[From Bill Powers (2004.06.26.0516 MDT)]
Bruce Nevin (2004.06.25 21:55)--
This is what I understand from this. When the observed organism is known to
be constructed differently from the observer, then only an analogy can be
known between the perception that the observer identifies as the CV and the
perception that the observed organism is controlling as the CV. However,
when the observed organism is known (or assumed) to be constructed like the
observer, as a fellow human being is, then for the observer who has
identified a CV (eliminating every alternative that she can imagine), an
equivalence can be known: the observer knows that both the observer and the
observed person are controlling the "same perception".
The similarity of construction of different people is not as firmly
established in my mind as it seems to be in yours.
I put "same perception" in scare quotes because in the PCT definition a
perception is a neural signal, and it is obviously not the same neural
signal. "The same perception" means a signal at the corresponding place in
two control systems that are constructed alike.
A numeric value in a successful simulation is also claimed to be "the same"
as a perceptual signal in the organism that the simulation models because
PCT claims that the simulation and the organism are structured alike, and
that the functions, connections, and signals in one correspond to those in
the other.
The reason these models have adjustable parameters is that people are NOT
constructed identically. If they were, the same model with the same
parameters would describe and predict any person's performance equally
well. That is not the case.
My claim is that the actual, functional, organization of the nervous system
is determined primarily by reorganization, with the inherited structures
only providing the raw material from which various classes of functions can
be constructed. There is a structural predisposition toward forming control
systems, I assume, because it seems that evolutionarily that type of system
has a strong advantage over any other type. But there is no predisposition
toward forming any _particular examples_ of control systems -- perceptual
input functions or output functions -- at least at the higher levels. While
the same physical structures may be present in all people, the synaptic
weightings and connections in the nervous system are highly plastic from
top to bottom, and it is the synaptic weightings and connections that
determine what kind of organization a given neural net will become. The
appropriate analogy is to a computer and its programs (though the brain is
not a digital computer and we're not talking about computer programs). The
computer provides the basic functions that can be performed, but it is the
program that determines what kind of machine the computer becomes, a game,
a simulation, a predictor of stock markets, a process controller..
This plasticity is actually essential for your concept of language and
culture. If people were constructed identically, functionally as well as
physically, there could be only one language if any, only one culture if
any. You and I could have no disagreements, either actual or apparent. We
would be copies of the same person with the same organization.
The ability of a model to give a plausible and predictive explanation for
behavior does not mean that the model is structured like the person or that
one person is structured like another. There are are multiple alternative
ways for the nervous system to carry out any given function, even such a
simple function as comparison of two signals to produce an error signal.
Comparators can be, and are, built into perceptual input functions
(temperature control in the body); they can be, and are, built into output
functions (in the brain stem); they can be composed of multiple neurons
that can handle both positive and negative signals, or single neurons that
can detect only positive or only negative errors.
But worse that that, they can consist of multiple stages of processing.
Instead of e = r - p being computed by a single neuron, the net might
compute e = r + 17*p - 8*r + 4*(p+2r) - 22*p and we would never know the
difference, because the processes that do those computations produce
exactly the same error signal from r and p (with a longer delay, but many
other computations could produce the same delay).
And that's far from the end of it. What we know about the brain is itself
the content of our own brains. It has long been suspected, and some think
proven, that the brain cannot even in principle comprehend its own
structure. Certainly the reality of our brains has far more degrees of
freedom than our intellectual representations of a brain have. We perceive
through the filter of our input functions, which simplify and idealize
experience, averaging, smoothing, ignoring and discarding differences and
variations at every level.
Probably the best analogy to our situation is the one in the little book
"Flatland, by A. Square." Here a three-dimensional being encounters a
two-dimensional world whose inhabitants are triangles, squares, pentagons,
and so on, and who live in flat outlines they call homes. They are
astonished when their visitor (a sphere, if I remember correctly) can
change size before their eyes (by rising and falling above and below the
plane of Flatland), and can even disappear from inside a house and appear
outside it withough passing through a door or window. The analogy was
intended to illustrate how an extra dimension would be quite invisible to
us, yet could explain certain strange phenomena like the curvature of space.
