[From Rick Marken (930408.2100)]
Bill Cunningham (920408.1320) --
Cliff Joslyn is right. It is too early for a poll.
This debate has been going on at least since mid March. There
have probably been over 100 posts on it. By last week we got to
the point where we were submitting numerical output of quantitative
models that the ITers agreed would test whether there was infor-
mation about the disturbance in controlled perception. The results
demonstrate unequivocally that there is no information about the
disturbance in perception. The only thing that the ITers can hope
for now is that the whole thing will start over and it will come
out differently. It's over. The poll is interesting because it
will show how people interpreted things. But I can't see much use
in going on with this debate anymore. If you think that there is
information about the disturbance in perception (whatever that
sentence might mean to you know) after reading this debate, then
that's what you believe and there is absolutely nothing I or anyone
else on the PCT side could do that could convince you otherwise. Of
course, the same goes for me. I will continue to go on believing
-1
that o = g (d) in a control system (not o = f(p)) and there is
apparently nothing the IT people can do to get me to believe otherwise.
I think I have the math, the models and the numbers on my side. But it
appears that the IT folks have got god (or at least Claude Shannon)
on theirs.
But it's not over until it's over, so here we go into extra innings:
Allan Randall (930408.1700 EDT) --
The following is what I
am assuming as given in any attempt by me to reconstruct D from P:
- the percept p
- the reference r
- the output function O()
- a programming language, such as C
The following are *not* assumed (to do so would be "cheating"):
- the output o
- the environmental feedback function F()
- the disturbance d
My question to you is this: given the above as the basis for the
entropy calculation, do you still assert that there is no
information about the disturbance in the percept? If so, we have
a real disagreement.
Gary Cziko -- HELP!!
What does it take? The first set of conditions above are what
Allan assumes are given "in any attempt to reconstruct D
from P". I met those conditions last week and sent two vectors (P)
of p values, the value of r associated with each, the output, O(),
function and encouraged Allan and Martin to reconstruct D from these
P values. When no reconstruction was forthcoming (as noticed by
Gary Cziko, custodian of the Ds) Bill P and I posted the results of
simulation runs showing why it would be impossible to reconstruct
D from P, even given information about r, O() and the computer
language of one's choice (including C, C+ or C++). To reconstruct D
you had to know F(), one of the pieces of information that Allan rightly
calls cheating since there is NO WAY that a real control system can
perceive F(). In the quoted section above it sounds VERY MUCH
like Allan is saying that, if you CANNOT reconstruct D from P given
the assumed conditions (r, O() and C), then one could "assert that
there is no information about the disturbance in the percept".
Since we have shown that, indeed, you cannot reconstruct the disturbance
from the percept under these conditions then I think it is fair to assert
that THERE IS NO INFORMATION ABOUT THE DISTURBANCE IN THE PERCEPT. A better
way to say this (so that onlookers can get a better sense of what this
means) is that THERE IS NO INFORMATION IN PERCEPTION THAT CAN GUIDE
THE OUTPUTS THAT COUNTER DISTURBANCES TO THAT PERCEPTION.
Actually, there is one better way to say this: BEHAVIOR IS THE
CONTROL OF PERCEPTION. Hmmmm. Where have I heard THAT?
When I say "reconstruct," I mean
actual reconstruction of the sequence of values, I believe in the
same sense Rick and you have been using the word.
This looks like the nail in the IT coffin. Allan means the same thing
by "reconstruct" as we do. And we have shown that you can't reconstruct
D from P: as Martin once said "QED". But somehow I think the idea that
there is "information in perception from which the system can reconstruct
outputs that control perception" will rise again like Dracula. Say, Tom
Bourbon, if you're out there, could you hand me a stake?
ยทยทยท
-----
More from Bill Cunningham
Hadn't intended to enter fray, but consider following in response to
Cliff's question #6, whether IT is relevant to PCT:
It had damned well better be.
So I take it that it would be a disappointment if it turned out that
there were no information about disturbances in perception?
Man is considered a social animal,
requiring considerable interaction (read transfer of information)
between members of the species.
How about putting the following in parentheses instead:
"(read: having mutual effects on controlled variables)"?
I think it would be worth it for anyone who thinks PCT is only about
individual behavior to read Tom Bourbon's article in American Behavioral
Scientist, v. 34, no. 1, 1990, pp 95-105. Then you will see how nicely
and precisely "pure" PCT deals with interpersonal behavior -- without
a trace of IT. You might also like to take a look at Bill Powers'
"Gatherings" program if you are interested in social behavior involving
more than two social animals.
Greg Williams (930408) --
The fact that someone is using statistics doesn't necessarily make it
"pretty useless as real science goes,"
Well, I think you're just the guy to pick out the research article to
serve as a starting point for a discussion of "how to do PCT research".
You seem to know what the important stuff is that is being discovered
by conventional methodology. I don't want to be accused of picking a
straw man piece of research. All I want is an example piece of research
with lots of data that seems real important -- to you. My goal is not
to criticize the research, by the way. Just to discuss the PCT perspective
on what it means and what PCT methodology might contribute to understanding
the kind of phenomenon delt with in the article. OK?
And by the way, I DO think statistical methods are excellent ways of
studying population phenomena. I don't, for example, expect PCT
modelling to replace exit polls as a way of predicting election
results.
"WHAT'S IT TO ME?" would have been the category for me.
Bill and I (at least) have tried to explain what we perceive
as being at stake in this debate. The main thing for me is
methodology; the idea that there is information in perception
suggests, to me, that one can learn something about behavior by
looking at how information in inputs (stimuli) relates to human
outputs (responses). This is just the conventional approach to
studying behavior and it can be quite misleading -- because
the nature of the relationship between inputs and outputs does
not depend on characteristics of the organism; it depends mainly
on the relationship of those inputs and the organsisms outputs to
a common controlled variable.
But, then, I don't really know whether that's anything to you or
not. Since we've explained "what's at stake" many times and you
still ask "what's it to me" I would guess "nothing".
Best
Rick