On 4/6/2011 3:28 PM, Darren New wrote:
> On 4/6/2011 14:07, Kevin Wampler wrote:
>> posted to be pretty easy (only missed 3).
>
> I got 22, which was the low end of normal.
>
> I have to wonder how they got people to act suspicious or surprised or
> something for the photo. I'm not sure you can take an actor's portrayal
> of an emotion and then judge someone else on how well they can recognize
> the subconscious bits of it.
>

The situation under which using actors wouldn't be valid would be if 
there were people who had significantly more trouble reading eyes of 
actors but had no more trouble than normal for "true" facial expressions 
(or vice versa).  This seems possible, but I'd be quite surprised if it 
had a statistical effect at the same level as overall face reading 
ability, so I don't think that the test is invalidated by this concern 
so long as you rate people on a relative to each other, rather than on 
an absolute scale.

Also, here's how I think they got the photos.  They had actors come into 
the studio and perform a set of pre-chosen facial expressions.  Each 
expression was photographed and the eyes extracted.  These photos were 
then given to a panel for four judges, who decided upon a word 
describing the internal mental state of the actor.  Only those photos 
for which the panel reached unanimous agreement were used.  At least 
that's the best I've been able to figure from reading the research paper 
on which the initial form of the test was based.

Post a reply to this message

On 4/6/2011 2:19 PM, andrel wrote:
>>
>> That's actually pretty interesting in itself, because I found the test I
>> posted to be pretty easy (only missed 3). You might find that you'll do
>> better than you expect when you take it though, I know I didn't expect
>> to do that well.
>
> Ok, you talked me into it. I did indeed better than chance. I scored 15
> that is 6 more than pure chance.
>

Not so bad considering the nature of the test IMHO.  I find it pretty 
impressive that it's even at all possible to answer these questions at a 
better than pure-chance rate.

This test in relation to the uncanny valley also reminds me of when I 
spent a while trying to make some facial animations.  I spent so long 
staring as slightly-off animations deep in the uncanny valley that I 
stopped being able to see what exactly was wrong with them.  Basically 
I'd spent so long looking at these super-creepy looking animations that 
they started to look "natural" in some sense.  This made debugging even 
more of a painful process than it would have been otherwise.  Suffice to 
say, I haven't tried to to program anything else involving human faces 
since then.

Post a reply to this message

On 4/6/2011 16:16, Kevin Wampler wrote:
> I spent so long staring

I find this happens with any image, animation, or sound effect with me.

Or even, for that matter, saying the same word over and over until it stops 
making sense and you don't remember how to spell it any more.

-- 
Darren New, San Diego CA, USA (PST)
   "Coding without comments is like
    driving without turn signals."

Post a reply to this message

On 4/6/2011 16:09, Kevin Wampler wrote:
> were people who had significantly more trouble reading eyes of actors but

I don't know. If you had an actor give a fake smile and asked "what is this 
expression?" almost everyone would say "smile."  But if the smile doesn't 
reach the eyes, that's not going to work well.

> panel for four judges, who decided upon a word describing the internal
> mental state of the actor.

Huh.  I'm still not convinced. :-) Not until someone explains how you pick 
four judges with good judgement in such things.

-- 
Darren New, San Diego CA, USA (PST)
   "Coding without comments is like
    driving without turn signals."

Post a reply to this message

On 4/6/2011 5:04 PM, Darren New wrote:
> On 4/6/2011 16:16, Kevin Wampler wrote:
>> I spent so long staring
>
> I find this happens with any image, animation, or sound effect with me.
>
> Or even, for that matter, saying the same word over and over until it
> stops making sense and you don't remember how to spell it any more.
>

There's actually a term for that (at least the meaning part, I dunno 
about the spelling part): 
https://secure.wikimedia.org/wikipedia/en/wiki/Semantic_satiation

Post a reply to this message

On 4/6/2011 5:07 PM, Darren New wrote:
> On 4/6/2011 16:09, Kevin Wampler wrote:
>> were people who had significantly more trouble reading eyes of actors but
>
> I don't know. If you had an actor give a fake smile and asked "what is
> this expression?" almost everyone would say "smile." But if the smile
> doesn't reach the eyes, that's not going to work well.
>

Certainly, but in that case you'd expect that *everyone* would have 
trouble answering the question, so it's not going to impact where you 
fall with respect to the average score (which is what they measure in 
the study).  The effect you're talking about is really determined by the 
variance in the ability of different people to judge acted versus real 
eye expressions.  I'm sure there is a variance, but I'd imagine that 
it's smaller than the variance in overall ability in reading eye 
expressions.

