[Aleader-dev] Re: direction

[Joshua N Pritikin]

On Sat, Aug 09, 2003 at 03:36:34PM -0500, William L. Jarrold wrote:
> On Thu, 7 Aug 2003, Joshua N Pritikin wrote:
> > One more note: For easy of discussion I assume "emotion" =
> > "affective state".]
> 
> That is fine for email and agreements between you and I, but prolly
> (prolly = my silly shorthand for "probably".  Some American's with
> regional accents might pronounce "probably" as "prolly")

I'm living in India, but I moved here only ~3 years ago and all
my education happened in USA.  In short, I'm familiar with most
email slang.  ;-)

I'll ask if I can't decode something.

[Here is some dialog which is my response to an earlier email
from you.]

> > In any classification scheme, some variation is unavoidable.
> > The challenge is to see how well we can minimize the variation. 
>
> Yes, you want to minimize variation within a category.  However the
> more categories you have, the less manageable and or harder to verify your
> theory becomes.  Yes, in the limit as time approaches infinity you
> will want to maximize the number of your categories.

Actually it's not that bad.  Roseman96's model doesn't include Aleader's
appraisal of "phase".  If we multiply the 17 Roseman emotions by
3 phases (before, during and after) then we get 51 emotions.  So I
argue that Aleader's model isn't much larger than Roseman96.

> I doubt that "real" NLP takes up much more memory than Good Will
> Hunting.  How much memory does the average feature length film
> require?  Any idea off hand how much Cyc and or KM require?

I took a look at LIWC.  It looks like a superficial approach to NLP.
I guess LIWC would easier than integrating with Erik Mueller's
ThoughtTreasure or KM.  On the other hand, I tend to prefer the deeper
approach, even if it takes more effort.

OK, maybe it's not impossible, even with my modest computing resources.
It sounds like a lot of work though.  I feel like I have already done
too much work without any pay-off.  I am eager to grab some low-hang
fruit.  After a few bananas, I'm willing to re-consider all the
options.

> You might also be interested in the CASP and the DANVA.
> Both of these are instruments that neuropsychologist type people
> who I work with at UT use.  And these instruments are used to
> detect impairments with social cue perception.

Yah, I can definitely see applications for Aleader here.  If we
published the right type of article and tweak the software a bit
then it should be an easy sell.  I see that DANVA costs $250.
If we start some kind of mail-order business then I bet we can
get at least one order per month.  ;-)

> Inferring emotion from prosody is something much different than
> the logical cognitive models that my dissertation was about.  But
> your system is capable of assessing someone's prosodic perception.
> But your formal system is miles away from modeling prosodic perception.
> The kind of formal system for that is more like a markov process,
> a neural network...something that handles non-digital, non-symbolic
> representations.

Well .. my main interest is in empirically validating the Aleader
aspraisal model.  Assessing prosodic perception does not really
interest me.  On the other hand, all my source code is GPL'd so
anyone is welcome to work on it.

In an eariler email, you mentioned Clark Elliott.  I guess I find his
research a little bit strange.  According to my philosophical beliefs,
what is important is to teach people how to recognize the
meaning of emotion.  Teaching computers how to express emotion
doesn't seem very interesting.  I think John McCarthy would agree
with me.

A question about attribution:  I took some of your email and dropped
it into the Aleader manual, almost verbatim: "A given personality in a
given film clip may be depicting several different emotions at once."
Do you care about this?  I don't mind adding your name somewhere.  Let
me know if I should do something.

[And now back to today's back & forth.]

> not good for getting published.  Much ink has been spilled on what
> is an emotion.  We need to have a good definition.

No doubt.  If we need to discuss the definition of "emotion" then
let's discuss it.

> Part of making a good definition is to say what our definition is NOT.

Agreed.

