[From Bill Powers (2003.05.13.1250 MDT)]
Marc Abrams (2003.05.13.1305)--
> Argyris clearly does have a theory (in my terms) about behavior: it is the
> "Cog Sci" theory that we plan the actions we must take in order to achieve
> goals or correct errors and then execute them.
>Was Rick being nit-picky, accurate, or wrong, or are you being loose?
Argyris says plainly that this is the theory he uses.
Designing action requires that agents construct a simplified
representation of the environment and a manageable set of
causal theories that prescribe how to achieve the intended consequences.
It would be very inefficient to construct such representations
and theories from scratch in each situation. Rather,
agents learn a repertoire of concepts, schemas, and strategies,
and they learn programs for drawing from their repertoire to
design representations and action for unique situations. We
speak of such design programs as theories of action.
This theme is repeated over and over: the person uses a representation of
the environment as a basis for designing actions so they will create
intended or desired results.That is the basic "Cog Sci" model that you will
see over and over in the psychological literature. It is what "Modern
control theorists" believe. They all think that behavior is a matter of
deciding on the right actions and then carrying them out, using perception
only afterward to see if the action had the desired effect, and using that
information as the basis for altering their method of designing actions.
It's a plausible model (if it weren't, no intelligent person would adopt
it), but once you see how control theory explains the same phenomena, it's
simply not acceptable.
> I think there's a great confusion going on around this word "theory."
By whom? I know what I mean by theory.
Maybe so, but you don't distinguish in your writing between theories as
explanations and theories as proposed observations.
>Yes. That is why I consider his work a theory about human interactions, in
>which he _uses_ a theory of behavior ( Cog Sci ) to explain those
>observations ( interactions ) I believe his theory is accurate. I believe
>the other theory he uses to explain his theory is wrong.
You see? You're using the word "theory" in two completely different ways
here. Argyris has a theory that people operate by Model I (that's one kind
of meaning). But the Cog Sci theory by which he explains how they do this
is wrong (that's the other meaning).
Instead of calling the Model I idea a theory, why not just call it a
proposed observation of principles that people seem to use? It's a proposed
fact, not an explanation. I'm sure he has methods for checking whether a
given person is using something like the principles he describes, so all he
has to do is apply the method and write down the results. He's not really
_testing_ the idea that people use these principles; he accepts that they
are, and interprets their behavior accordingly.
>To the theoretician, it doesn't matter whether people use one
> set of principles or another;
Is this true of the researcher or practitioner? As a researcher, how do you
know what a "principle" is and how would you differentiate it from a
"system" concept if you did not have a pre-defined "set" of examples of each
to research?
I have an endless supply of principles to look at, my own and those that
other people use. Of course nothing is as clear and sharp at these higher
levels as it is in simple motor behavior, but it's not hard to find
agreement with others about which level is involved in specific cases.
As a theoretician and researcher, I'm interested in testing the basic
theoretical propositions of PCT, so it makes little difference to me which
set of principles (or any other kind of perception) I look at to test the
ideas. Of course as you imply, the practitioner is primarily interested in
practical results and with the specific personal or social problems at
hand. The theory is the framework within which the practitioner understands
what is going on, but the emphasis is more on problem-solving than on, for
example, improving the theory. Practitioners tend to take theories at face
value and apply them.
>any set will serve as an example,and the
> model can be tested by seeing whether people behave in relationship to
> those principles as the theory says they do.
This being true of course if the premise is true and the data used
accurately represents the data theorized.
Huh? Teaching your grandmother to suck eggs again? Of course we test the
premises as far as possible, and the whole point is to check the data the
theory predicts against the data obtained from the real system.
>HPCT doesn't care _what_ principles are involved.
You left out a word here. The HPCT [ theoritician ] doesn't care ... An HPCT
researcher _might_ be, and a practitioner, most definitely does.
Fair enough. I meant that there is nothing in HPCT from which anyone could
deduce a preference for one principle over another.
> Argyris clearly does have a theory (in my terms) about behavior: it is the
> "Cog Sci" theory that we plan the actions we must take in order to achieve
> goals or correct errors and then execute them. THIS is his "theory", as I
> use the word.
No. I disagree. His theory is about _interactions_. He uses someone else's
behavioral theory to "explain" his "social theory".
There's nothing magical about interactions as opposed to actions. Look at
the Crowd program. The program itself simulates individual control systems
pursuing their own ends. But these individuals affect each other and
interact very strongly, and the model shows the results which (as McPhail
and Tucker noted) are quite realistic.
What's missing from Crowd is any social intelligence. The entities have no
concepts of the existence or nature of the other entities (though in one
setup, entities can be given the goal of following another entity). That
would be harder to add, but it could be done. At one point McPhail thought
he had a programmer who would do it, but that fizzled out.
At any rate, the same theory applies whether we speak of individuals or of
sets of interacting individuals.
>This is the "computed output" or "plan-and-execute" model on
> which Rick and I have commented. It has nothing to do with the set of four
> principles that Argyris proposes; it would apply to any set of principles,
> just as PCT would.
Ah! So here is the rub. I am interested in explaining the phenomena Argyris
has accumulated over the years. I think I can. Argyris's explanation of his
observations are inaccurate.
Good. That's how I would like to see it said.
>And it is only in certain limited circumstances that we plan at all before
acting.
Do you know what those "limited circumstances" might be? Have you tested for
them yet?
