Seeing it my way

[From Rick Marken (950912.0930)]

Chris Kitzke (950905.0030) --

Rick, can you ever expect anyone to see anything your way?

Yes. Science is really about trying to get people to see things your way.
This was Bill Powers' point in two marvelous posts (950905.0830 MDT) and
(950907.2000 MDT) that I think would be worth re-reading.

I have been trying to see things Bill Powers' way for the last 15 years. I
think I'm pretty close to seeing things his way -- the PCT way.

In science there are two aspects of "seeing things another person's way". The
first is learning what "their way" is: with PCT this means learning Powers'
approach to the control model and how it maps to behavior. The second is
trying to show that "their way" of seeing things is wrong! In PCT this means
subjecting the PCT model to constant, challenging experimental test.

In order to properly subject another person's way of seeing things to
experimental test you have to be sure you have got "their way of seeing
things" correct. In physics, this means that you have to learn Newton's laws
and how to apply them. In PCT, this means that you have to learn Powers' laws
and how to apply them. The only way to learn another person's way of seeing
things is by trying out various understandings of how the other person sees
things until the other person (or a qualified representative) no longer
corrects your understanding. This is how physics classes work; it's also how
CSG-L works.

Of course, another person can't make you see things their way; they can just
hope that you will be smart enough to see that they seem to have something
there -- and that you will, then, try , of your own
initiative, to see things their way.

Once you think you have a pretty good idea that you are seeing things the
other person's way (getting A's in physics shows that you know Newton; being
rarely contradicted by Bill Powers shows that you know PCT) then you can
start subjecting this way of seeing things to the test; you can start trying
to find holes in the other person's way of seeing things. It is this second
aspect of seeing things another person's way -- subjecting their way of
seeing things to rigorous, skeptical experimental test -- that distnguishes
science from religion.

It took 200+ years of RIGOROUS experimental tests to find any significant
holes in Newton's way of seeing things; I predict a similar life span for
Powers' way of seeing things. Unfortunately, there are not nearly enough
people trying to show (BY EXPERIMENTAL TEST) that Powers way of seeing things
is WRONG.

The greatest respect you can show for another person's ideas is to try, like
mad, to show, via experimental test, that those ideas are wrong. The greatest
disrespect you can show for a person's ideas is to try to show, by argument
and reasoning, that those ideas are wrong. The former approach is the path of
science; the latter is the path of inquisitorial religion.

How could the goal of learning PCT be to make things better when wants are so
different, not to mention perceptions?

I don't think the goal of PCT is to make things better; it's to CORRECTLY
understand how people (and all living systems) work. We have to understand
this before we can have any hope of making things better.

Ken Kitzke (950912.1007) --

Have we discovered a new theory that better explains "how" people act to
"control" their perceptions while producing the same conflict, anger and hurt
so prevalent among the multitudes of misguided scientists and
practitioners? Is that all there is?

Ideas, no matter how good, do not guarantee good behavior (goals).

Can we work together cooperatively to expand the understanding and successful
application of PCT? Can we jointly endeavor to keep the theory pure?

I think we should endeavor to keep our understanding of the theory accurate
(not pure -- sorry, that word gives me the creeps). But if you want
successful applications of the theory we need criteria for success. So far,
none have been suggested. If you want to expand understanding of the theory I
suggest that the best way to do that is test the hell out of it and try to
show that it's wrong.

diatribes have crept into our dialogues.

Bill Powers - always so thoughtful and helpful but your perceptions about
competition will never get you on the "Tonight" show..."Good competition
makes everyone better."

This is interesting. You are making a plea for harmony and cooperation on the
net while arguing for the virtues of competition in the marketplace. If
competition makes for better late-night TV, why doesn't it make for better
PCT applications? I hope this question doesn't seem too competitive -- or too
cooperative;-)

Hans Blom (950912d) --

Ok. Challenge mode off.

Gee, that was quick.

And my point was that "accurate control" by the classical PCT type of
controller is possible only if it is designed for the environment it is to
operate in. Not exactly, maybe, because a PI-controller is rather robust,
but approximately.

This has never been in dispute; the control system must have the right
polarity and dynamics; but that is a LONG way from saying that it must have a
model of the environment.

Can we finally agree on this now?

Yes. We can agree that a control system must have negative feedback and
appropriate dynamics. What we don't agree on -- not even close -- is that a
control system needs a model of the environment (like the one in your model-
based control program) in order to control like a living organism.

The point is that you have already specified what this system can
learn: exactly nothing.

This is wrong, but irrelevant. We are not dealing with learning (yet). We are
dealing with an existing control system. How control is learned is a
different problem. I understood you to be claiming that an existing control
system (one that has already been learned) can control only if it has a model
of the environment.

A conceptual problem: disturbances. In my vision, disturbances do not exist
in the real world.

Why make this difficult, Hans? You know how to program a compensatory
tracking task, right. Disturbances clearly exist in what most of us would
consider the "real world" part of that task. So why not set up this tracking
task and show us that a model - based control system is needed to control the
cursor? Or that a model-based control system can control better than a simple
PCT control system?

This is my challenge to you: show that a model-based control system controls
as well or better than a simple control system in the simplest control
situation. If you can't show this, then I will have to assume that your
continued discussion based on the assumption that organisms use model-based
control is based on a persistant hallucination :wink:

Best

Rick