Friendship

[From Rick Marken (960716.0845)]

Me:

When I said that threading a needle blindfolded can't be done, I meant
that it can't be done as needle threading is normally done; ...

Hans Blom (960715) --

Then everything was just one big misunderstanding. Now, can we be friends
again?

I'm sure you're a real nice guy, Hans, but I'm afraid friendship is not in
the cards for us -- until, of course, you start saying and doing things that
reflect an understanding and acceptance of the fact that organisms are
perceptual control systems.

I know that your interest is in "improving" PCT through the addition of model-
based control. The only problem with this is that you have presented no
evidence that the basic control model needs improvement (we're currently
accounting for 99% of the variance in most of our data) nor have you
presented evidence that the model-based system is the improvement that will
account for whatever unexplained variance exists in our data.

I find your efforts to "improve" PCT particularly annoying, not just because
there is no evidence for the necessity of these improvements (yet) but
because the particular improvements you suggest are precisely the models of
behavior currently in fashion with conventional behavior theorists (who,
incidently, will not give PCT the time -- or the light -- of day). Model-
based control has nothing to do with perceptual control; regardless of it's
merits in engineering design (which I believe are minor), model-based control
obviously scores points with life scientists because it is compatible with
their cause-effect view of behavior. Offering this model as an addition that
"improves" PCT strikes me as being similar to offering "impetus" as an
addition to Newton that obviously improves his "limited model" of dynamics.

I know that you think we all just see the world in a different ways and that
we all should be friends because it's all just a matter of opinion. But I'm
afraid that I see the world quite differently. I think that, in science, some
models of reality (imagined realities, as Bruce Gregory would say) are better
than others; the better model is determined by observation (experiment). I
labor under the belief that observation shows that PCT is the current best
model of behavior; this model is completely inconsistent with cause-effect
models of behavior, like model-based controllers. This is just the way it is.
I believe that, in science anyway, there is a right and a wrong (model of
behavior, in this case).

I also find that the difference between right and wrong models of behavior
_matters to me_. I believe that all people operate on the basis of an
implicit model of the nature of the other people they deal with and that many
people -- including policy- makers, executives, parents, teachers, etc.,
people who can make a difference in the quality of people's lives -- have
been operating on the basis of the wrong model of people. I think the world
will be a better place (for all people and living things in general) once we
are operating on the basis of something more like the correct model of human
nature -- PCT. It is important to me, therefore, that people understand PCT.
So it's hard for me to be friendly with people who don't understand PCT- -
particularly those who won't try or those who say they do understand it when
they say things that clearly indicate that they don't.

Since I think the evidence for PCT -- and against any cause-effect model of
behavior -- is overwhelming, I also find it difficult to be friends with
people who may understand PCT but don't want to accept it. I don't dislike
such people; I just find it hard to deal with them. This is especially true
when I know that these people have had access to the data relevant to making
a judgment about PCT. I guess I assume that people who have seen the evidence
and still don't accept the fact that people are perceptual control systems
simply have another agenda. I feel the same way about people who are familiar
with the evidence in the OJ trial and still think OJ might not be a murderer;
there is obviously another agenda there -- and it's hard for me to be friends
with people who have agendas that get in the way of their sweet reason.

So I don't dislike you at all, Hans. It's _some_ of your ideas -- the ones
that are inconsistent with PCT that I dislike. But as long as you cling to
those ideas, I'm afraid I'm not going to feel all friendly and cuddly towards
you. That's not much of a loss for you, though. I'm kind of a bore in person
and you'd get sick of my Bob Dylan imitations in no time;-))

Best

Rick

<[Bill Leach (960715.1231)]

[From Rick Marken (960716.0845)]

I find your efforts to "improve" PCT particularly annoying, not just because
there is no evidence for the necessity of these improvements (yet) but
because the particular improvements you suggest are precisely the models of
behavior currently in fashion with conventional behavior theorists (who,
incidently, will not give PCT the time -- or the light -- of day). Model-
based control has nothing to do with perceptual control; regardless of it's
merits in engineering design (which I believe are minor), model-based control
obviously scores points with life scientists because it is compatible with
their cause-effect view of behavior. Offering this model as an addition that
"improves" PCT strikes me as being similar to offering "impetus" as an
addition to Newton that obviously improves his "limited model" of dynamics.

Model based control is anything but "minor" in engineering design. One of
the very few engineered controllers that I can think of that is not usually
thought of as a "model based" controller is the "go/no go" thermostat that
we so often use for examples here on PCT (and even it is designed based upon
models).