> The following
> blurb shows how Ortony et al have defined emotion.  This is an excerpt
> from my master's thesis (well, technically it is from my prospectus) that
> was commented out from the final version...
> 
> %The first step in this process is defining emotions as affectively
> %valenced construals of situations.  However, they define emotion as
> %distinct subset of affect. (actually this is not from [OCC] but rather
> %from [Ortony, Clore and Foss 1987a]) The best examples of emotions are
> %those that a) refer to internal, mental conditions as opposed to
> %external or physical ones, b) are clear cases of states (I'm not exactly
> %sure what they mean by this), c) and have affect as opposed to behavior
> %or cognition as predominant focus.  As these constraints are relaxed,
> %one gets non-emotions or poorer examples of emotions.  Though the

This is probably wrong, but I am tempted to write "a summary of
the Aleader appraisal instructions."  It is tautological that
we are including all the emotions which can arise from any
situation which can provide answers to our appraisal questions.
Erm ..

One point which may be helpful is that Aleader emotions
happen "outside" in the sense that the emotion is a relation
between two people.  I speculate that Damasio has targetted
his emotion research from the opposite angle, trying to
measure the _physiology_ of emotion (& maybe thinking).
I'm not very familiar with Damasio's work, so I'm just
guessing here.

> ...As I recall OCC has a nice blurb on how "abandoned" does not refer to
> an emotion.  Our paper should say that although "abandoned" does not refer
> to an Ortony like emotion but it does refer to an affective state.

Yah, that's a nice tie-in.  I have taken note.

> > > I don't see haughty as being a sub-case of admires, unless admires
> > > includes self-admiration?
> >
> > To be precise, the appraisal category is:
> >
> >   "I am _expecting_ you to admire me."
> >   (I=Will, you=Chuckie, spoken in a pushy tone of voice)
> 
> OCC states that admiration, by definition, involves one person to feel
> good about something that another has/does.  The emoter is focusing on
> the praiseworthiness of another's role in some situation.  The emoter
> finds that the admired person is upholding some kind of standard.  I find
> this to be a good coherent definition of the emotion (and the affective
> state) referred to by the word "admiration".

Yes, that is exactly how Aleader defines the general category of
admiration.  However, Aleader leaves the point-of-view unspecified.
It sounds like OCC style admiration is always from the admirer's
point-of-view.  To contrast, Aleader considers the point-of-view
(the admirer or the admired) as a refinement of admiration.

> As you state, the "You think I'm afraid of you big fuck" clip is about
> Will expecting Chuckie to admire him.  I agree with that statement well
> enough for current purposes.  However, I will add that the affective
> state depicted is *related to* admiration but it does not fit the OCC
> definition of admiration.  Will is *expecting* Chuckie to admire him.
> Will is *not* focusing on the praiseworthiness of Chuckie's comportment
> in this situation.

I think you meant: "_Chuckie_ is *not* focusing on the praiseworthiness
of _Will's_ comportment in the situation." (Will and Chuckie are
reversed?)

> Expectations fit in via standards based emotions such
> as admiration and reproach.  E.g. people are expected to hold up the
> standard of "exhibit good manners" or "don't lie" or "don't kill people."
> Back to the example at hand, Will is expecting Chuckie to admire him.
> Will is sitting there like, "come on you fuck, admire me!" and Chuckie is
> not complying with this expectation.  In fact, I think Will is feeling the
> opposite of admiration.  He is feeling reproach in this situation.

Are you saying that Will's

1. comment is a reaction to his prior feeling of reproach
2. comment is an expression of reproach
3. he feel reproach after making his comment

?  (Roughly, I'm asking you to clarify "phase" in Aleader terms.)

> Will
> expects his friends to uphold the standard of admiring those who deserve
> to be admired.  And Will believes that he is deserving of admiration.
> Chuckie is "just sitting there".  So, Will perceives Chuckie's role
> as violating the standard/expectation of "Admire those who deserve to be."
> According to OCC, if someone focuses thought on a another's
> blameworthiness with respect to a given standard/expectation and these
> thoughts are "above threshold" (kind of a handwave for present purposes)
> THEN reproach is felt.
> 
> Hmm, let me do a "parity check" here to make sure we are on sort of in
> sync....  

Yah, sounds reasonable.  I bet you'll pick [1] from the list above,
but maybe not ...