There are probably vast numbers of ways in which the world can change to
which we are utterly blind. What we see as the same is not necessarily the
same at all. And this means that each person could experience a unique
version of reality, and be completely unable to detect the fact that his
world was not the same as the worlds of others.
This, of course, is a worst-case picture. Absence of proof is not proof of
absence, so it remains possible that we are identical in most respects and
comprehend all that matters in reality -- but that picture is also an
extremum.
I have looked for many years for a way to find the truth of this matter
without just giving up and saying, :"O.K., here is what I choose to
believe." I have looked for ways of showing that with sufficient
triangulation among different perceptions, it would be possible to find a
unique solution to the equations of reality, so that only one form of
reality would remain possible. Even considering the triangulations made
possible by the existence of different independent individual observers, I
have not found a way to transcend my own perceptions and see what lies
beyond them, even given all the aids at my command (which are not nearly
enough for the job, evidently).
What I have been able to do, as you saw at last year's CSG meeting, is to
show that an intraordinal set of control systems can perceive and control
variables which are arbitrary random functions of a common set of
environmental variables, creating controlled variables out of nothing, and
yet being able to act on the environment in such a way as to achieve
independent control of each variable. That is a worst case, too, but it is
a most interesting one. My mind balks at taking the next step, which is to
set up two systems in a common environment, and to introduce
reeorganization so that they will adapt to each other. Will they not only
create organized perceptual worlds for themselves to control, but adapt to
the other's modes of control?
I don't accept these findings as a conclusion, only as evidence of a
possbility. But I am still looking for some way to show that there is an
alternative to them.
I say that my mind balks and that is what I mean. I come up against some
kind of intellectual limit. I feel too disorganized to proceed -- what is
the first step? I have often felt this way in the past, sometimes for
years, but eventually the next step reveals itself and I can take it. But
not yet.
Limiting ourselves to fellow humans -- the scope of discussion is social
realities, after all -- when the observer has identified a controlled
variable, the observer and the observed comprise a single dyadic system
linked through the environment variables (EV):
output output
________ ________
/ \ / \
Observer's \ / Other's
[CV] Control EV Control [CV]
Hierarchy / \ Hierarchy
\_________/ \__________/
input input
The first thing you have to do is look long and hard at this diagram you
have drawn, and see it as your own creation, in your own perceptions. You
have set the diagram up so it contains (as you describe it) two identical
control hierarchies. Of course if they are identical, the CV's you indicate
are identical. But isn't that the conclusion the truth of which you are
investigating? It seems to me that by drawing the diagram this way, you are
simply expressing the conclusion that you already believe, not providing a
basis from which it can be deduced, discovered, or defended.
A mathematician would set up the diagram with enough free variables and
parameters so that in principle any arrangement would be possible on the
two sides of EV. Then he would see if it is possible to show that the
variables and the parameters would have to be equivalent or even identical
on both sides if the respective parties thought they had reached agreement
on the CVs. A proof would have to include the means by which the parties
communicate the nature of the CVs to each other, for only in that process
would there be any hope at all of establishing the identity. And the threat
of infinite regress would always lurk in the background, because how does
one establish that the communication itself is understood the same way by
both sides? Through more communication, raising the same question again?
The parts of the two control systems are in correspondence, and in
particular the [CV] perception that one constructs of the environment
variables is "the same" as the [CV] perception that the other constructs of
the environment variables. It follows then that by virtue of identifying a
CV controlled by a fellow human the observer knows that she is perceiving
"the same perception" as the observed person is perceiving. This knowledge
is a perception of course, constructed by means of perceptions that we call
PCT. What else could knowledge be, but perceptions.
This is completely circular, is it not? You're saying that becauee the two
control systems are in correspondence, they are in correspondence. But how
do you establish that they are in correspondence in the first place? You
can't do that by drawing a diagram that indicates the arrangement that
would exist if the two control systems were in correspondence. What if tney
are not? How would you draw that?
I think that further deductions should be postponed until this basic
problem of circularity, of begging the queation, has been settled. At this
moment, it looks to me as if you're simply stating your faith in, or your
preference for, one particular conclusion that is to be drawn. I don't see
where this conclusion comes from, other than the fact of being stated and
illustrated by a drawing.
Best,
Bill P.