goal-Jarrold or a
Individual-Intention has been met/violated/etc are essential drivers of
this part of this region of the ontology.

Joshua, one more thing, I encourage you to give at least a rough one
sentence defintion of each new term you use.  Maybe we should have a
glossary page on the website?  There is a glossary page in my diss, if
memory serves.  That might be a helpful starter.  As I recall that
glossary was pretty quickly written and needs some fine tuning.  With
luck, all such fine tuning will occur between Dec 29 and Jan 4 when
I attempt to put the final touches on my dissertation (i.e. implement
the minor revisions the committee asked for)....Of course, in all
this defining,we can't forget that some poor fool will
eventually have to write this definition in PowerLoom/CycL/KM/whatever
so be nice.

>
> > > 1. Billy is not hungry.
> > >
> > > or
> > >
> > > 1. Billy is crying.
> >
> > ...hrmm, well, this last exemplar, "Billy is crying" seems like it
> > does not fit too well with the former two exemplars.
>
> I was trying to give an example of an "anti-goal".
> Typically anti-goals occur in reaction to
> something, so it is a little bit awkward to give
> an example of an isolated anti-goal.

Maybe this is an example of an anti-goal:

Billy doesn't want a diagonally cut sandwich. (statement of anti-goal).
Mommy gives Billy such a sandwich.  (emotion eliciting situation)
Billy cries.  (if an anti-goal is met and Billy is sufficiently
bratty and inflexible, then he cries).

>
> > The category you
> > are trying to define is called "individual-intention".  At first I
> > thought it was something like goal or focal goal or goal structure, but
> > now when you add "Billy is crying" my internal case-based reasoning system
> > is quite confused.  Can you try to define "individual intention"?
>
> I assume that both "goal" and "lack of goal" are obvious
> enough.  By anti-goal I mean a goal that increases the
> distance between where I am now and where I want to be.

Ah, so any given goal might have an evil twin, or even several
evil twins.  Whereas the given goal might be a recognizer for
a desirable state, an anti-goal is a recognizer for an undesirable
state....We might want to do this instead via slots on a
given goal object.

Whatever.  The best evidence of KR progress is new bits of ontology
which make the model make new believable inferences for scenario
cues or classes of scenario cues or which improve the believability of
existing inferences for said scenario cues.

>
> Maybe this is the part of the goal-satisfaction model
> which distinguishes between goal-blockage and goal-failure.
> For example:
>
> Tracy wants a banana.
> Mommy gives Tracy an apple.
>
> If Tracy's goal is ridged then she may construe Mommy's

I assume you meant to say Tracy's goal is rigid
(not "ridged" (-:).

> action as fulfilling an anti-goal.  She may start crying
> now in protest: "I don't want an apple, I want a banana!"

That would be a fun an interesting direction to take the
model.  Add personality attributes to it.  Then some situations
would be in the form of ...

{Tracy/Eric/Toby} wants a(n) {apple/banana/hot chocolate/ride in a car/
ride in a train}.  (He/she) is a {good kid/shy kid/bratty
kid/firestarter}.

...with a little creativity we can add some interesting
parameters to the overall model that do not overly tax
the KR abilities of our system yet do cause our scenarios
for people to rate to be more interesting....That is
why I am a little loathe to move to a different KR system
than KM at this point.  We have a working model that is
easily extendable.

>
> There is a correspondance (to some extent):
>
> unfulfilled    fulfilled
> -----------    --------------
> goal           happy
> anti-goal      sad
> no goal        indifferent


Sure.

>
> When you read the powerloom KIF file then you'll
> need to remember how to interpret the terminology:
>
> personal = goal
> suffer   = anti-goal
> purity   = no goal

Hmmm, okay.  Well, I hope there are good comments
in the KIF file.

>
> I will probably change the terminology after we figure
> out what makes the most sense.

Okay, but, err, umm, what is our overall goal here?
What new/better piece of model functionality are we
in the process of getting here?

>
> > Oh, I'd say lets start with accepting any format.  Keep a close eye on
> > 'em.  If we find that we've got 20 stories after 2 weeks we might have
> > to put some tighter constraints on things.
>
> Yes -- OK
>
> Things are coming together in my head.  :-)

Oh goodie.

Bill

>
>
>