> The original line from the film is "You think I'm afraid of you
> big fuck." [a] Does "you" refer to Chuckie or Will?  Joshua, you
> basically stated that subtext of [a] was "I (Will) am _expecting_
> you(Chuckie) to admire me."  Correct?

Correct.

> Hmm, I used the word "parity check".  Not sure if you have had the right
> courses/self study to know what I am referring.

Yah, I know a little bit about digital circuit design.  While in
junior college, I built a simple counter with wires, chips and
solder.  It was fun.

> What kinds of classes did you take at CMU?

I finished two years of a four year software engineering &
math program.

I left because the courses increasingly seemed irrelevant to
what I was really interested in.  Somehow I am not very good
at doing things which don't interest me.

> Any significant self-study beyond that?

Yah, tons.  That's one of the reasons I got bored at college.
I was forced to take courses like "How to program in C" while I
was working with a team to write a multi-user Star Trek game in
my spare time.  ;-)

Maybe I should have applied for a PhD program or something,
but it never entered my mind.

> Now that I think of it, what computer languages do you know?

My favorites are C and Perl.  I also know Borne Shell, C++ (ugh),
Bison, Lex, PCCTS, Lisp/Scheme, Prolog, Pascal, SML, Haskell,
and a bunch of others which I temporarily forgotten.

The question you should ask is: what computer languages do
you _not_ know?  Well .. KM, except that I almost know it
now because I read the manuals.

> I'm not sure what you mean by "funny cases".  I don't think there
> are another other elements to be added to the above 3 element list.

Good then.  Maybe the list won't expand.

> > Which reminds me ... I have tried to explain this next point
> > a few times.  Tell me if you understand (or if you already
> > understood previously):
> 
> Thanks for asking.  Questions like this seem to be causing me to feel
> hopeful optimism about our working together.  (-:

In an attempt to increase your comfort level further, I would like
to point out that I have some experience doing team projects:

  http://search.cpan.org/author/JPRIT/

The "Event" module is the most popular.  While I did most of the
coding, grep'ing the ChangeLog shows about 25 contributors to the
project.

> So, to answer your question about your next point.  I sort of understood
> it but not entirely.  And I understand it slightly better having read
> the paragraph below.  However, I still have some lack of understanding.
> See my questions below.

Let's keep going until it's crystal clear.

> This helps.  If you haven't already, put the above paragraph in your
> writings near the beginning.

Done.

> However, you will need more.  There are three big open issues
> here.  There is:
> 
> Issue 1: "What rules govern cognitive appraisal?"  (I.e. for a given agent
> how can we predict the affective states experienced for a given set of
> perceptions).
>
> Issue 2: "What rules govern affective mindreading a.k.a. cognitive
> empathy?"
>
> Issue 3: "Are there answers to issues 1 and 2 that can be consciously or
> deliberately used by people for the purposes of personal growth."

I'm not sure whether I understand the issue questions.  I'll
give it a try though:

Issue 1: I've tried to spell-out the rules in the Aleader manual
in the Classification chapter.  Doing a KR model of the appraisal
process would probably help to make the written instruction more
precise.

Issue 2: For Aleader, I think we need to flip this around: "Do
the rules which govern cognitive empathy apply to our own personal
cognitive appraisal process?"  I have been assuming all along
that the answer is simply "yes".  Am I naive?  I'm not sure how
to discuss this issue, whose other opinion, compare, contrast??

Issue 3: I don't want to tackle this until I get some feedback
about whether I understand issues 1 & 2.

> Please mention these issues your (our?) writings.  You (me too!)
> hypothesize that pursuing issue 3 is worth trying.  Nonetheless, it could
> turn out that the answers to Issues 1 and 2 are so complex, so
> non-intuitive that there's not much personal growth that can be obtained.
> 
> Please mention something like the above (copy it in is fine with me for
> present purposes) to the relevant part of your (our) writings.

OK .. it probably goes in the introduction somewhere.

