It's good to get reports about PCT making a difference somewhere.
Thanks.
···
-------------------------------------------------------------
Oded Maler (930510.1000 ET) --
In principle it might be that higher-level variables (H) are
controlled like lower-level (L) ones, but how much is this
*likeliness* meaningful?
The *likeness* is clearly not at the level of muscle tensions,
but is in terms of organization. You said somewhat jokingly
I was "going to Paris" controlling for an audition as a means
to control for a permanent reserch position as a means to ...
How about thinking seriously of these activities as control
processes? Each each of these cases, you can think of the process
as producing output, or as producing perceptions. The latter, of
course, would be the PCT interpretation.
You can anticipate the questions a PCT researcher would ask: how
do you know you're going go Paris; how do you know that what is
happening is an audition; how do you know what kind of position
you are being offered? In each case the answer is that you can
perceive the thing in question. And you could be asked about the
nature of the corresponding reference signals, and the errors
that may or may not exist, and the actions you take when
disturbances occur that tend to make the perception differ from
what you intend to percieve. Your reference perception of the
research position, for example, might be very specific, or it
might concern only a few general desires, with many different
research positions (but not all) being able to match it. You
would probably not accept a research position calculating the
statistics of spoiled meat in a slaughterhouse, for example. So
there is some reference picture for what you consider an
acceptable position. And so forth.
The control of L and H are made on time-scales which are of
different orders of magnitude, and the variables themselves
have completely different relationships with their
"corresponding" EVs.
True, but isn't this already part of HPCT? The differing time
scales are discussed in BCP, and it's inherent in the model that
the higher perceptions correspond to different types of external
situations than those controlled by lower systems. However, the
lower-level perceptions are the elements from which the higher
are constructed; higher perceptions are not drawn directly from
sensory receptors. Going to Paris entails control of many lower-
level variables: relationships, events, transitions,
configurations, sensations, and intensities. You have to adjust
reference signals for all these lower variables in order to
achieve the result that is perceived as going to Paris.
Take a perception of muscle tension and your perception of,
say, the correctness of US economic politics. The former has
at least some hypothetical relation with an electrode one can
put in the appropriate place in your body. Can you suggest a
similar CEV whose measurement will have some relationship with
your higher-level perceptual variable?
Yes: a similar perception in another person. The "correctness of
US economic politics," however, is not a single variable but a
collection of independent dimensions, each one of which could be
modeled as a controlled variable. Correctness, in any event,
can't be measured in the environment; it's an opinion, and as
such exists in someone's head as a perception and a reference
condition. When you start explaining what you think is wrong with
the policies, you have to start naming variables that, in your
opinion, are being maintained by the US in the wrong state, and
then you have to say what you consider to be the right state so
you can define what the error is.
According to the current HPCT model, when you have isolated a
particular one-dimensional variable as a variable controlled by
some person, the implication is that you could put an electrode
in that person's head in the right place and measure the
magnitude of the perception that represents the state of that
controlled variable. To encompass a complex perception you would
have to find the degrees of freedom involved in it and through
the Test discover the specific state of each dimension that is
desired; then you would need an electrode for each dimension. If,
from all these dimensions of the perception, the person is
perceiving some unitary thing, then we would expect that at a
higher level there would be a single perceptual signal that is a
function of all the measured ones. An electrode at that level in
the right place would measure the magnitude of that single
perceptual signal.
This is quite unlike the "distributed" concept of perception that
others have offered (Martin Taylor, for example). So I suppose
that some day we will find out which model is right.
Can you poke some millions electrodes in the hearts, stomachs
and pockets of (even a sample of) the American population and
define a CEV as a function of those?
That would just yield the average perception of the correctness
of US economic politics; it wouldn't tell us anything about the
"objective" characteristics of those politics. Correctness is
relative to an observer who perceives something and compares it
with a desired state of the perception.
Could you find a resonable correlation between this
hypothetical signal and your percept?
Someone might, ultimately, be able to measure the correlation
between signals indicating one person's opinion on this subject
and signals indicating another's, although I suspect that it
would be simpler just to ask. In the external world there is no
correctness or incorrectness; things simply happen and affect
each other as they do. The result is desireable or logical or
proper only in terms of what specific people perceive and want to
perceive.
I think that describing, say, the Iranian revolution, as "yet
another revolt of mideastern people against an helenic-
influenced elite", is not less useful description than whatever
you might say about perceptions of individuals - were they
perceiving themselves as Persian or Muslims, as poor or rich,
etc.
As a generalization it is probably as good as any, but no better.
It simply reflect someone's way of applying the principle that
what has happened before is likely to happen again. It tells us
nothing about why people whom we characterize as " mideastern" do
something called "revolting" against people whom we characterize
as "hellenic."
To explain why such a generalization seems to hold true, we would
have to explore the properties of these peoples and their
relationships to each other. For example, it may be true that
those whom you categorize as "hellenic" have a tradition of
trying to control the behavior of those you categorize as
"mideastern," or of acting in a way that violates the preferences
of mideastern people. If that is the case, we can explain the
revolt easily as the natural consequence of trying to alter
arbitrarily the behavior of a lot of control systems.
If this is a recurring situation, as your generalization
suggests, then we would have to ask why hellenic peoples keep
doing what they're doing even though the result is always revolt.
