Afterthoughts on PCT versus other theories

[From Bill Powers (940607.0200 MDT)]

Trying to get adjusted for GMT, so far with the opposite results. I
am some where in Australia at the moment.

Jeff Vancouver (940606) --

Afterthoughts from re-reading your post. You say

However, my primary mission (reference signal) is promoting
interdisciplinary collaboration. To that end, I have come to
adopt the philosophy of von Bertalanffy. That is, seeking a
higher-order paradigm that will allow scientists working in
different areas to collaborate and learn from one another.

It's not that easy to put into words what your "primary" reference
signal is. All you have to do is ask yourself, "Suppose I did
succeed in promoting interdisciplinary collaboration. What would
that get me?" This will show you that achieving interdisciplinary
collaboration is only a means to satisfying an even higher-level
reference condition -- and not just one single condition.

Actually, interdisciplinary collaboration is a lot easier to achieve
than collaboration within a discipline like psychology. The
stickiest situations arise when PCT comes up against people offering
different theories about the _very same phenomenon_. Bandura and
Locke reject control theory while describing the very phenomena that
control theory is designed to explain. Even without PCT, you have
cognitive scientists, Skinnerian behaviorists (there are plenty of
them alive and kicking -- er -- responding) and personality
theorists (like Lord and Hanges, Hyland, Bandura, and Locke) all
operating completely separately and completely at odds with each
other. Just consider the recent split of the whole field of
psychology right down the middle, clinicians against self-proclaimed

I find the usage of the term "theory" in various branches of the
behavioral sciences rather odd. When Bandura says he has a "theory"
that there are "proactive" behaviors, he seems to think that simply
announcing this phenomenon amounts to offering a theory. But to me,
it is only a description of something that we either can or can't
observe. If the phenomenon is replicable, we have to accept it as
real -- but that leaves the job of theory, as I have always
understood the term, undone.

PCT is not simply a description of purposive behavior dressed up in
a new vocabulary. We don't just substitute "reference signal" for
"goal," "perceptual signal" for "stimulus," and "output" for
"response." And when someone makes the translations in the opposite
direction, the result is not an understanding of PCT, it is only
switching words and continuing to apply them in the context of the
same old model. In fact, a facile adoption of the terminology of PCT
is an excellent way to avoid getting the point.

The point of PCT is to explain how it is that an organism can select
some physical condition that does not currently exist and bring it
into existence by acting on the environment. The first step in
getting psychologists to understand the explanation is to get them
to accept that organisms can in fact do this sort of thing. There
have always been a few psychologists who accepted that fact,
although usually with poor justifications. But for most of my
career, at least, the vast majority of "behavioral scientists"
didn't even accept that as a valid description of behavior. This
meant that I couldn't even take the first step toward explaining PCT
-- what good does it do to offer an explanation of a phenomenon that
your listeners believe to be illusory? I have spent a large part of
my career just trying to demonstrate that the phenomenon itself

Now that more people are coming around to the view that behavior is
purposive, goal-directed, intentional, etc., it is at least a little
easier to start explaining what PCT is about. Or it should be.
Unfortunately, too many people STILL think that when you have
described a phenomenon, you have explained it. So when Bandura and
Locke say that people pursue goals, and that goals cause behavior
through a method of discrepancy reduction, they think they have
explained goal-seeking behavior. This illusion could be shattered in
an instant if you could just get them to focus on the question,
"What is a goal, that it can have such an effect?" Or, "How is it
that a descrepancy between a goal and the actuality can produce just
those detailed motor activities that have consequences tending to
reduce the discrepancy?"

PCT is sitting here with detailed answers to such questions, and has
been sitting here for nigh unto 40 years, but to no avail: that sort
of question doesn't seem to come up among the likes of Locke and
Bandura. In explaining that behavior is goal-directed, they seem to
think they have reached the foundations. I hope you can understand
how frustrating that is to me. I want to say, "Good, now you
understand the problem. How about listening to my solution?" But
they seem to think that their description of the problem IS the

In fields like AI, neuroscience, and the new incarnation of AI,
Artificial Life, almost the opposite problem exists for PCT. Here we
have explanations galore in terms of neural circuits and system
designs, all good stuff, but practically no appreciation of the
phenomena of ordinary purposive behavior. Thus many of these people
are using their armory of explanatory tools to explain phenomena
that don't occur in nature. Look at the neural net people. What
behaviors are they trying to explain? Responses to stimuli!
Unfortunately, when you approach the organism at the level of the
functions of its components, as PCT does, you can construct
gazillions of models that do _something_. But unless you can tie the
models to the sorts of behaviors that organisms actually produce,
the whole effort is just an exercise in imagination; mathematical

PCT consists of two equally important parts: the definition of the
problem (people seem to behave in very specific purposive ways) and
a model that solves the problem (how they must be constructed in
order to do that). It is necessary to take both parts of PCT
seriously to understand PCT. In fact, you can't really understand
either part of PCT without putting some serious effort into
understanding the other part. A lot of our problems on the net have
come from people with a fair understanding of one facet of PCT, but
hardly any understanding of the other.

The PCT model, with its hierarchical control-loop structure and
mathematical properties, helps us to recognize purposive control
behavior when we see it; seeing examples of real behavior helps us
to select possible architectures and discard others that are equally
plausible on computational grounds alone. This interplay requires
approaching PCT always from both sides, giving neither side a
disproportionate emphasis. Observation keeps us honest; computation
keeps us rational. Which one should we do without?



Bill P.