[From Bill Powers (930223.0800 MST)]
Martin Taylor (930222.1500) --
All that we have available is a set of scalar signals that are
no different from each other whether one represents "cup" "in"
"democracy" or the intensity of a patch of light. Only an
observer who sees both the signal and the referent can say that
the signal represents the complex in the outer world
Aren't you forgetting that the observer, too, can experience only
the cup etc. signal? There is no observer who can say what any
signal represents in the outer world. All the observer can say is
that when the other's brain contains a certain signal, the
observer is having a certain experience in his own brain.
If there is some method the observer can use to find a referent
for a signal observed (by electronic or magical means) inside
another system, then that method is also available to the person,
organized like the observer, in whom that signal exists. As far
as I can see, the only possible method is to look at other
perceptual signals, that being the only kind of information that
is available. We explain or identify one perception in terms of
others. The observer explains the signal in the other by
employing a world model, but that world model's form and behavior
are known in the form of perceptual signals -- as everything that
is known is known.
I'm trying to be strict in adhering to PCT principles.
So am I. And it causes problems for me, because I can't explain
why a cup looks to me as it does. It doesn't look like a neural
signal. The only solution I have ever found is to say "Well, this
IS how neural signals look, to an aware observer." I feel as if I
have a beautiful logical structure here, with one essential
proposition missing somewhere in the middle of it.
The NATURE of the REFERENT of the perceptual signal is NOT
perceived, at least not by THAT perceptual signal.
Right, and this is true of the referent of EVERY neural signal,
isn't it?
The problem in this whole discussion is that you're describing
world models in terms of what they must do, which is to model
the world, but without any indication of whether a feasible
method for doing this exists, even in principle.
Well, I take it that when control engineers show graphs of the
convergence of their algorithms for doing this, they are not
simply lying.
No, they aren't lying about that. But they're cheating, because
before any such algorithm can work, the basic structure of the
internal world-model must be supplied by the engineer in a form
that matches the elements of the external world, with only a few
free parameters left to be adjusted. The mechanism for producing
and adapting an FIR is supplied, and it is determined beforehand
that an FIR would be a suitable kind of model for the environment
that actually exists. This is not, in general, a suitable model
for an environment perceived at any general level. For example,
it would be useless (as would be a modeled mass or mass-spring-
dashpot) for modeling the control of an event like giving a cup
of tea a quick stir. Engineers think in terms of just a few
restricted kinds of control, in environments having simple and
familiar characteristics. This is fine for engineering, but
completely inadequate for modeling the whole range of human
control behaviors.
Your question leads me to wonder why it is that we perceive
acceleration readily and can control it, but can do so much
less readily if at all for the third derivative of position.
Acceleration has a sensory concomitant which we have sensors to
detect: force. By controlling force we can control acceleration.
In driving a car, sideward forces are used as a sign of lateral
acceleration; driving in a straight line is aided (and control is
speeded) by controlling for zero lateral force, and using visual
control only for lateral position and velocity. This is called
seat-of-the-pants control, an expression which shows acute
observation on the part of car drivers (and airplane pilots).
Acceleration is also perceivable at a high level, although much
less precisely: rate of change of a velocity. You can see that
another car is decelerating, but not with high resolution.
To the observer, there may be highly significant properties of
some relationships among the observed values of the vector
elements. This does not give those relationships significance
to the perceiver within whom the vector exists.
The question is, to whom or in what sense are these observed
relationships "highly significant?" If the behavior of higher
systems is strictly a matter of controlling perceptions derived
from sets of lower-level perceptions, then there can be
behavioral significance in the inherent relationships only if
those relationships are represented as specific signals. This
applies to the observer as well: the observer can know of them
only if they give rise to a specific neural signal in the
observer, the signal the observer sees as the "pattern-ness" of
the selected set of signals. What other kind of significance
could there be?
I would have thought that the co-moving set of functions
identified by "the whole organism" was a far from arbitrary
collection, having been so organized over evolutionary time
because they worked better together than separately.
If all I can say about a set of functions is that they are "co-
moving" that doesn't give me much sense of a system. It may not
be an arbitrary collection from some hypothetical objective point
of view, but it's arbitrary for me until I can see the
organization in it. To speak of "the whole organism" just means,
to me, to speak of a lump of matter. Not very helpful to the
understanding.
Why did life evolve with separately moving entities of very
similar structures, rather than as one planet- enveloping mass?
Presumably small things were better able to control their
perceptions than big things would have been.
