reliability, philosophy, society; T-illusion

[From Bill Powers (950219.1017)] --

Thought I was going to have an early night after two days of beautiful
Utah, but I made the mistake of reading the mail before going to bed.


Bruce Buchanan (950218.14:50 EST)--

Point of interest from various posts: you keep saying "controlling
variable." Could you explain the relationship between what you call a
controlling variable and what we call a controlled variable?
     In practice, for instance, one might question whether it is worth
     pursuing dialogue with someone who insists that reliability and
     responsibility i.e. consistency and dependability of language and
     action, are not only unimportant but so indeterminate as to be
     meaningless (which is the impression I have got from this

I will try to correct that impression. I am not arguing against anyone's
recommendations for how we should get along with each other; only
pointing out that if you have to use persuasion or logic to prove that
something is a good idea, it's probably not inherent in the system. Some
groups of people value and encourage spontaneity and creativity; others
demand predictability and conformity. Many social systems have been
proposed and tried involving different kinds and degrees of assignment
of reliability and consistency, with varying outcomes. If you have a
good one to propose, fine -- but it's not a scientific or philosophical
truth, it's a proposal. Do you propose, for example, that we would
communicate better if people used words in consistent and reliable way?
I would probably think that a good idea.

     Here we are getting to what I think is a basic source of confusion,
     perhaps an error from a philosophical perspective (and part of the
     error is to rule out consideration from a more comprehensive

The philosophers I've read don't agree with each other well enough to
establish something we could call a "philosophical perspective." While
philosophers have raised questions that I think have been useful and
sometimes, to me, inspiring, I haven't been particularly satisfied with
their answers. It is sometimes difficult to know whether a philosopher
is arguing from a more comprehensive perspective, or just a vaguer one.

     The point is that everything cannot be reduced to PCT. Other
     systems must be understood on their own levels in terms of their
     own entities and phenomena, and according to their own observed

I quite agree. I don't use PCT in understanding or designing electronic
circuits unless they happen to be control systems. I don't think PCT
applies to the interactions among independent control systems: in
particular, to social interactions.

In short, I am quite capable of looking at non-PCT systems in their own
terms. This, however, does not mean that I have to buy into whatever
anyone else says about them. It doesn't mean that I suddenly begin to
accept as facts statements to which I can think of 10 counterexamples in
half as many minutes. I can put PCT aside and play the other guy's game.
But I don't leave my critical faculties behind when I do so.

     ... as Michael Polanyi pointed out, while emergent properties
     depend upon the more basic properties of the systems from which
     they emerge, they have new properties at their own levels, with new
     relationships and new dimensions, as I think is recognized by PCT.

Did he "point it out" as something that was obviously there and needed
only to be noticed, or did he "say" it? I opt for "say." I go even
farther than Polanyi, since I say that higher-level systems (in human
organization) are physically distinct from lower-level ones, not merely
broader ways of talking about the same collection of elements.

     For example, binocular vision depends upon the integrity of the
     visual processes in two separate eyes, but by reaggregation of
     selected features brings out depth perception, which is not,
     however, in turn reducible to any two dimensional fields. No more
     can social and economic phenomena be reduced to adequate
     explanation within PCT, even if the principles involved in PCT and
     physiology, etc., do underlie some aspects of the higher level
     descriptions of relationships and events, i.e. phenomena.

Yes, I agree. Social phenomena are emergent. They follow different
rules. But they do not follow any ONE set of rules that anyone has so
far discovered. If two people get into a positive-feedback relationship,
certain predictions can be made about the consequences. But there is no
social rule that says any two people will get into such a relationship.
If you see the symptoms of positive feedback, you can say after the fact
that positive feedback must have happened. But you can't predict that it
WILL happen, because that involves predicting which people with what
internal characteristics are going to bump up against each other. To say
anything interesting about emergent phenomena, you have to understand
what they are emerging FROM.