Most controllers (engineered at least) are "self-tuning" (this is really
what "proportional gain", "derivitave gain", and "reset" actually are when
adjusted by the controller itself) and the tuning capability is based upon
models of the environment in which the controller functions.

The first controllers that began "wearing the term 'model based' on their
shirtsleaves" were the adaptive controller (probably the second or third
generation versions). These are essentially controllers, usually with
considerable memory and analysis power, that "learn" how their output
functions effect the perceptions under control. To say that a bit more
accurately: These controllers monitor their own control error and change
the parameters (and even the output functions used) to reduce average and/or
peak error based upon operating history.

The manner in which these adaptive controllers actually work is quite
possibly closely related to how a living control system might function
(remember the that original "govenor" was really nothing more than a
replacement for a human operator). One of the difficulties in discussing
such systems is that the "control of current perception" that ALWAYS exists
in such controllers (at least in the ones that actually work) is often very
difficult to find (recall Bill Power's efforts with Hans' relatively simple
example).

We know from many sources that _new_ motor function tasks are controlled
poorly if at all when first attempted. We can often observe that improvement
can be either "smooth and gradual improvement" or seemingly sudden quantum
improvements usually with additional smooth and gradual improvement. While
far from any sort of proof, this observed behaviour is very suggestive of
both "self-tuning" (smooth, gradual improvement) of (an) existing control
loop(s) and "adaptive control" (quantum change in control quality)
suggesting a fundamental change in control loop structure.

As much as Hans seems to be argueing for "generated output" I get the
feeling that he does not really mean it the way that most of us usually take
him to mean. Though with that statement I will also assert that it is his
responsibility to develop a way of speaking on the CSGNET that is consistant
with our terminology to express his ideas.

HPCT concepts from the "configuration level" and above probably all involve
modeling ideas (though admittedly while the thoughts about the configuration
level are certainly model based I am not so certain about the functioning
within the level itself). The program levels and above can be argued to be
model based without much difficulty. Program level discussions almost always
use model based concepts (controlling for having milk available and all of
the attendent control of current perception necessary to bring the
_currently_ furture perception under control as a current controlled
perception sometime in the furture, for example).

I am sure that the wording of my previous paranthetical statement is
confusing at best but I worded it the way that I did because the difference
between "control" of a future perception and "control of perception" as we
normally mean the term is crucial to the PCT understanding of model based
control.

In a sense, the "go by a gallon of milk" program (probably driven by the "I
want to perceive that I have milk available control loop") will "generate"
output. In this case, references for current perceptions that, if
controlled, will, based upon the model, result in satisfying the "milk
available" condition. That the store selected for the purchase is out of
milk, the road is closed, the car breaks down, your wallet has been stolen,
etc. are all irrelevent as far as the original program(s) is/are concerned.
Indeed, a control failure due to one of these "disturbances" may appear to
the observer to "trigger" a different program. Of course what is happening
is that if the reference for the "future" perception still exists and is at
a higher priority level than other controlled perceptions (degrees of
freedom issue) the organism will continue to try various control loops and
"methods" to reduce the error to within acceptable limits (again in this
case some minimum quantity of milk available).

Bill's posting of last night is an excellent example of how model based
control might be involved in our lower level PCT control loops as well as an
excellent example of why we always seem to be talking past Hans (and vice
versa of course).

This leads me to the idea that (at least some) "learning" vs. the sort of
"control of future perception" that the milk example represents are
qualitatively different. Adaptation and tuning of lower level control loops
has been posited as a form of learning and this process is undoubtedly also
a model based control operation.

I (maybe arrogantly) believe that Bill Powers and myself are very much in
accord concerning model based control (we have discussed this subject quite
a bit) though naturally what I say is my opinion and does not carry his
influence or necessarily his approval.
-bill
b.leach@worldnet.att.org
ars: kb7lx

[Hans Blom, 960717b]

> Then everything was just one big misunderstanding. Now, can we be friends
> again?

I'm sure you're a real nice guy, Hans, but I'm afraid friendship is not in
the cards for us -- until, of course, you start saying and doing things that
reflect an understanding and acceptance of the fact that organisms are
perceptual control systems.

Rick, thank you for your long and carefully worded reply. Although,
as you correctly note, my world view is different from yours, I think
I now understand yours better. I accept it as honorable. Let's just
leave it at that.

Greetings,

Hans