I don't think of evolution as selecting among a set of possible
designs, and especially not doing so for rational reasons. The
only way a planet-wide mass could have been selected against
would be for it to have existed at one time, which it never did
as far as I know. Evolution can deal only with designs actually
implemented.
Why so many different kinds, all more or less identically
repeated millions of times? Presumably each of these co-located
collections of functions worked as a "system" pretty well in
the environment in which members of the class tended to find
themselves.
There are so many kinds because reorganization and mutation are
not systematic in any terms we can recognize (i.e., they are
"random"). They survived because they were capable of controlling
the effects of the environment on themselves that were important
to survival. They did not survive simply because they were
"systems." Even a nonviable organism is a "system," according to
the permissive usages of the term that prevail.
I think of the control hierarchy as a control "system,"
too.
I don't, although I do think of any one control system at one
level in the hierarchy as a control "system." The whole hierarchy
is a system, I can agree, but it is not a control system. It does
not have an identifiable input function, comparator, and output
function controlling a controlled variable at a specific
reference level. To borrow a term from Minsky, it's more of a
community. Or you could call the whole hierarchy the output
function of the reorganizing system (my version).
I think you are using "DNA" as a metaphor for a whole complex
of interacting chemicals in a structured environment. I think
of it as a double helix made of interlinked sequences of four
complementary amino acids. It doesn't act. Its structure is a
store.
It's possible that you are right. What's missing from that view,
however, is that which reads the store. There must be a mechanism
that scans the stored code and uses it as reference signals for
processes that manufacture enzymes and proteins.
It's also possible that the required control systems are actually
in the DNA. There are long stretches of DNA that seem not to be
involved in any genes; I've been reading recently that people are
starting to suspect that these "silent" stretches have regulatory
functions. All that's needed for a chemical control system is
something with catalytic activity to provide amplification and
comparison. Some combinations of the amino acids could have such
activity.
One factor supports your idea rather strongly: there is more that
is passed along from parent to offspring than just DNA. In
reproduction by division, all of the soup in the cell is divided
between the offspring, including complex structures like
mitochondria and cell membranes. This is also true in sexual
reproduction. If the tape player is in the soup rather than the
DNA, it would still be passed along the generations (including
the instructions for self-repair). In this case, perhaps the
"silent" stretches of DNA carry the states of variables involved
in control, like perceptual signals, error signals, and output
signals. This whole arrangement would then permit continuous
operation of control systems right across the boundaries
separating generations. Of course this would apply only to the
cells actually involved in reproduction, but there are rather
long periods, even in sexual reproduction, where those cells
don't yet exist, and the only operative control systems are in
the cells that do exist.
···
---------------------------------------------------------------
Bill Leach (940222.2030) --
Before trying to guess what behavioral side-effects depression
might produce, it would be a good idea to try to say what
depression is. What is it that depressed people do or don't do,
that normal people don't do, or do? Beneath the symptoms seen and
felt as depression and described with that word, there must be
something out of the ordinary in the functioning of a person's
control systems. If we could pin down the meaning of depression
to specific changes in functioning, we might be able to guess
what control processes have gone wrong -- perception, comparison,
action, memory, reorganization? We might be able to guess which
levels of control are still functioning normally and which are
not. By applying the same sorts of experimental procedures we use
in tracking experiments, we might be able to measure people's
ability to control variables at different perceptual levels, and
see if the parameters are abnormal in comparison with those found
in a baseline population. The test for the controlled variable
can be used to see what the people are perceiving, and what
reference levels exist for those perceptions. The method of
levels, used in talking therapy, can help us to explore the
structure of the people's goals.
In this way, rather than simply treating symptoms with random
methods until something works to remove the symptoms (the present
method, for the most part), we could do a systematic exploration
of a person's structure of control systems and narrow the problem
down to specific malfunctions. We would not have to rely on
"typical" symptoms or "indicated" treatments. Each person could
be investigated as a unique case and the nature of the problem
identified in some way other than general similarity to other
people with problems showing the same superficial appearances.
This approach to mental and emotional problems would completely
do away with most present approaches. It might even work, which
is more than can be said with confidence of most approaches.
RE: clinically depressed people give more accurate descriptions
of the world than normal people: Maybe the message is that normal
people just don't understand the situation. If they did, they
would be depressed, too.
Even as I write this, I am beginning to question my definition
of "depressed".
That's a good start.
----------------------------------------------------------------
Best to all,
Bill P.