The fact that there are tennis balls and that they
show a certain behavior in a gravitational field, in an atmosphere,
and in relation to a tennis racket is merely an accident, an
invention, which might not exist in a universe with a different
history ...

     Well, if conditions in the world were different, we would not have
     the problems and the questions we do. But to follow this line of
     reasoning leads nowhere, and, from a philosophical point of view,
     it raise pointless questions as to objectives and methodology.

So much for the superior comprehensiveness of the philosophical
perspective. In the physical sciences, it makes a great deal of
difference whether we are talking about a matter of happenstance history
or fundamental principles that would remain true regardless of history.
The present orbits of the planets are entirely a matter of the history
of the solar system, and in themselves are of no scientific importance.
The fact that they are all ellipses, and that ALL orbits of ALL planets,
whatever solar system we are talking about, must be ellipses (to an
impressive number of decimal places) IS important from the scientific

Of course it is of great practical importance that the particular orbits
in our actual solar system be known, at least to those who predict the
appearance of the skies or plan expeditions to other bodies. However, I
don't think that NASA engineers (or philosophers) would be insulted to
be told that the detailed celestial mechanics of our solar system are,
historically, accidents. Confusing historical accident with fundamental
laws led to the proposal of Bode's law, a meaningless generalization
from one case.

      More specifically, failure to accept that we live in this
     universe, on this earth, with the history we were given by the
     accident of birth in a particular culture, is to reject the
     possibility of an inquiry into many complex problems, the results
     of which might well be justified in relative terms as the best we
     can do.

No, it's not to reject that possibility. People can work on these
complex problems all they like, with my blessing. But this earth and
this history we were given is a sample of one, and we shouldn't
overextend our generalizations about them without at least conjecturing
about what else might just as well have been. When we can see that a
different outcome might just as well have occurred, we will be less
eager to propose that the way things are is the result of some
fundamental principle of social life. There may be fundamental
principles, but you can't get them from a sample of one.

If there are social
rules, for example, they exist only because people are capable of
perceiving rules and making behavior conform to them. This property is
totally independent of WHAT rules are put into effect.

     As written this still seems to be beating a straw man. If what is
     meant is that *social rules depend upon general human capacities to
     perceive and conform*, it would be hard to disagree. But what is
     implied also suggests that there are no other reasons can exist
     which may make rules necessary or desirable.

That isn't the question I'm asking. I'm asking what makes rules
POSSIBLE. That's what I mean by asking "What is a rule, that we can
follow it?" I don't mean what particular rule, I mean any rule at all.
It's thinking in terms of "any rule at all" that makes one realize that
particular rules aren't built in to human nature or human relations; any
rules are possible. Particular rules are made up by people in answer to
practical problems. When you realize that there is a deeper level at
which to consider rule-making and rule-following, the question "what
rules do people follow" begins to look rather empty.

     This leaves dangling at least two important questions which are the
     reason for the discussion in the first place i.e. (1) what might
     be the logic of social structure and relationships which requires
     some sort of rules, whatever they may be, independently of who
     might occupy the social roles in question; and (2) what might be
     the functions and activities which account for such human
     capacities in terms of their evolution and development i.e. what
     functions those capacities serve.

But first, don't we have to establish what those capacities ARE?

I see a brain with
the capacity (I conjecture) to perceive and control in terms of
intensities, sensations, configurations, transitions, events,
relationships, categories, sequences, rule-governed processes,
principles, and system concepts....

     Well, on the one hand these comments seem almost modest e.g. ("I
     conjecture") and on the other hand they strike me, correctly or
     not, as an imperialistic hard-science approach to all questions of
     human nature.

