# Useless; syllogisms; self-stoking cycles

[From Bill Powers (951019.0530 MDT)]

Hans Blom, 951019 --

My thesis is still that all behavior, all of it with no exceptions,
is produced for the purpose of controlling some perception of the
behaving organism.

What is the meaning of "thesis" in this context? Something yet to
be proven? An axiom, i.e. a ground assumption that you need to
construct further assumptions?

I said "My thesis is ..." deliberately, instead of saying "I believe
that...". I mean it in the the same way that Newton might have said "My
thesis is that each particle of matter in the universe attracts each
other particle, with no exception, in proportion to the product of their
masses and in inverse proportion to the square of the distance between
them." Having stated the thesis, one then proceeds to test its
implications by seeing how much of the behavior of natural systems can
be accounted for by assuming it. I suppose that is what you would call
an axiom, except that the intent is to use the axiom to explain natural
phenomena rather than to develop a self-consistent mathematical system.
So the test is not only one of internal consistency, but consistency
with data about physical systems. Failure of either kind of test would
invalidate the axiom. For example, Newton's axiom is consistent with a
deduction that the orbits of planets are elliptical, but would have been
invalidated if the observed orbits had not proven to be elliptical (or
at least conic sections).

Some time ago, I provided you with a demo in which behavior (all of
it with no exceptions) is produced for the purpose of controlling
some IMAGINED perception of (that is, prediction by) the behaving
organism.

There was, however, a means provided to assure that the imagined
perception (x) was kept in a match with the actual perception (y). The
whole point of your system is to assure that a perception, y, of the
actual plant output, x, remains equal (on the average) to the reference
level set by the signal r. So the overall purpose of your system is
still to control a perception of the external world.

Your model assumes that the perceptual signal y is in one-to-one linear
correspondence with some real variable x, give or take a noise term. In
PCT, it is assumed only that the perception is some function of a
collection of physical variables x1..xn, and that it is the value of the
function, not the individual values of the x's, that is under control.
This is the difference between a model of an organism and a model of an
engineered control system in which the nature of the controlled variable
is given by the customer's requirements and is known independently of
the controlling system.

Yes, I am aware that I am going up a level, and that this response
may appear to be oblique to the ongoing discussion...

Yes, and usefully so.

···

-----------------------------------------------------------------------
Rick Marken (951012.1550) --

Your explanation of why conventional observations are useless for a PCT
analysis was a great imnprovement over just saying that they are
useless.

I think the conventional researcher is in the same position as a
person observing the behavior of the subject in my "mind reading"
demo. The observer sees all kinds of interesting "behaviors"; the
subject moves the numbers in many interesting patterns; for
example, some numbers move in circular patterns, some in angular
patterns, some make broad, sweeping movements, others make more
constrained movements.

It's important to emphasize that these are REAL patterns; the numbers
really do move around this way; anyone looking at the screen would see
the same patterns. And the objective correlation of the hand movements
with these patterns is quite high by any accepted standard of
statistical significance. The longer you run the experiment, the more
certain it becomes that there is a relationship between the movements of
any one number and the handle movements.

If the movements of the numbers on the screen were taken as causal
variables and the hand movements as their effects, the number movements
would account for a respectable percentage of the variance of the hand
movements. A multivariate analysis would account for nearly all the
variance of hand movements (this would be worth actually carrying out,
just to emphasize the point). And the _least_ correlation would be found
between the hand movements and the number that is actually under
control.

meant by irrelevant side-effects of behavior. It also shows exactly what
is wrong with the standard approach to causal analysis. Even though the
various side-effects of handle movements are clearly visible, and even
though their relationship to the handle movements can be established
beyond reasonable doubt by objective means, they are still irrelevant to
what the person in the experiment is really doing.
------------------------------------
There's another facet of this problem, of a different nature, that Bruce
Abbott and others who might argue on his side should ponder. It is the
subjective component that all experimenters bring to observations.

Suppose an observer chanced across the experiment in which chicks were
separated from the brood and the mother and were placed in chambers with
and without mirrors. This observer is not informed of the purpose of the
experiment and does not know how the experimenters are interpreting
their own observations. He says:

Ah, listen to that angry peeping! Look at that aggressive running
around, that hostile pecking at the walls of the cage! Observe how
isolation in the plain room elevates the chicks' rage, how the
implacable threatening peeps are pumped out endlessly in a frenzy
of hatred! And observe how cowardly these threats are, for as soon
as images of other chicks are provided in the mirrored chamber, the
threats become muted and the belligerent behavior is suppressed.
The counterthreat from the other chicks that are apparently present
is sufficient to overcome the hostility and force the formerly
warlike behavior of the chick into socially-controlled channels.
The parallels with human behavior are unmistakable: just as the
belligerent cries of the chick are suppressed by socializing
forces, so are the angry cries and instinctively belligerant
actions of the child suppressed by the presence of counterforces,
and by the superior size and strength of the parents.

Once an observer gets a story started in his mind, the world obediently
offers supporting evidence that makes the story continually more
convincing. Every little detail fits; every new detail that is noticed
leads to an AHA and a further building of confidence that one is right.
Robert Pirsig used the happy phrase "self-stoking cycle" to speak of
such positive-feedback phenomena, specifically relating them to
Phaedrus' obsession with seeing evidence about "quality" in everything.

It is this tendency toward self-stoking cycles that science tries to get
around by its insistence on testing, testing, and testing. And the
testing it insists upon is not just seeking further corroborative
evidence that one's interpretation is right; it is an insistence on
looking for ways in which the interpretation might be _wrong_. Anyone
who has been absorbed in an obsession knows how easy it is to see
corroboration of an idea one is convinced is right. The conviction
itself is the driving force, the interpretation that turns neutral
events into overwhelming evidence. The point of science is to find ways
of investigating nature that make conviction irrelevant, that are
capable of leading to unavoidable conclusions even against one's
previous convictions.

My point about "separation anxiety" is not that it is based on poor
observation, but that by its very nature it is a subjective
interpretation by an observer. It is a story that, once accepted, shapes
the way every succeeding detail is interpreted. It is part of a self-
stoking cycle. I don't need to go into undiplomatic details about all
the self-stoking cycles we can see in human behavior, particularly in
their intellectual behaviour. I only want to point out that there is a
problem here and that the only antidote is skepticism and a demand for
demonstrations and proofs. Since we are all prone to becoming obsessed
with self-stoking interpretations, our only defense is to resist being
carried away by plausibility, and to look to nature for an anchor. It is
only when _no other interpretation is possible_ that we can feel safe in
accepting any proposition about reality. And even then, experience will
teach us that we have not actually thought of all possible
interpretations.
-----------------------------------------------------------------------
Best to all,

Bill P.