[From Rick Marken (940223.1500)]
Martin Taylor (940223 12:00) --
re: self-organizing systems
What doesn't work without nonlinearities is the amplifier that
provides the gain, without which there is no effective negative feedback.
The non-linearities are not necessarily in the feedback loop itself.
So a self-organizing systems can be represented by the following
system of equations:
o = f(i)
i = g(o) + h(d)
There is no reference input; output is a function of input only and
input is a function of output and disturbance. This is a negative
feedback loop if f() is an inverter (o=-i). The amplification in
the loop is provided by g() -- and you are saying that g() amplifies
becuase of non-linearities that are not in the loop itself, which is
fine with me.
So a self-organizing system is a negative feedback control system. The
quality of control will depend on the amplification provided by g()
(assuming appropriate dynamic stabilization). Assuming reasonably high
amplification, the system will keep i = 0.
Me:
Well, then a self-organizing system must control
a perceptual variable relative to a fixed reference level (zero).
Martin:
If you care to identify a perceptual variable in the loop, you can certainly
say that it controls to a fixed level, but I don't see how you can say
that the fixed level is zero, unless you do it be definition.
I meant that it keeps the perceptual variable (or i if there is no
perceptual transform of the input) at a value that corresponds to
a reference SIGNAL level of zero. If there is no reference signal
input, the control system acts to keep the controlled variable = 0.
If the controlled variable is the untransformed input variable, i,
then i will be kept at 0. If (as is more likely) the controlled variable is
actually a function of i then the value of that function of i will be kept
at 0 - - and the input, i, will be kept at whatever value results in the
value of the function being at 0.
In a self-organized system, there need
not be any such stage of transformation, and one might identify the
CEV itself with the perceptual signal. In other words, there's no
obviously necessary "inside" and "outside," to which parts of the loop
can be assigned.
Interesting point. In living control systems there is an obvious
"inside" and "outside". But there are many artificial control systems where
the boundaries are not obvious.
Me:
One quick question -- just what is it
that we're trying to explain? Is self-organization control? If not,
what is it? and why is it explained by a control system?
Martin:
Who we, LA-man? You said self-organization was a dormitive principle,
explaining itself.
In the American Psychologist article (which I now find might have been
the wrong one to turn to for information about self-organizing systems)
phenomena (like bifurcating chemical reactions) were used as models of
other phenomena (like multiple personalities). This is a version of
dormative principle where a phenomenon is explained in terms of a similar
appearing phenomenon. So instead of saying that multiple personalities
result from a tendency toward multiple personalities, the self-organizing
approach (in American Psychologist) says that multiple personalities result
from self-organizing systems that produce multiplepersonalities. It looked
pretty dormative to me.
What I tried to explain to you was that self-organized
structures appear through feedback loops dependant on non-linearities
in far-from-equilibrium energy flows.
But you failed to explain what a "self-organized structure" is. It sounds
to me like self-organized structures are the phenomenon we are trying to
understand. What is a self- organized structure? How does it "appear through
feedback loops"? Is a self- organized structure the same as a controlled
variable? If not, why is a control system needed to model the existence of
such structures?
They almost always appear in such
flows, and that is the phenomenon for which the feedback process provides
an explanation.
You have not described what "they" (meaning self organized structures) are.
How do I know one when I see one? PCT may have been a crazy cult but they
did have a nice clear and replicable way to identify controlled variables.
Does self-organizing systems theory have a clear and replicable way to
identify self-organized structures. I hope it's not something I can only
detect after long periods in sensory deprivation (though it would be fine if
were something I could only detect after imbibing copious quantities of
Jack Daniels).
In your language, is that control?
No. In my languade it's nothing at all -- little more than saying that self-
organized structures are self organized structures. I still have absolutely
no idea how I would identify a self-organized structure. That PCT cult keeps
looking better and better -- even if they are mean to me.
I'm not proposing to play word-games here;
Nope, you're not proposing; just playing.
I'm not sure crowd rings qualify as self-organized structures (but I'm
not sure they don't, either).
Very candid of you. Now, could you please point to a SIMPLE example
of a self organized structure -- and a SIMPLE explanation of how
a negative feedback control systems (with no reference input) explains
its existence.
Oh, and while you're at it, could you also tell me how I can measure the
undercertainty of the disturbance given the percpetual signal to a control
system. If I don't re-join the PCT cult, and the self-organizing systems
group doesn't work out, maybe I'll give information theory a shot.
> Leave PCT for a study of self- organization if you want.
Well, I'm teetering on the verge of going back to Bill Powers with hat in
hand. [Poor Bill. Jesus had these kind of problems with Peter, didn't he?]
I must say, my short stay in the self-organizing systems world has been
most non-illuminating.
Richard Thurman (940223.1315)--
In instruction and
in learning -- teachers, coaches, and lovers 'provide feedback' to us
by telling us how we are doing.
Isn't this type of 'feedback' the same Gary C. and Bill P. were talking
about last August when they wrote of coaching speed skaters.
Well, yes. The teacher is doing some "providing" all right -- but it's
not really feedback that's being provided -- just a perception. In
the case of coaching speed skaters, it's a perception that the skater is
willing to take as a surrogate for another perception he apparently can't
have (of the "lowness" of his body). The skater is trying to control
for having the coach stop saying "lower". The caoch is just part of the
feeback loop between the skater's outputs (body position) and the
words heard ("lower" or not). The coach may feel like he's "providing"
feedback but he is really just providing a perception that the skater is
free to control (or not control) for as needed (for other, higher level
reasons).
Rather, many students are actually controlling for the perception of a
grade, or for perceiving the coach yelling "Thats It!" In this manner
the teacher, or coach is part of the feedback function.
Yes. The coach is part of the feedback loop. The student can control
the percppetual variables in that loop (or not) as needed.
Is this coherent? How am I doing?
Yes. Fine.
I need some feedback (oops--input)!
Sorry, I can only provide perceptions (or, more appropriately,
disturbances -- see, Bill Leach, I listen).
Best
Rick