<Martin Taylor 940222 15:00>
Bill Powers (940222.0830)
I think something needs to be made more explicit here. I assume
that the world we experience consists entirely of scalar
perceptual signals, some in control systems. That means not just
scalar signals representing positions and velocities, but scalar
signals representing cups and cupboards and "in" and numbers and
conclusions and logic and everything else. We experience many of
these signals awaredly and simultaneously, which is what creates
the complex world of experience. But any one dimension of that
world, at any level whatsoever, is actually a simple single
scalar neural signal that can vary only in magnitude.
Exactly the position from which I have been working. 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. Otherwise, all the hierarchy has is the ways in which its actions
affect its perceptual signals--its untyped scalar variables.
And if the "world models" in the various ECSs happen to
correspond (for the outside observer) to changes in cups and
cupboards, so it may be. The ECSs in question don't know.If the ECSs don't know, then nobody knows. What else is there to
do the knowing? Are you proposing a whole new human system that
doesn't employ ECSs or at least their perceptual components?
Is it clearer now? I'm trying to be strict in adhering to PCT principles.
To perceive something is to have some value for an untyped perceptual signal.
The NATURE of the REFERENT of the perceptual signal is NOT perceived,
at least not by THAT perceptual signal.
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. I haven't
done it myself, so I have to take the word of people who publish the
results of experiments in doing it that they have actually done what they
claim. I'm not, perhaps, as cynical as you.
I have to think more about what you say about the limitations of using
a simple FIR as the model. I agree that the problem of double integration
provides a nasty issue, and my approach may be too simplistic. It isn't
the approach used in the engineering systems, but it is one that is simple
to implement in neural systems, so might be feasible at least for some
situations, if not all.
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. If it were simply a question of cascading
perceptual differentiators, there should be no problem other than noise
in the system.
All this relates to a bone of contention that was unearthed
earlier, namely the concept of vector perceptions.
I'm not sure where there is or was a bone of contention about vector
perceptions. I do not accept what you seem to impute to me:
According to the vector perception hypothesis as I understand it,
the mere existence of the relationship is enough to give it
behavioral significance, even if there is no perceptual function
in the system that can receive the values of the two variables
and output a signal indicating the value of the regular function
of them.
This is NOT my view and never has been. The significance of the
vector perception, as I have claimed right from the beginning, lies
entirely within the observer of the behaviour, who does have a single
perceptual function (or set of them) that provide significance to the
relations among the different elements of the vector. 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.
Other than that imputation, I agree with what you say about vector
perceptions. To me, what is interesting about them is that relationships
among the elements may be enforced by facts in the world (unknown to
the vector perceiver), and it is these relationships that provide the
potential for the perceiver to develop higher-order control systems that
work and are useful.
Also, I agree with what you say about all-enveloping systems. But again,
there is a matter of emphasis. Some people are interested in the microscale
description of the system elements, whereas others accept that the smaller
elements perform their function in some appropriate manner, and are
interested in what happens when those smaller units are put together in
various manners.
So to me, "system"
is not connected to physical subdivisions of larger or smaller
size, but to a recognizeable organization of functions.
Reasonable.
I do not
see a recognizeable organization of functions in "the whole
organism,"
Surprising.
or in general in arbitrary collections of functions
with boundaries arbitrarily drawn.
Reasonable. 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. 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. 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.
So I think of the control system as the input function, the
comparator, and the output function, with the external link being
part of that which is controlled (although control is possible
only for a certain range of external organizations).
Yes, I think of the elementary control unit (a phrase I now prefer to
ECS--it was coined in error by one of the participants in my PCT seminar)
as just that. I think of the control hierarchy as a control "system,"
too. And it is only the hierarchy within one organism that deserves this
title. Larger (such as social) structures may be "systems" but they are
not "control systems" for all the reasons that have often been expressed
on CSG-L. The interesting features of such social structures derive from
the ways "control systems" interact when they have no common high-level
reference signals. The "reorganizing system," if one exists, deserves to
be called a "system," because of its coherence with respect to its function
in respect of the hierarchy within one organism. Together, the hierarchy
and the reorganizing system might well be called a "learning system."
Together, they form a complex that learns to control a considerable set
of perceptions.
That probably didn't merit a whole long paragraph.
Well, it induced another. What is "to deserve?"
Well, the model I have in mind shows DNA as a collection of
control systems which control for the states of variables
important to accuracy of replication, by means of varying the
mutation rate.
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.
Martin