This gets us into the question of why people follow traditions --
that is, why significant numbers of people adopt, in this case,
goals that bring them and their nations into guaranteed conflict
with millions of other people. We are now talking about the way
higher levels of perception and control operate in individuals.
Mass phenomena such as this suggest the possibility of human
characteristics that are common to many individuals, leading to
similar behavior in at least many of them.
So by leaving the generalizing mode and beginning to look for
actual explanations, we can come up with the beginnings of
hypotheses about how people work, and start looking for control-
system phenomena (which are not hard to find). We would probably
find that there is nothing particularly unique about hellenic
peoples -- that mideastern peoples do the same things to each
other and to Africans as well, producing similar revolts. In
fact, we would probably conclude that people all over the world
make this same mistake, trying to get what they want for
themselves without considering the effects on other control
systems, with the inevitable outcome that control theory
predicts.
In your tracking experiments, since you know that whatever the
controlled variable is, it is a function of the "objective" x,y
coordinates on the screen, you can be pretty sure about what is
controlled (although you'll never know thru which lower-level
paths this perception of x and y emerges).
It isn't necessary to know what the controlled variable is, and
in fact we have done experiments in which a person's action
affects many variables simultaneously, with the person silently
selecting one of them to control. By applying one or another
version of the Test, we can deduce quite reliably which variable
is under control. And as to the lower-level paths by which this
control is accomplished, we can't narrow that down to a specific
neuron, but it's pretty clear that the path involves brainstem
and spinal control loops, some of which have been traced out in
considerable detail -- enough to permit building a working model
of the lower systems. This model could be incorporated in our
models of tracking tasks, but as the behavior of the model makes
the relationship between arm position and the reference signals
specifying that position a very simple function, there is no need
for that elaboration. The lower control systems are adeqately
represented as a time constant and a delay, which appear in the
tracking model without the details that create them.
In general, the very idea of being able, even in principle, to
construct global dynamics based on simple local rules, looks
much less promising these days - this is all the Chaos fury is
all about. (It is not only a trendy trend as some may think).
But control systems convert chaotic relationships in the
environment into regular, repeatable, intentional relationships.
The affairs of the world are conducted, at any given time, by a
very small number of people in positions of power, who can launch
or not launch armies as they judge necessary to achieve their own
purposes. Human affairs are produced by intentions, not by
chance. If you have a way to deduce intentions, such as the Test,
you can understand why people do what they do even on a global
scale. Or we would be able to do so if people would start using
control theory instead of all the old traditional methods that
are incapable of leading us to such understanding.
As for the specific levels you have chosen, I don't know if
they are explanatorily optimal, and whether they are invariant
over all humans and cultures.
Neither do I. I've been pleading for years for some research to
be done on this question. So far, people seem content either to
memorize the levels and apply them as if they are written in
stone, or sit back and think of hypothetical problems with them.
The question concerning why perceptions aggregate higher and
higher in this form or another is both a mathematical and a
neuro-biological question, which currently is far above my
competence (but the situation is not hopeless, considering the
current rate of my progress in understanding continuous
dynamical systems, I might be able to consider such problems
before 2000).
Well, good -- at least you're considering working on the problem
some day. In the meantime, it is also possible to find out
whether the levels exist and are correctly defined by seeing what
kinds of variables people actually control, at different levels.
There is a vast untapped source of research projects here, which
anyone interested in human behaviour could start exploring at any
time. Neurobiological and mathematical research relies heavily on
the way we informally characterize behavior. If we can tell the
mathematicians and neurobiologists what types of variables people
actually control, these researchers will have a far better idea
of what they're trying to model.
----------------------------------------
I just browsed and so part of the "attractor" discussion. I
view attractor as a very useful mathematical term, telling you
to what state is a dynamical system going to flow given some
region of initial conditions. A good control system is a one,
when considered together with the disturbing environment, has
its reference signal as an attractor.
Why use metaphors when you can speak literally about how the
system works? The whole language of attractors and basins of
attraction is based on dubious and often wrong analogies. A
"basin" of attraction is a metaphor built on the behavior of a
ball in a bowl. But there is in fact no attractor pulling the
ball toward the center of the bowl; the actual "attractor" (if
you want to think of gravity that way) is the pull of the earth
on the ball STRAIGHT DOWN. It is only the constraint applied by
the surface of the bowl to the ball that converts the actual
attraction into an acceleration toward the center.
It's the notion of a basin of attraction that has lead dynamic
systems analysts like Kelso into thinking that a control system
(in which the reference state is analogous in its effects to a
point attractor) must naturally settle down to a limit cycle.
This is, of course, a model of a malfunctioning control system
(no matter how small the cycles get, Avery). But the dynamic
systems analysts don't know enough about real control systems to
realize that, so they draw all sorts of erroneous conclusions
about control processes -- and print them in the literature where
they will launch yet another train of authoritative
misconceptions.
You can get the appearance of a point attractor from a system in
which there is absolutely no oscillatory phenomenon at all: for
example, a race-car driver steering around a circular race
course. If you just look for mathematical forms, you will
erroneously apply the mathematics of attractors, and be led
completely away from the correct explanation of the behavior.
Pure mathematics unassisted by facts doesn't tell us anything
about nature. It tells us only about abstract forms, which may or
may not correspond to something that actually occurs.
------------------------------------------------------------
Best to all,
Bill P.