Social Control, Close Shaves

[From Rick Marken (960821.0900)]

i.kurtzer (960813.1615)

the larger the population that is using the word (though potentially
...then the less it will change as the virtual control sytem is in
effect as function of the gain across all members.

Kent McClelland (960820.1145 CDT)

I thought you made a good point in observing that the more widely a
collectively controlled perception (like a word in a language, for
instance) is held, the more slowly it will be likely to change.

Change is evidently easiest when the group controlling the perception is
relatively small or at the margins of society, and deviant groups like
ethnic-minority youth often become the creative "avant garde" for new
styles and usages... Does that make sense to you?

It makes sense to me. But then S-R made sense to me until I modeled a
simple tracking task;-) Do any of your modelling efforts speak to the idea
that the number of systems involved in control of a variable influences the
ease with which the variable can be changed? It should be relatively easy
to test this. Basically, the idea is that the rate of change in a commonly
controlled variable should depend on the number of control systems
controlling that variable. It might also depend on the distribution of
gains of the systems involved and on the distribution of references for
the state of that variable.

The interesting question for me would be how some commonly controlled
variable changes as the reference for that variable changes in individuals
in the population. For example, I'd bet that about 80% or more of the
people in the US have a reference for keeping drugs illegal (or,
equivalently, for jailing people who consume drugs -- except alcohol and
tobacco, of course). Some of these people eventually realize that
controlling this perception has side effects that are worse (disturb
other variables) than personal drug use; but so far the people who are
controlling for drug decriminalization are no match for those who make up
the virtual control system that is aimed at maintaining the perception of
drug users in jail.

It would be interesting to see how temporal changes in the relative
distribution of gains and references in a population of control systems
controlling for a certain perception influences the behavior of that
perception over time. It sounds like PCT sociology to me. But, then, you
know that because you are the founder of that field!

Bruce Gregory (960820.1655 EDT) --

You [Martin] maintain that PCT involves an _implicit_ model of the
world. I am not sure that your interpretation is one that helps me
apply PCT

You are correct. This interpretation is simply incorrect. There is no
implicit or explicit world model (a model of the environmental physics
that relate system output to controlled input) in PCT.

so I reach for my razor....

And, as you end the refrain, thrust home;---)

Best

Rick

[Martin Taylor 960821 14:10]

Rick Marken (960821.0900)

Bruce Gregory (960820.1655 EDT) --

You [Martin] maintain that PCT involves an _implicit_ model of the
world. I am not sure that your interpretation is one that helps me
apply PCT

You are correct. This interpretation is simply incorrect. There is no
implicit or explicit world model (a model of the environmental physics
that relate system output to controlled input) in PCT.

Ah, so you are asserting that a control system will control its perceptual
variable equally well no matter what the form of the output function,
no matter what the gain, and no matter whether the output for positive
error is positive or negative.

So be it. But I'd love to see a demonstration.

And that's only for a single control loop. You are also asserting that a
complex hierarchy will control _all_ its variables (and the intrinsic
variables) equally well no matter what the weights (and the signs of the
weights) are that link the different elementary control systems, and no
matter which ECS is linke to which others...

A dramatic claim, which requires dramatic evidence.

Oh, it's certainly simpler than a model in which the linkages, weights,
functions, gains, and signs affect the performance of the system. So Occam's
razor says you must be right--if, and only if, your randomized system
performs as like the biological system as does one in which the weights,
functions, gains, etc have been tuned.

But you simply _know_ there's no implicit model. Do I hear an implied
"So there" at the end of "There is no implicit or explicit world model...?"

I agree with Bruce that knowing whether there is an explicit or implicit
world model may not help him to apply PCT in a particular situation, but
it sure is useful to keep in mind when you are thinking about reorganization
(learning).

I
agree that psychologically it makes perfect sense to use the
tools you are familiar with to solve the problems they can
solve.

That isn't what I said, though it is a sensible statement with which I also
agree.

But this seems to me to be a quite different issue from
the question of the minimal tool needed to get the job done.

To understand what I said, perhaps a little metaphor from a quite different
realm might be useful. Consider the difference between a marginal tax rate
and an overall tax rate. Two people under different tax regimes may both
pay (say) 17% on an income of $50,000, but person A may be in a country
where he pays 50c on the next dollar he earns, whereas person B pays 17c
on every dollar. Their marginal tax rates are very different, though their
overall tax rates are the same.

Now, Occam's razor deals in marginal rates, at least as I see it. Once one
has invested in a model that provides good returns, and no further investment
is required (i.e. no further assumptions or ad-hoc data) for it to account
for new data, the Occam's cost of using that model is zero. To discard that
model and use a new one that is seen as simpler by someone else (it cannot
be simpler on any _absolute_ scale) is to make a new investment. The cost
is considerable, and can be justified only if the new model provides
greater precision or a wider range of circumstances for which predictions
are made. It has to cover what the old model already did cover.

I know virtually
nothing about MCT, but it seems from what you and Hans say, it
involves an explicit model of the world that is absent from
PCT.

That is true as I understand MCT.

Further, MCT requires all the machinery of PCT.

That is false, as I understand MCT. PCT has a considerable machinery of
hierarchic linkages and perceptual functions, and these do not occur in
MCT. The explicit model takes their place. I could not say, from what I
know, which is simpler.

If PCT gets the job done, there is no _need_ to invoke MCT,

True. And if MCT gets the job done, there is no _need_ to invoke PCT. The
question is: can a situation be described in which either one of PCT and
MCT gets the job done whereas the other does not? So far, I have seen no
valid proposals for such a situation. What I have seen is speculations
from both sides that a hybrid system is probably what exists, with the "M"
more prominent at higher levels of a hierarchy, and the "P" more prominent
at lower levels.

What does Occam's razor say to the PCT-only position if a situation can be
found in which MCT "gets the job done" easily, but PCT does only with complex
extra assumptions? Or vice-versa?

It's all perception, and we have no other access to the "real" world. If
there is anything in the "real" world that we can call "absolute simplicity,"
how would you know if you were correctly perceiving it?

-----------------social control----------

Do any of your modelling efforts speak to the idea
that the number of systems involved in control of a variable influences the
ease with which the variable can be changed?

Perhaps not Kent's (about which I know very little), but there is a lot
of related work (from a source you seem to despise--the Santa Fe Institute).
Typically, in cross-linked system of many nodes (such as the interactions
among control-hierarchies (people) in a society) a phase change occurs when
the cross-couplings get strong enough or dense enough. In the social context,
that means when enough people can affect one another and care enough about
their effect. The system goes from a kind of fluidity into a sort of rigid
state where changes are more difficult. But dramatic changes can occur in
the "rigid" state, small events sometimes shifting the system into a quite
different but equally rigid state.

I was intending to bring this up as another way to address Hans's questions
about fighting, and about therapeutic crises. Both seem to me to be related
to this kind of coupling, but where the couplings involve conflicts within
as well as between hierarchies.

One reasonable book to read on this topic (easy and thought-provoking) is
Stuart Kauffmann's "At Home in the Universe." (I've probably made three
spelling errors in his name, for which I apologize). He does not directly
address interacting control systems, but I think very little of what he
says would be altered if he did.

Martin