PCT and chaos (was Economics--a view...)

[Martin Taylor 2004.03.06.1313]

[Peter Small 2004.03.06]

[Martin Taylor 2004.03.05.10.00] wrote:

I can hardly think of a more inflammatory comment that could be made
on this mailing list, which is predicated on the concept that
perceptual control is the only effective model of what really goes on
in the real world!

Perceptual control is fine in two-person situations but becomes
vulnerable to chaotic instability if more than two people are
included in the same model.

True, depending on the coupling parameters. But in my view, that has
little to do with the viability of PCT.

The current thread "Zero Sum Economics" is evidence enough that
perceptual control is not a suitable model to represent real world
situations.

I disagree. I'm not even clear why you say it, but I imagine you do
have a clear idea. My suggestion is that you start a new thread (i.e.
change the subject title--actually, on re-reading this message, I've
already done it) to proceed with this aspect of the implications of
PCT. It has wider implications than just in economics. It relates to
the entire world of social interaction and culture.

[Martin Taylor 2004.03.05.10.00] wrote:

The perception of risk is among those that can be controlled at some
desired level

That seems to arise naturally out of the interactions among control
systems, at least one of which is controlling the perception of risk.
Perceptions of future costs, possibilities, worst-case scenarios...
are all controllable perceptions.

Risk is an assessment of probability. Probability is determined by
the uncontrollable and unpredictable disturbances of the system (from
internal and external sources). So, I don't see how perception of
risk can be considered controllable. Perhaps you could enlighten me
here?

OK. One can increase risk by walking closer to the edge of a
crumbling cliff, or decrease it by walking further inland. If one is
controlling also for seeing what is at the base of the cliff, there
is a conflict between controlling that perception (that you are
seeing the cliff base) and the perception of risk (assuming risk is
being controlled at a low reference level). The conflict can be
reduced by controlling the risk perception using a different action,
such as lying down at the cliff edge while being tied by a stout rope
to a big tree further back.

Risk is a perception that has as inputs not only the perceptions of
the probabilities of different events, but also the perceptions of
the distribution of consequences of those events. The probability of
the cliff crumbling will not change by the act of tying oneself to a
big tree, but the perception of the consequence of the cliff
crumbilng changes drastically, and so does the perception of risk.

[Martin Taylor 2004.03.05.10.00] wrote:

I don't think "designed" is the right word to apply to evolving
systems. Bill Powers has always asserted that the alteration of
control structures to accommodate the real world usually cannot be
designed. The effects of structural changes has to be explored by the
system as a whole. The colloquial term often used to describe this
process is "e-coli learning".

"Designed" is a good a word as any, although "Strategy" might be more
appropriate. However, Bill Powers was talking about the PCT control
structures when he says they cannot be altered to accommodate the
real world. He is right. But, I'm not referring to the PCT control
structures; I'm referring to the design of the complete system.

Indeed, that's what we are both talking about. It can't be designed.
I've never seen Bill Powers say that the PCT control structures can't
change to accommodate the real world. He has a full theory of how
they do exactly that. It's called "reorganization" in PCT-speak. The
mechanism of reorganization may not be fully described. Indeed,
several differen local and global mechanisms have been proposed, but
without experimental test there's not way to choose among them as to
which more nearly describes real life.

It may seem paradoxical, but although you cannot predict how a
dynamic complex system will behave, you can control it to behave in
the way you want it to behave. The basis of this control is to
"design" a system that lives on the edge of chaos and "design" ways
in which it can be disturbed.

The argument is that natural systems tend to evolve to live on the
edge of chaos for this very reason. Structures of elementary control
units are no different. But it is very hard to design systems that
live on the edge of chaos, if only for the reason that the coupling
constants in the environment have a habit of changing unpredictably.
Evolution seems to be the only practical way to get there, and
evolution seems almost always to arrive there.

I think you must be aware of this distinction, because of how you
responded to the last paragraph of my post:

[Peter Small 2004.03.05.10.00] wrote:

Without having a cognitive framework that includes the concept of
dynamic complexity and chaos, controlled systems are very hit or miss
affairs.

[Martin Taylor 2004.03.05.10.00] responded:

An insight that has not been lacking in the PCT world, though it is
often ignored because much of the simulation and experiment is done
on much simpler systems.

In view of the many misunderstandings, perhaps my inclusion in this
discussion group is not appropriate.

On the contrary. I think it is most refreshing to have someone else
who is interested in the same kind of issues as I have been. Perhaps
misunderstandings are to be expected, but when one has been observed,
there is an opportunity to correct it. It is the unrecognized
misunderstandings that cause trouble.

The control systems I am
considering are using concepts that are quite different from those
used in designing PCT control structures.

Again you use this word "designing". It's an engineer's concept (I'm
one, too), which implies the existence of a "designer". Who would
that be?

What we have to look for is how control systems evolve to be
effective in a complex environment, not how they should be designed.
Given the simplest plausible underlying elementary structure, and
some specification of an environment including the rules of
development (such as Powers' notion of reorganization of the control
structure, or Hebbian learning within an elementary control unit),
what is likely to happen?

I look forward to further interactions between us--chaotic or complex :wink:

Martin