[From Bill Powers (931209.0600 MST)]
Bob Clark (931208.1400 EST) --
Despite Martin Taylor's objections, I thank you for injecting a
note of physical sense into the discussion. I agree with almost
all of your statements. I think that people easily forget that
concepts like energy and entropy are not primary physical
concepts; they are derived, they are generalizations from more
detailed physical relationships. While some very useful
computational shortcuts come out of considering these generalized
concepts, they are not a substitute for the underlying details.
This is easy to see when you ask how a system works. If I answer
by decribing only the energy or entropy relationships, you are
left with no conception of what the system is, much less how it
works or what it does. But if I describe the system in terms of
variables and functions relating them, you can calculate energies
and entropies (etc) from the described relationships, and show
that the more abstract conditions are indeed met. That is how
they were found in the first place.
I think that this relationships gets turned topsy-Turvey
sometimes; it sometimes seems that the generalized relationships
are given the status of the more fundamental aspects of real
systems, while the actual detailed relationships are relegated to
the status of mere examples -- as if it is the generalized
principles that make the system work. My position is exactly the
opposite: the way real systems work is what makes the generalized
principles true.
I think this is in agreement with what you say: "[Energy]
is a property of the situation -- it is not 'in' anything, nor is
it any 'where.'" Energy can be reduced to the product of force
and distance, which are measurable aspects of the situation, but
force and distance cannot be reduced to energy.
Energy is an abstraction from, not a cause of, actual processes.
What makes an arm move is not energy expended in the muscles, but
forces applied by muscle contractions. The law of conservation of
energy can be reduced to the fact that forces acting through
distances in a system isolated from external forces remain in
balance with other forces acting through distances within the
same system. If they did not, there would be no conservation law.
ยทยทยท
-------------------------------
As to "dynamics," I think the people who are doing "system
dynamics" are in fact doing what they say: considering the
physical dynamics of certain systems. My only objection is to the
models they choose to analyze dynamically. They don't seem to
realize that other models are possible and in many applications
preferable. The fact that they can analyze ONE kind of system
using dynamical equations seems to have given them the idea that
they've invented a new approach to nature. What they're doing is
like deciding to investigate the properties of phlogiston using
algebra, and characterizing their approach as "algebric systems
analysis" instead of "investigation of phlogiston" (taking it for
granted that algebra is the essence of phlogiston, and never
doubting the existence of phlogiston). The system dynamicists are
investigating a particular kind of system, and the fact that they
use dynamical systems analysis, which is commonly used in many
fields, is incidental. What they are really studying are systems
of a particular type, whatever the analytical methods.
Psychological usages of the word "dynamics" are based only on a
loose association with things that move or change, or that cause
movement ("motivation") or change. There is no rigorous system of
thought behind such usages, so trying to figure out what they
mean is a waste of time.
-------------------------------------------------------------
Hal Pepinski (931208a) --
I can think of numerous examples of "control" phenomena that
have been defined as you seem to define PCT premises, Rick,
such at WI Thomas's theorem that things that are defined (read
"we choose to define") as real are real in their consequences,
or Robert King Merton's extension of the individual Thomas
theorem into a collective phenomenon: the self-fulfilling
prophecy.
In PCT the point is not how you control WHAT you perceive as
reality, but how you act on the world to alter the STATE of
things that you already perceive as real. This is not like
holding up your hand and deciding to see it as a fish, but like
holding up your hand and deciding to see it in a certain place,
like resting on a table. To cause that perception, you must use
your muscles to alter the position of the hand from where you see
it now to where you want to see it. That is the kind of control
of perception that we mean.
The means of altering perceptions in the PCT model is always some
action on the world. It is not just a passive change in the way
your perceptual functions work. You're referring to reorganizing
perceptual systems so that given the same world, you perceive it
differently. Control under PCT means that given the same
perceptual systems, you act on the world to change its perceived
state. You hold up your hand, and turn it first palm up and then
palm down. You still see a hand, but by using your muscles you're
changing the state of the perceived hand -- in this case, you're
controlling its perceived orientation (from your point of view).
Some time back I remember writing that if one takes the
Swahili-language broad view of what "see" means, PCT amounts to
presupposing that we see what we want to see and act
accordingly.
Just the opposite in two regards: (1) we act in order to alter
the state of what we see rather than what we see causing our
actions; and (2) these actions alter the STATE of what we see,
but do not alter WHAT we see. Every item of experience from the
concrete to the abstract has both an identity and a state. The
state consists of all the attributes of the item that can vary
(the lower-level perceptions of which it is composed). A car, for
example, has many variable attributes such as its position, its
velocity, its color, its price, its cleanliness, its gas mileage,
and so forth. When we control a car, we do not control its
identity, but its variable attributes. We continue to see the car
as a car (barring some major reorganization of perception), but
by our actions we can alter the attributes of the car that we
perceive -- those attributes that we can affect by acting. I can
alter the value of my car by using a sledgehammer on it.
Hal (931208b) --
What do you and I get for defining one referent signal rather
than another kind as real for ourselves: one form makes me and
those around me feel out of control; the harder I try to reach
it, the less I succeed. Another form of referent offers me a
real chance to achieve control. One kind of referent for
control is self-defeating; another makes control attainable.
It's perceptions, not reference signals, that get defined as
"real." You're talking about choosing different ways of
perceiving, not about choosing the reference state for a given
perception. Once you have settled on what you consider a good way
of perceiving a situation, one that gives you real control, you
must then select the reference condition for that perception. If
you select a reference value of zero, this means you will act to
bring that particular perception to zero, meaning that you want
it not to be experienced. Setting the reference value to a high
level means that you want to experience a lot of that particular
perception, so you will act to achieve that result. No matter how
you end up perceiving the world, the resulting perception can
still be in many different states, and in order to control it you
have to specify which state you want, and act to correct any
difference that exists.
