[From Bill Powers (930329.0700)]
Ken Hacker, Greg Williams (930328 etc.) --
I'd like to clarify some things about PCT and prediction. I've
made some contradictory statements here: PCT is about
understanding, not prediction; PCT can predict behavior better
than other approaches can. This sounds a lot like talking out of
both sides of my mouth, if you haven't been paying close
attention.
The basic PCT model says that the actions a person performs arise
from two independent variables: external disturbances of a
controlled variable, and the setting of the reference signal. In
order to describe behavior using a PCT model, it's necessary to
deduce what the controlled variable is, how the reference signal
is set, and what the external disturbances are. In situations
where this has actually been done satisfactorily, the PCT model
predicts behavior very accurately. The accuracy of the prediction
is an indication of how well the requirements have been satisfied
because, according to the model, accurate prediction would not be
possible if they weren't satisfied.
To establish the circumstances where such predictions can
succeed, it's necessary to persuade the human participant to
agree to maintain a constant goal during an experiment (and
actually to do so), and for the experimenter to know the
quantitative sum of all important disturbances acting on the
controlled variable. Either of these requirements can be
difficult to meet under field conditions where the experimenter
can't predict all disturbances and has not deduced the nature of
the relevant controlled variable and its reference level. This is
very much like the S-R theorist's excuse that one can't control
all stimuli. So what's the difference?
One difference is that under PCT, the point of the analysis is
not to predict the visible (audible) actions of the person, but
to predict relationships between actions and disturbances that
maintain controlled variables stable against the disturbances. In
order to predict actions, EVEN WHEN ONE'S MODEL OF THE CONTROL
SYSTEM IS PERFECT, it is necessary to predict disturbances --
that is, to predict future states of the environment. I may have
a terrific model of your umbrella-using control system for
keeping dry, but I can't predict whether it's going to rain, nor
where your umbrella is going to be when you need it. So when
PCTers claim to be able to predict behavior, it must be
understood that the behavior is almost totally dependent on the
environment. Even knowing the complete structure of a person's
control systems at all levels would not permit predicting
behavior unless the disturbances that are going to occur are
known.
However, a correct PCT model would be able to predict the states
of certain environmental variables with great accuracy even
without any ability to predict disturbances. Those are the
controlled variables that correspond closely to controlled
perceptions. With a one-level model, you can measure the
characteristics of the control system and derive a model that
will predict the controlled variable quite well as long as the
reference signal remains constant. With a two-level model, you
can predict the second-level controlled variable too, over a
somewhat longer period of time, until the second-level reference
signal changes. There is reason to think that as the model is
applied to higher and higher levels, the reference signals and
hence the controlled variables remain constant for longer and
longer times, so that the prediction of the controlled variable
can remain correct for longer times.
However, it continues to be true that behavior -- output -- is
only as predictable as disturbances. For higher-level variables,
disturbances tend to be more predictable, too, because the
effective disturbance is only what can get past all the lower-
level systems and disturb a higher-level variable. So it is
possible to predict higher-level behavior for longer periods of
time. This observation, I believe, is common to most studies of
human characteristics, from whatever viewpoint.
This is why I put the emphasis on understanding more than on
prediction, even though prediction can be very impressive in
laboratory experiments. The more completely you understand the
organization of a person's control systems, the more clearly you
will know what is in principle predictable and what is not, under
realistic circumstances. You will also be able to understand the
relationship between actions and environmental processes when
unanticipated disturbances arise; in effect, every action that
arises when a new disturbance occurs tells you more about what
the person is controlling for. The actions bring disturbances to
your attention; the disturbances test your concept of what the
person is controlling for. The concept of a control system
explains what you are seeing even when you didn't predict it, and
couldn't possibly have predicted it. This helps you to see how to
interact with other people, and to understand how they interact
with others.
···
-------------------------------------------------------
One of the frustrating aspects of our conversations on PCT and
other approaches is that right now there is only a tiny handful
of people actually trying to develop ways of exploring behavior
using the principles and methods of PCT. The rest of you guys,
however justified your objections and criticisms and comments,
are sitting on your butts and defending your personal status quo
and not doing a lick of experimental work to extend the scope of
PCT -- except from your armchairs.
Every person who has ever thought of a new idea and claimed that
it is better than old ideas has been accused of arrogance, of not
showing proper respect for the great achievements of others, of
making overblown claims, of not understanding what is good about
the old ways. Maybe those accusations are correct, but so what,
if the new idea is in fact better than the old ones? People who
love the status quo have always got their knickers in a twist
over upstarts who dare to get cocky about their own achievements
and go around pointing out the Emperor's lack of modest attire.
That's the problem of the complainers. It has always been their
problem. They're responsible for it. They can very easily make it
go away: get into the trenches and start showing that the new
idea is wrong, or get into the trenches and start working to make
it even better. All the wailing and gnashing of teeth and
complaints about violating the comfortable standards of polite
deference to the ideas of others is, in my mind, the sound of
oxen being gored and signifies mostly that the point is being
missed.
If I'm arrogant or Rick is arrogant, we have a right. We're doing
the work, not watching a TV show.
-------------------------------------------------------------
With sweetness and light,
Bill P.