Turing, Giving Up (sort of)

[From Rick Marken (920918.2100)]

Well, this is my last post from home. I'm giving my kid my
Mac; he's going off to college tommorrow. I was going to
get a replacement before he left -- but the anticipated
"trickle down" failed to materialize. But this should help me
buckle down to some real work on the weekends.

But here are some quick replies to swat around until monday.

Bill Powers (920918.1600) says, to me:

I think there's one aspect of the Turing test you may be overlooking.
It's not just an examination of output spontaneously created by a hole
in the wall behind which either a machine or a person might be
lurking. It's an interactive test; that is, you send messages through
the hole and get messages back.

I know that. I should have been clearer, I guess. My point was that
the Turing test is NOT aimed at the discovery of controlled variables;
but that's all that's wrong with it. And that is really all that is
wrong with conventional psychological research also. But that is a BIG
"all". Skinner tested the effect of "schedules" on behavior until
his ears turned blue; he just failed to consider the possibility
that the organism was controlling some aspect of the reinforcement;
so he forgot a few, little aspects of "the test" -- mainly, he forgot
to hypothesize what variable might be controlled. He did introduce
disturbances ("schedule changes") but he didn't monitor the
hypothetical controlled variable under disturbance and change
hypotheses if it turns out that the variable is not protected from the
effects of the disturabnce. The differences between the "behavioral"
approach and the PCT approach to behavioral research are
fairly small -- but crucial. I think the Turing test provides a
good framework for pointing out these crucial differences (as you
did in your post).

I also believe that idea of "intelligent behavior" is silly -- but
I think that's the way may ai types talk about it.

Gary Cziko (920918.1925) says:

Rick Marken (920918.1330) attempting to cop out says:

Not just attempting -- succeeding.

Have you stopped to think that any reviewer who is not familiar with PCT is
being called (at least partly) "blind" in your paper? Do you expect him or
her to respond positively to that?

Good point. But the one very brief review was actually fairly positive;
it said some nice things. There was no evidence that the reviewer thought
that he/she was being referred to as "blind" -- actually, I think that
the paper might have had a better chance if this were noticed. The
reviewer basically said "so what" -- not "what do you mean 'I can't
see the true nature of behavior'". The reviewer was clearly blissfully
blind to my imputation that he/she was blind.

I know that there are journals out there that specialize in non-mainstream
and more radical approaches to psychology.

Why try? Hardly anyone pays attention to my published stuff anyway.
The CSGNet forum is plenty for me. See my reply to Estes below.

If this were my article, I wouldn't give up yet.

You haven't done this as much as I have. I've published enough
articles for the time being. If I do another unbelievable experiment
maybe I'll try again. Right now, it just doesn't seem worth it.
What's the point of publishing anyway? For me, the goal is to get
some colleagues to help with the research and modelling; to show what's
been done and hope some people will help carry the ball forward. Extra
heads really help. But I don't get extra heads from my published stuff;;
the conventional psych audience just ignores it because it is simply
orthogonal to their goals. So, forget em.

If you don't feel up to this yourself, I propose that you draw on the
expertise represented on CSGnet. I'd certainly be willing to help where I
could. I think that this article could become a classic if you get it
published, although perhaps not in your lifetime (think how proud your
grandchildren will be!).

If you want to work together to try to make it publishable I'd be
honored to have you as a joint author. I think all of your ideas about
how to express the ideas in the paper are great. I'm just pretty
much through with trying to impress people who don't even know what
the show is about.

P.S. Rick, could you share the review(s) with us?

The review was VERY short so I think you can glean its contents
from this letter I am sending to Estes (the editor). I wrote it
as my swan song to publishing in conventional venues -- but if
you want to enter the fray as Marken and Cziko or vice versa, that
would be great.



18 September 1992

William K. Estes
Department of Psychology
Harvard University
33 Kirkland Street
Cambridge, MA 02138

Dear Dr. Estes:

  I give. Apparently conventional psychology does not want
a dialogue with control theory psychology so I'll retreat back into
my lunatic fringe. It is clear from the reviewer's comments on my
paper 'The blind men and the elephant: Three perspectives on the
phenomenon of control' (MS 92-61) that psychologists have no
fundamental problem with the basic assumptions of their discipline.
I can see that there is no use pointing out that 'the emperor has no
clothes' when everyone knows this and seems to accept it as
evidence of imperial splendor. I will just have to continue this
conversation with the few behavioral scientists who see that there is
a conversation to have.

  I will note that the reviewer made a few factual mistakes
about the paper. First, Powers' model does not say that all behavior
is best viewed as an attempt to "maintain sensory stimulation
constant". The model says that it is intentional behavior that is best
viewed as an attempt to maintain sensory input at constant or
varying reference levels that are specified by the organism itself;
other (non-intentional) behavior is best viewed as the result of
ordinary physical laws. The reviewer suggests that the control
theory view of behavior is 'obvious'. But if the input control view
is obvious then why is virtually all psychological research based on
the assumption that this obvious view is false. If organisms are
input control systems then the independent-dependent variable
approach to research reveals almost nothing about the organism but
plenty about the environment in which it lives (as shown by
equation 5 in the paper). This is the 'behavioral illusion'. If the
input control view of behavior were obvious this illusion would be
well understood and experimental psychology would already be
based on the methodology implied by control theory -- the test for
controlled input variables. This is the only methodology that can lead
to a correct understanding of the nature of the behavior of closed
loop control systems.

  The reviewer asks how control theory can make a real
difference. I explained how, in the very last paragraph of the paper,
viz. by changing the goal and methodology of psychological
research. Psychologists should look for controlled inputs, not
caused outputs. I forgot to mention that by doing this they can start
shooting for relationships between variables on the order of r = .99
instead of r = .56 or so. The reviewer asks if control theory can do
better at explaining puzzling phenomena. From my perspective,
nearly all of the purported phenomena of psychology are puzzling
because they cannot be accounted for precisely (less than 2% error,
say) and reliably by the behavior of working models. Control
theory can account for several phenomena with this level of
precision;. But, with only a small number of people doing this
work, it's tough to make progress quickly. Nevertheless, I think
it's better to have a high quality account of a few phenomena than a
low quality account of many phenomena.

  Trying to publish articles on control theory is like pulling
teeth -- only worse; in the case of control theory the teeth are
actively pulling back (because, of course, they are control
systems). And when these articles are finally published nobody
pays attention to them anyway. Worst of all, nobody seems to try
to publish experimental results showing that we (control theorists)
are wrong. Instead, we get approving nods (the reviewer noted
that "it is undoubtedly helpful to remind oneself from time to time
that the organism is usually part of a feedback loop"), misinformed
proclamations ("in sensitization, for example, the feedback can be
positive") and, finally, dismissal ("it would all be much better if the
author could show how his ideas make a real difference"). If
demonstrating the fallacy of a basic assumption of behavioral
research (the assumption that behavior is a dependent variable) does
not make a "real difference" then I surely don't know what does.
Clearly, we are presenting solutions to problems that do not exist
for conventional psychologists. C'est la vie.


Richard Marken


Richard S. Marken USMail: 10459 Holman Ave
The Aerospace Corporation Los Angeles, CA 90024
E-mail: marken@aero.org
(310) 336-6214 (day)
(310) 474-0313 (evening)