They represent my best attempt to generate a scientific approach to
questions of human nature. I described what I consider to be a
scientific approach in a post a few days ago; if you've lost it I'll
send you a copy. I don't really distinguish among science, hard science,
and trying honestly to understand how things work. It's all one thing to
me: trying to understand. The "imperialist" label is rather out in left
field; if you want to discuss how we can come to understand how things
work, fine. But if there are other approaches I refuse to accept for
myself, and point out the flaws in, does this make me an imperialist? I
happen to think that some theories are better than others, and some ways
of reaching understanding are more effective than others. So what? So do

     Yet the set of specific secondary or emergent attributes which
     could arise, whether food habits, social rules, tool-making,
     language habits, and all the features which make a civilization
     operate, could not have been just "any set(s)". They had to be
     habits and rules, etc. which fit the conditions and enabled mutual
     survival, and these constraints are more often than not quite

They may seem rigorous if you don't look at the details. You can make
any two things look the same if you describe them at a sufficiently high
level of abstraction. There is not just one set of habits and rules that
fit conditions and enable mutual survival. There is an infinity of
alternatives, and quite a large number of them have been tried. There is
not just one "civilization" but many, reflecting only a small sample of
the ways human beings can promote their own survival and that of others.
I hope you don't mistake this for a PCT idea -- I don't think you need
PCT to see that this is true.

     Yet I would argue that there is a case to be made, not incomptible
     with PCT although not part of its theory, for considering that
     ultimate meanings for human beings are related not only to physical
     survival but to the individual's perception of membership, of
     acceptance within a group and in the universe somehow.

Well, this leaves me with no understanding at all of what you mean by
the word "meanings." Please explain.

     We perhaps agree that all observations and discussions must begin
     from where we are. And we are individuals enmeshed in social
     structures in a physical world. We do not know enough to explain
     all this in terms of a single theory which reduces everything to to
     terms of perception alone. As I think I have suggested, I do not
     believe that such a reduction is possible in principle.

At the risk of bring up epistemology for the nth time, can you give an
instance of ANYTHING you know about that isn't a perception in the PCT
sense? If you can think of one, I'll try to put it into the model.
Rick Marken (950218.1530)--
Bruce Abbott (950216.1040 EST)--

Good work on the T-illusion. Interesting how quantifying it shows us
that there isn't "a" T-illusion; the amount of it differs considerably
among people. I found the same thing in working with Pat Alfano (student
of Dick Robertson) on a moving-dot illusion (essentially the same method
Martin Taylor has described using years ago). Pat was originally
interested in it because she had severe motion-sickness. After we had
tried out the method for measuring the motion aftereffect enough dozens
of times, in the course of working up a systematic experimental program,
both of us ceased to have any motion illusion at all. I haven't tried it
lately to see if it has come back, and quite possibly the immunity was
p[eculiar to the experimental setup, but this is another case where
measuring "the" illusion turned out, inconveniently, to be impossible.

To explore this would really need some protracted experimentation with a
number of subjects. Bruce, wouldn't this be a nice one to hand off to a

It's interesting (and reassuring) to see that regressing vertical on
horizontal yields the same results as running the model. In effect, both
procedures are doing the same thing: determining the reference condition
of the relationship between vertical and horizontal. In the case of the
regression analysis, the assumption is that the observed ratio is what
the subject intended; in the control model the intention is made

I wonder if you wrote the program in your post the way it was run.

p = vert/horiz;
e = p - ref; { ref = 1.0 initially }
h = h - k * (e*horiz);

As presented, p depends only on vert, because horiz is a constant! You
seem to have left out the dependence of horiz on handle. And I don't
know why you multiplied e by horiz.

The data recorded are

      horiz^[i] := 2 * (XLength + oldt);
      vert^[i] := YLength + oldc;

Since the horizontal length is independent of the model we can leave it
as it is. But if the cursor position is numerically equal to the handle
position (the code says it is) we have to write the model this way:

p = (Ylength + c)/horiz

e = p - r

h = h - k*e

c = h;

Where Ylength = Maxy div 3 or 479/3 or 159.

This should improve the correlation of the model's handle position with
the real handle position.

I'll try this tomorrow and see what I get. But of course the amount of
the illusion will be mine own.
I think I've finally got my mind as tired as my body.

Best to all
Bill P.