> > > My intuition tells me that you are into both the researchy and the
> > > personal growth aspects of this CD.  However, my intuition also tells
> > > me that you have not yet developed separate strategies for these two
> > > aspects.
> >
> > Correct.
> 
> Great.  We have at least some resonance then.  Bill's
> hopefullness/optimism potential just went up another notch.

Cool.

I just want to emphasize again that my ears are wide open.
Suggest how to make the project more scientific-ish and
I'll probably go for it.  If I forget, remind me.

(The only thing I don't want to do is literally _destroy_ the
writings I've done on philosophy and religion.  I don't expect
you to suggest that though.)

> > My main worry with putting the whole thing on a web site is that
> > I'll attract a cease & desist court-order for broadcasting
> > copywrited material (the films).
> 
> Well, you don't *have* to use films that have big expensive lawyers
> behind them, right?  I mean you can shoot your own.  You can find films
> that actually *want* more distribution.  There are plenty of well made
> films that are very emotionally evocative.  Student films are a good
> example.

I selected the films quite a while ago.  At the time, I was
concerned about fine-tuning the appraisal model.  I selected a
film for two reasons:

+ I felt like the emotional portrayals reflected reality.

+ I liked the film enough that was willing to watch it hundreds
or thousands of times.

Unfortunately I didn't seriously consider re-distribution rights.

20/20 hindsight ...

I am ready to analyze re-distributable films.  Can you recommend
any?  My downlink is 8kbytes per sec so I'm not going to be much
help in solving this problem.  The films should be at least one
hour long.  We'll analyze only the first 20 minutes.

Once we get new films, it will take time to go through
the analysis.  I estimate that 20 minutes of film takes me
about a week to analyze.  I wonder if we can put together
an initial article using material only on the existing CD?

I guess it depends how many human subjects we can corral from
our respective colleges.

> > (Can you imagine my state of mind before I figured out how to make
> > a bootable CD?  I had this cool project, but it was impossible to
> > show to anyone!)
> 
> Are you saying that distribution via CD is legal but on the web is
> illegal?  (If so I would feel at least slightly surprised if that were
> true).

I have been acting according to the belief (correct or not) that
the chance I'll draw a lawsuit is proportional to how much noise
I make about giving away films.  At this point, I'm just seeking
serious collaborators.  If we use a web site to collect test
results for a research study then a lot more people are going to
notice what we're doing.

> A nice thing about a conference is that the bar for participation can be
> lower.  You can sometimes get by with just a proposal -- and get good
> feedback on how to improve it.  Or you can get by with more preliminary
> data from a recently executed proposal.

Sure, but I don't want to do a lot of expensive air-travel unless
it is _really_ going to be worth it.  Round-trip USA-India tickets
are about $1200 per seat (minimum) plus VISA hassels.  Europe is
somewhat cheaper, but not much.

Contrast that with my extended family's yearly cost-of-living
of about $6000.  Do the math.  ;-)

> > The other point is that, as an academic outsider, I suspect I
> > will learn a lot from the publication process.  I bet my whole
> > presentation will improve tremendously just by immersing myself
> > in academic procedures and protocol.
> 
> Yes.  Btw, my advisor said that reviewers anonomously review stuff.  Thus,
> to the extent that anonymity really can be preserved, it should not matter
> that you are not part of the academic regime.

Yah, a research article seems like the way to go.

> > Hrm, will a "human subjects review" also apply to any
> > Indian subjects?
> 
> Ask the professor with whom you meet.
> 
> By the way, these issues fall under "scientific research ethics." So, if
> you intelligently use that phrase in your language (e.g. with your
> professor) you will seem more like an insider, less like some whacko that
> the prof needs to be worried about.

OK, I put it on my list.

> > + Read my comments on Roseman96.  I emailed them yesterday.  If you
> > didn't receive them then let me know.  I'll re-send.
> 
> Eek.  I have not read this yet.  I believe I saw it in my mailbox.

Here's another reminder.  ;-)

> I'm happy enough with our
> current modus operandi.  The key is to focus on deliverable writings --
> proposals, drafts.  Many important nuggets are in the above email.

Agreed.

-- 
.. Sensual .. Perceptual .. Cognitive .. Emotional .. Oh My!