Anticipation, assumptions

[From Rick Marken (970326.0840 PST)]

Mary Powers (970325) --

Anticipation, expectations, and planning are products of the
program or strategy level in the imagination mode.

Imagination at this level is called thinking. The idea is to
cook up reference signals for perceptions of the anticipated,
expected, or planned, and either try to bring them about or sit
back and hope they happen.

This level is so important to some people that they think it's
the whole ball game. It isn't.

So succint. So clear. So precise. That Bill Powers is one lucky guy.

This is also an excellent reply to that master of imagination, Hans
Blom. I think he asked a rhetorical question about anticipation and
planning a few days (weeks?) ago. I think he expected the answer to
be "model based control". In fact, the answer is "control of
imagination" which (like Hans' model-based "control" model) is not
the kind of control that will get you want in the real world.

James Chriss (970327) --

You suggested that I go see how your model works rather than
worrying about underlying assumptions of your model or PCT
theory. But this is the sort of instrumentalist bias which I
cannot abide by.

Well, there's not much I can do about that. I guess you're going
to have some problems abiding me. You won't be alone;-) (That
reminds me; where IS Bruce Abbott these days;-))

When you make a model that is supposed to represent some aspect
of the world, the making of the model itself in no way assures
that the model is somehow isomorphic with the empirical social

I agree with this completely!

In order to judge the extent to which some working model can be
said to represent some aspect of the empirical social world,
your axiological, ontological, and epistemological assumptions
about the nature of values, knowledge, and reality have to be
analyzed to see what you're assuming about the way the world works.

I agree with this somewhat, too! You just left out what I consider
the most important basis for judging a model -- the extent to which the
behavior of the model matches the behavior of the system under study.
Once you have a model that works, then I agree that it is worthwhile to
look at its assumptions to see if they are consistent with other
relevant knowledge. On this basis the PCT model stands
head and shoulders above all other behavioral models; the PCT model
is consistent with relevant knowledge of physics (the model of the world
in which control systems operate) and neurology (the model of the
organic basis of perceptual control).

For example, say I build a bowling machine...I have replicated, to
some extent, the goals of the human system with regard to the
activity of bowling. But that thing, that machine, is in
no way isomorphic with human beings;

Maybe. Maybe not. You've got to keep on testing to determine the limits
of your model. You have to see whether the model is
controlling the same variables that the human bowler is controlling;
and whether it is doing this in the same way. I say you have a ton
of experimental research to do to determine how well your bowling
machine models the human bowler. If it turns out that the bowling
machine does keep predicting every detail of human bowling then I
think it is, indeed, time to see whether the machine components
(assumptions) are consistent with other knowledge (such as the
physiology of the bowler). If they are not, then the problem is
to build an equivalent model that is consisent with such knowledge.

If you build a model with a comparator, for example, show me the
corresponding comparator in the human being.

Bill Powers already noted that there is strong evidence for the
existence of neural connections that can (and do) function as
comparators. A comparaison is a VERY simple operation to implement
in neurons. The hard thing to do is to compute perceptual varables.
Why don't you like comparators?

Give me your underlying assumptions about why you think that model
is isomorphic to some empirical reality. Then we can begin to make > some progress.

I can't give _assumptions_ about why I think a model is "isomorphic
to some empirical reality". I can just give my reasons for thinking
so. My reasons are 1) the model keeps fitting empirical observations and
2) the components of the model are consistent with other relevant
knowledge (physics and physiology in the case of PCT). I think the
only assumption I make in my work is that it IS possible to invent
models of purposive behavior that behave like the reality I observe (and
operate upon). So far, I have been give no compelling reason
to abandon this assumption. A compelling reason would be a demonstration
that reality is capricious; that it is possible to
model a behavior on one occasion and not be able to model the same
behavior (produced under the same circumstances) on another.