Another point is that they *do* find that people perform much better 
than random chance at selecting the right expression from the eyes 
alone.  They also found some other statistically significant in the 
abilities of different subgroups which perhaps make intuitive sense: 
women performed slightly better than men, and autistic spectrum people 
performed worse than non-autistic people.  To me this points that they 
are clearly measuring *something* useful.  It's probably not exactly 
equivalent to what they'd really like to study (ie. real expressions) 
but it seems about as close as you can easily get in a controlled study 
-- certainly close enough to draw some cautious conclusions from.  I 
mean, you're studying humans here, you shouldn't expect a proof of 
correctness.

As an aside, they found a few other interesting things too.  For 
instance for "simple" expressions like happiness, sadness, anger, etc. 
the whole face conveyed significantly more information than the mouth or 
eyes alone, and the mouth and eyes both conveyed about equal amounts of 
information.  For "complex" expressions like sarcasm, pensiveness, 
refection, etc. on the other hand the eyes seemed to be the most useful 
thing in conveying the expressions.  Also interesting was that the 
autistic spectrum people only seemed to have significant trouble with 
the complex expressions, but not with the simple ones.  I actually found 
it to be a pretty interesting paper to skim -- not that I actually know 
anything about the field of course.

Post a reply to this message

On 4/6/2011 17:42, Kevin Wampler wrote:
> ability of different people to judge acted versus real eye expressions.

Exactly.

> but I'd imagine that it's smaller than the
> variance in overall ability in reading eye expressions.

I wouldn't assume that, personally. But I know almost nothing about such 
subjects, really.

> To me this points that they are clearly measuring
> *something* useful.

Certainly.

> "complex" expressions like sarcasm, pensiveness, refection, etc. on the
> other hand the eyes seemed to be the most useful thing in conveying the
> expressions.

I think it's likely that the eyes are used much more for "intentional" 
expressions, too. Things you want the other person to see, like skepticism 
or disapproval?

> Also interesting was that the autistic spectrum people only
> seemed to have significant trouble with the complex expressions, but not
> with the simple ones. I actually found it to be a pretty interesting paper
> to skim -- not that I actually know anything about the field of course.

That's cool. Do you still have a reference to the study sitting about?

-- 
Darren New, San Diego CA, USA (PST)
   "Coding without comments is like
    driving without turn signals."

Post a reply to this message

On 4/6/2011 6:42 PM, Darren New wrote:
> I wouldn't assume that, personally. But I know almost nothing about such
> subjects, really.

You and me both.  It does seem like something that would be interesting 
to study, although I can imagine that getting the photos of "real" 
expressions would be tricky.


> That's cool. Do you still have a reference to the study sitting about?

There's actually two papers, the initial study:

http://psych.colorado.edu/~tito/sp03/7536/Baron-Cohen_97.pdf

And the follow-up on which the actual test I linked to was based:

http://www.autismresearchcentre.com/docs/papers/2001_BCetal_adulteyes.pdf

Note: somehow the line break in the above URL actually matters, and if 
you remove it the link won't work.  If it still doesn't work, just do a 
Google search for the title:  The "Reading the Mind in the Eyes" Test 
Revised Version.

The primary author for both papers is apparently a relatively well known 
autism researcher: http://en.wikipedia.org/wiki/Simon_Baron-Cohen

Post a reply to this message

On 06/04/2011 06:18 PM, nemesis wrote:

>
http://files.sharenator.com/steven_segal_emotion_chart_Steven_Seagal_Emotion_Chart-s513x708-65813-580.jpg

LOLrus.

And so true to life...

Post a reply to this message

On 06/04/2011 06:22 PM, nemesis wrote:
> Invisible escreveu:
>>> Why wait for a great sunny day to shoot when your digital sun and blue
>>> sky look better with no sign of smog?
>>
>> I think the difficulty of waiting for a sunny day is utterly eclipsed
>> by the sheer intractability of rendering realistic smoke, flames,
>> dirt, surf...
>
> I'm seeing realistic enough smoke, flames and dirt in games these days,
> let alone in movies...

This stuff is improving all the time of course. They have now managed to 
get believable fur and basic water working in CGI. But I've yet to see 
anybody manage really high-quality smoke or flames yet.

Post a reply to this message