NetNews Usenet Archive 1992 #31/NN_1992_31.iso/spool/bit/listserv/csgl/2038

Text File | 1993-01-03 | 5.7 KB | 108 lines

Comments: Gated by [e-mail]
Path: sparky!uunet!zaphod.mps.ohio-state.edu!usc!howland.reston.ans.net!paladin.american.edu!auvm!AERO.ORG!MARKEN
X-Delivery-Notice:  SMTP MAIL FROM does not correspond to sender.
Posted-Date: Sun, 03 Jan 93 15:25:03 PST
Message-ID: <[e-mail]>
Newsgroups: bit.listserv.csg-l
Date:         Sun, 3 Jan 1993 15:25:03 PST
Sender:       "Control Systems Group Network (CSGnet)" <[e-mail]>
From:         [e-mail]
Subject:      3 Pencils & 4 Rubber Bands, Etc
Lines: 95

[From Rick Marken (930103.1000)]

Gary Cziko (930103.0240 GMT) --

>I did it!  I just designed the most awesome manual PCT demonstration of all
>time.

Nice work!! Looks like the end of psychology as we know it, right?
Wrong! Greg Williams (920103) points out:

>Skinnerians would have no problem with the fact that (even slightly)
>DIFFERENT "responses" resulted from different "stimuli." And if you show them
>results where successive "responses" are IDENTICAL, yet the "stimuli" in each
>case are different, they will talk about "stimulus generalization" or say that
>the organism can "lump" different-appearing stimuli (to the experimenter) into
>ONE kind of "discriminative stimulus." But it gets even worse. If successive
>"responses" are judged as different by the experimenter, they might say that
>they really are all in the same "operant" set of responses.

Greg makes an EXCELLENT point in this post; psychologists in general
(and Skinnerians in particular) are not going to be persuaded by these
demonstrations of principles because it is very easy to SAY it's just
"stimulus generalization" or "discriminative stimuli" or "operants" or
whatever. You can't "persuade" people with these demos unless they are
1) willing to be persuaded and 2) willing to deal with the problem
QUANTITATIVELY. Gary's demo is a good example. I developed the computer
version of this demo specifically to anticipate the kind of criticisms
I thought psychologists might have of Powers' demonstration of the failure
of the causal model in tracking tasks. In several of his early papers,
Bill showed that, in a compensatory tracking task, the correlation between
input (cursor), i, and output (handle movements), o, can be nearly zero
while the correlation between disturbance (which is invisible) and
output is on the order of .99. Bill's demo simply illstrates in
practice what the equations for a closed loop control system show
analytically -- that the output of a control system depends on the
disturbance to the input, not on the input itself. This is an amazing
finding (from the conventional perspective) -- because the cursor is
all that the subject sees -- it MUST be the cause of what the subject
does. Most models of tracking assume the o = f (i) -- the output is
some function, f, of the input. The function,f, characterizes the
way the subject transforms inputs into outputs; f is a functional
model of the subject, from the conventional perspective. Indeed,
all psychological research is based on the premise that you can discover
f (for a particular task) by varying i (the independent variable) and
measuring its effect on o (the dependent variable). Powers' little demo
showed that there is no visible functional relationship, f, between
i and o when behavior occurs in a closed loop. Obviously, this was a
finding that would not be easy for psychologists to swallow -- seeing as
how it would call into question the validity of virtually EVERY PSYCHOLOGICAL
LAW that had been discovered to date.

I assumed that psychologists would say that Powers found a low
correlation between i and o because 1) f was highly non-linear and
thus would not be captured by the correlation or 2) there was a lag
in the relationship between i and o so that o = f(o-t). In other
words, I anticipated QUANTITATIVE objections to Powers' demonstration.
So I tried to think of a demonstration that would obviate these objections.
My "repeated output with different input" demo does this. What I show
is that it is easy to produce virtually the same o on two occasions
while i is COMPLETELY different on each occasion. Now the person
claiming that there must be SOME function that produces o from i must
find a function that can map i1, i2, ... iN (all different temporal
variations in cursor position) with the SAME o. This is just not a
mathematical possibility (even allowing for the slight statistical
differences in o on each occasion).

So Gary's demo quantitatively rules out the possibility that o = f(i)
when behavior occurs in a closed loop, negative feedback situaiton.
It PROVES that sensory input is not the cause of behavioral outputs
-- no matter how ridiculously counterintuitive this seems. But will
this demo convince a psychologist who is busily doing research based
on the assumption that o = f(i). NO WAY, JOSE. S/he can always
describe the results VERBALLY  -- invoking the shiboliths of
scientific psychology -- "stimulus generalization", "response
generalization", etc -- and they can get back to work.

As Greg said -- demos like this are no problem for the scientific
psychology establishment. I confidently predict that if you (Gary,
or anyone else) tries this demo with a standard psychologist -- they
won't even break stride; they'll have a quick explanation, see no problem
and go off, comfortable in the knowledge that there is no problem
at all. I don't think any demo -- no matter how clever -- will ever
wake the psychological establishment from its dogmatic slumbers. Only
those who are willing to learn -- AND WILLING TO THINK QUANTITATIVELY--
have any hope -- and I think all of them are already in CSG.

I say this, Gary, so that you won't be too frustrated when you find
that your brilliant demo produces virtually NO revelations amongst
your colleagues.

Very nice work though.

Best

Rick