Of course. A tracking experiment, or any experiment in which a joystick or
mouse is used to control the appearance of a display on the computer
screen, always includes at least one unpredictable disturbance which makes
it impossible to plan actions in advance (successfully). By extension,
planning action is impractical in any circumstance in which disturbances
are large enough that they account for the main part of observed behavior
(like standing up straight, where no actions have to be taken unless there
are disturbances). Planning can succeed only in an environment where the
effects of actions are almost perfectly predictable (where disturbances are
absent or very small), or when the outcome desired does not have to be very
exact to satisfy the controlling person (low gain).
Even when planning is practical, what is planned has to be an outcome, a
perception, and not the action that produces the outcome. There are always
disturbances acting to _some_ degree, and it takes very little disturbance
to make a significant difference when an action is simply repeated blindly,
over and over.
>No but I can certainly see myself planning a trip and what route to take and
>when to stop for fuel and meals and what attractions I might want to see,
>and then of course, adjusting for disturbances as they are encountered along
>the way
Good examples, but if you reflect on them you will see that each element of
the plan is a _reference condition for a perception_ and not a goal for the
_action to take_. The route you follow is a consequence of your turning the
steering wheel right or left (or neither) at each intersection, but your
plan concerns which way you want to perceive yourself turning, not which
way you intend to turn the steering wheel. Your plan is to see a nozzle in
your gas tank while you watch the numbers change, and to see the needle on
the gauge at F -- not to lift the hose and nozzle, move them, and stick
them in the tank, or ask your wife if she has the credit card, or use the
restroom.
If all the main events in the plan occurred by themselves, with no need for
effort by you, you wouldn't produce any action at all, and the trip would
unfold by itself. But of course you have to act, because disturbances of
all sorts exist. You can't even just coast into a gas station when you run
out of gas -- if there's a headwind you will have to stop at a closer
station, and so forth.
> I believe you'll find that the four principles are found in every single
> person Argyris has studied simply because they are defined so as to fit
> anything anybody does.
Actually this is not quite accurate. Argyris has been at this since the
early 50's. His first book was published in 1956. In 1974, he published with
Donald Schon _Theory in Practice_, which introduced his Model I concept and
the 4 principles. They have not changed over 5,000 people, and 27 years. A
pretty good track record.
Or a pretty stubborn man. To say that someone has not had a significant
change in his ideas in 27 years is not exactly an indication of an open
mind. If he sticks to his ideas that firmly, what do you think your chances
are of persuading him to change _anything_ in his approach? You may well be
able to get something worthwhile from his writings, but it may turn out to
be a one-way process. I have met only three people in my life who were
already well-known and successful in their own right, and who then turned
around and accepted control theory as their primary view of behavior:
Donald T. Campbell, Martin Taylor (who said "my theory is a subset of
yours"), and Phil Runkel.
>There may be a deep reason for that -- the statement
> that people try to achieve goals certainly has some deep implications in
> PCT -- or there may be some heavy forced interpretation going on here, as
> it does in Glasser's scheme where every single person can be proven to
have
> the same five basic needs.
I'm sorry I missed something here. When was Glasser "proven" wrong?
I don't
necessarily agree that those are the only five or if in fact those 5 are any
more or less important then any other 2-1,000 that might be uncovered. But
to say he is _wrong_, is to say that system level concepts don't exsist. or
intinsic variables don't exist. Again, he may be pedalling some half-truths
but don't hit him up on his "five basic needs".
He wasn't proven wrong, because you can't prove something like that is
wrong. No matter what anyone does or says, you can fit it into one of
Glasser's needs, so how can you be wrong? That's what makes it a useless
classification scheme. Your basis for saying, for example, that a person is
showing a need for "belonging" will have to change from person to person
and even example to example for the same person, in order to go on saying
it's "belonging" that this person needs -- but you can do it. I've seen it
done, often. In one case the person said he wants to go to the movies, so
you can say that this is a need for belonging because many other people
will be in the theater. In another case, the person said he is a Christian,
which you can point out shows that he wants to belong to a religious group,
and so forth. In another case the person says he wants to be alone, which
means he belongs to a group called "meditators."
Something similar happened with the now-famous studies of facial
expressions as showing emotions. A classification scheme with 8 basic
expressions was worked out. The only problem was that they were defined so
as to cover all possible combinations of ways in which facial muscles could
contract, so it was impossible that any facial expression would not fall
into one of the eight classes. It's like that scam at the racetrack, where
you go around whispering a tip on the winner (in return for a promise of a
small part of the winnings) to as many people as there are horses in the
race -- of course you predict a different winner to each person. You can't
lose.
Popper said that the essential feature of any good theory must be that it
is falsifiable. This does not mean, as one person seemed to say in a letter
I saw, that the best theories are those that are proven false, but that if
there is no way a theory could be proven false, it's worthless. My theory
is that there is a small orange cube on the floor of a crater on the other
side of the moon. Until someone goes there, it's possible, isn't it? You
can't prove I'm wrong, can you? But Popper would throw that theory out:
it's not, for the forseeable future, falsifiable. There's no point in
wasting any energy on it.
That's how I feel about the theory of the 8 basic facial expressions. If I
divided the possible combinations of muscle contractions into 9 groups,
there would be 9 basic expressions, and so on. Same for the 5 needs (which
were 4 at one time). If I classified all possible needs into 11 types,
there would be 11 basic needs.
And at the moment I'm having similar suspicious about Argyris' 4
principles. Why not 3? or 5? or 350?
I'll admit, though, that his description of how people pursue goals does
ring a little PCT-ish bell, even though he seems to think poorly of that
principle since it's part of Model I.
Best,
Bill P.
