[Heartlogic-dev] Re: putting it all together

[Joshua N Pritikin]

On Tue, Mar 09, 2004 at 11:38:54PM -0600, William L. Jarrold wrote:
> On Tue, 9 Mar 2004, Joshua N Pritikin wrote:
> > The hypothesis about how to combine appraisals has lots of independent
> > parts --- each combination of #$GoalStatus's can be considered
> > separately. The example I gave, #$Protest, is one of the least
> > intuitive parts of the hypothesis.
> 
> Yes, I agree with the last sentence with a certainty of about
> .8.

OK  :-)

> I hope you don't mind my using certainty factors.  Most people get
> irritated or look bemused or quizzical.  Why?  What is wrong with
> giving a numeric value to the intuitive certainty of our statements?

Seems reasonable to me.  No objection here.  Perhaps the folks who
react with irritation are suffering from neurotypical syndrome.

> > Why did I start with the least intuitive part?  Because I just wanted
> > to get you onboard with the idea of combining appraisals.  It sounds
> > like you are onboard (more or less).  Cool.
> 
> I am an enthusiastic supporter of testing the combinability of appraisals.
> We need to validate the notion of compound appraisals.  This is one issue
> that my dissertation explicitly leaves as future work.  Am I wrong? If so
> thank you for identifying a gap in the Conclusion chapter.  If
> so, Study 1 gets this point across, at least sorta kinda.

I don't remember.  I'd have to re-read your conclusions.  I vaguely
remember that you _do_ mention it.  Maybe.

> > Looking at your dissertation (and Kathryn Sanders 1989), some words
> > are taken as primitive such as give, receive, sell, and steal.  These
> > words can be defined by the combining hypothesis.
> 
> Huh?  give, receive etc are actions.  Not affective reactions.  Right?

These words _are_ affective actions.  What is ownership?  Recall the
central tenet of cognitive appraisal theory: that situations do not
have significance in of themselves, but only by virtue of an
individual's interpretation.

For example, consider land ownership.  What does it mean?  It means
that I have the cognition: "I own this land."  I am basically
unchanged.  The land remains unchanged.  It is only because people are
mad to own land that we write down our cognitions on pieces of paper
and file them with a government registrar.

> > Perhaps you feel
> > like playing Mr. Devil's Advocate, so you don't have to believe my
> > claim.  I'm only asking for your cooperation on setting up the
> > hypothesis.
> 
> My biggest concern now is tractability.  In my dissertation I have
> started simple, very simple on purpose.  The next step should
> probably be a baby step too.  I think a study of goal substitution
> study would be the next step.  Instead of just banana vs apple
> lets look at bannana (target) vs one of (apple, coal, skittles,
> cough syrup, kiwi fruit, asparagus, world peace)....But maybe not.
> 
> A good place to put a lot of energy is to write rules lotsa rules.
> Get lotsa generativity out of a model.  Get a model that is fun
> to play with just like Josh White says.
>
> I'm kinda brainstorming other potential directions to make sure you
> really have thought about these and still persist in the direction of
> #$Protest and friends.

Yah, I definitely want to help integrate our trivial model with large
commonsense ontologies.  However, I will also stubbornly pursue my
combining hypothesis.  If you can believe it, this combining
hypothesis dates back before the beginning of the Aleader project.
Only now I feel like I am successful in explaining it to you.

> > So yah, "What are people going to rate?"  Good question,
> 
> Damn right it is a good question.  It is an *essential* question.
> 
> Think of your items.  Also, think of your data tables.  How will they look
> like in a final article.  Also, spell out your testable (i.e.
> operationalizable hypotheses...e.g. see my H1, H2, etc things in
> my diss).

The hard part is just setting up the experiment.  Take a look at my
mock-up:

  http://openheartlogic.org/opine.cgi?leaf=mockup

> > First, we get ratings of each appraisal from every point of view (for
> > each #$appraising-agent for each #$mindreader).  Then we take
> > believable pairs of appraisals, combine them and rate believability of
> > the hypothesis output.
> 
> How do we combine them?  Simply add them?  In what order?

See the mock-up for examples.

> > Add in experimental manipulation
> 
> What are the independent variables:
> 
> o Valence reversal

Yah, basically.  There are nine categories of appraisal pairs which we
can map to affective actions.  The equivalent of valence reversal
would be to just map them the "wrong" way.

> o whether the appraisal is compound vs atomic, and
> if it is atomic then
> is it #$appraisingAgent 1 vs 2
> or
> is it #$mindreader 1 vs 2

Uh, no, not like that.  The proposal only deals with pairs of
appraisals.

> > Am I going too fast?
> 
> Nope.  I might rather see how happy and sad combine.  When do they
> neutralize?  When does one trump the other.  I'd rather do this,
> before testing for which atoms combine and how such atoms combine
> to form #$Protests and its ontological kin.

OK, but I argue that combining happy and sad is more sophisticated
than what I am proposing.  There is no conflicting affect in my
combinations.  I am combining perspectives, not affects.

-- 
A new cognitive theory of emotion, http://openheartlogic.org

signature.asc
Description: Digital signature