Suppose you decide that you would like to perceive your own
attitude toward another person as one of acceptance. Once you
have decided to control for acceptance, you still have to choose
how much acceptance to aim for. A little? A lot? Maximum
possible? Zero? You will probably not choose the maximum possible
as a reference level, because you have an agenda of your own that
you would like to be handled on at least a moderately equal
basis. You might set your desired level of acceptance rather high
if you're a therapist, and less high if you're negotiating a
contract, and very low (but not necessarily to zero) if you're
being mugged. So you control the perception in the dimensions in
which it can change. The kind of perception doesn't change; just
the amount of it. The amount of it is controlled by varying your
actions on the world, and thus altering the perceived amount.
I hope this makes "control of perception" clearer.
-------------------------------------------------------------
Osmo Eerola (931208.1703) --
I agree; you would think that any undergraduate control engineer
would find that PCT makes perfect sense. It should be possible to
explain PCT to such an engineer in about 10 minutes. That is what
I used to think, too, and what I hoped would happen. I had hoped
that real control engineers would grasp this model of organisms
quickly, and that at least some of them would start using their
great skills in mathematical systems analysis to carry our simple
model to a new plane of advancement.
Experience, however, has proven otherwise. Engineers are taught a
particular version of control theory that requires some detailed
transformations to become the PCT model. Most of them have proven
to be extraordinarily resistant to making those simple
transformations. The standard engineering approach, apparently,
has been so completely absorbed that any deviation from it causes
great distress. Also, the idea that some unknown person might
have come up with a different slant on control seems unacceptable
to many control engineers; they have identified with their
profession to the extent that any suggestions of change
constitutes an attack, which is repulsed because of loyalty more
than reason. It is hard for an insider to take an outsider
seriously, particularly an outsider who seems to be rocking the
boat a bit.
As others have pointed out, there is nothing new in PCT with
regard to control theory. What is new is the way of matching a
control model to the organization of organisms. PCT can live
without the approval and acceptance of the control engineering
community (we are not without control engineers, and the others
can always catch up). But the main purpose of PCT, which is to
offer a unified framework for the understanding of living
systems, can't be achieved until the massive resistance from the
community of life scientists ceases. In this field we are not up
against just a reluctance to make minor changes. We are up
against a multitude of rotten theories and unworkable methods
which are defended to the death by those who have made their
reputations by using them. Here the problem is not scientific: it
is psychological and political. We are not just rocking the boat;
we are overturning it.
Since you are basically a life scientist, your participation on
this net is at least as important as that of control engineers.
You are certainly in the area where our greatest problems of
acceptance lie.
---------------------------------------------------------------
Martin Taylor (931208.1830) --
[A vortex] certainly IS dissipative, but a sustained vortex,
such as one seen when water goes down a drain at the same rate
as water is supplied upstream, does not vanish. It is a stable
structure, at least it maintains its structure against mild
disturbances. It is not at all like the case of a ball-in-a-
bowl, in which a disturbance supplies energy to the ball, to be
released (into heat) when the ball drops down to its previous
position.
I think a vortex is probably much like a marble in a bowl. To
disturb a vortex you must either add or subtract some angular
momentum from it, which moves it away from its equilibrium state
of minimum potential energy. If you make it spin faster it will
expand; if you make it spin slower, it will contract (I think).
If you maintain your disturbance, the vortex will simply change
shape and size accordingly; it will do nothing to restore its
former state. If you change the physical situation, say the size
of the hole in the bathtub drain, again the vortex will simply
change accordingly; it will not do anything to restore its
original state. The vortex is always in passive equilibrium with
the external forces tending to make it spin and the external
forces tending to oppose its spin. It simply comes to the state
in which the effects of all those external forces are balanced.
It has no preferred state. It has no power source of its own on
which it can call to maintain its state against variations in the
external forces acting on it.
I think you threw us off the track by speaking of the general
flow as a power source for the vortex. It is not: it is a
disturbance acting on the vortex, tending to make it spin faster.
What matters is not the energy, but the forces and momenta
involved. Those forces, once they start the vortex spinning in
either direction, will cause the vortex to speed up until the
counterforces rise enough to bring about equilibrium. We are
talking about the effects of converging forces that happen to
have a somewhat complex effect. At equilibrium, we could
calculate the incoming power and the dissipated power and find
that they are equal. But the forces are what cause the vortex to
work.
I think we have been overlooking something about "closed loop"
systems. It is not just the closure of the loop that matters.
What matters is that it is a _unidirectional_ closed loop, at
least at one point. Sensing the state of the controlled variable
_does not affect the state of the controlled variable_ (to any
meaningful degree). When the error signal creates an output from
an effector, physically disturbing that output does not affect
the error signal backward, through the output function. This is
what makes the control system different from a marble in a bowl,
back EMF in a DC motor, or a vortex. In all of those systems,
action and reaction forces are always equal, when dynamics are
taken into account. In a control system, action and reaction are
NOT equal around the loop.
The neuron, the vacuum tube, the enzyme, and the transistor (and
a few other rarer inventions) are devices that create an extreme
degree of unidirectionality in the interaction between otherwise
normal physical variables. To do this, they must draw upon some
external source of energy, but that is not what is important;
what is important is the unidirectionality which removes the
requirement that action equal reaction, a requirement that holds
in all passive systems.
----------------------------------------------------------------
Best to all,
Bill P.