"Deep" understanding

[From Rick Marken (940217.0950)]

Bill Leach (940215.20:57 EST(EDT) --

I would guess that there are probably quite a few things that people
generally think of as characteristics for humans, that PCT would predict
a behavioural bias;

Bill Powers (940216.0920 MST) gave the PCT "position" on human
"tendencies and biases" rather nicely, I think.

Martin Taylor (940215 20:45) --

Me:

The conflict between information
theory and PCT is a result of a difference over whether it is important,
in principle, to seek "continuity with the past" as a way of "gaining
acceptance for and contibuting ot the development of PCT".

I don't give a hoot about continuity with the past,
in regard to scientific theories. What works is what works

So your higher order goal is different than mine. Your higher order
goal is a theory that "works". Hmmm. I thought that was what I wanted
too. Let's see....

Martin:

one cannot really understand
hierarchic control at any deep level without coming to terms with its
informational implications. Understand at a deep level does NOT mean
the ability to program models that mimic behaviour. That ability is a
proof of concept, that one's ideas are plausible, and very valuable as
such. But it does not represent a deep understanding, in my view.

Ah. I think I see the glimmerings of your higher level perspective.
What you seem to mean by a theory that "works" is "a theory that gives
you _deep_ understanding". So a model that acts almost exactly like a
living system might "work" from MY perspective but it doesn't "work" from
your perspective -- because it doesn't give "deep understanding".
Fortunately, this "deep understanding" is provided by IT -- says you.

I think I can agree with you that PCT does only "work" in the old
fashioned scientific sense. It's just a model that, acting on its own,
produces results (behavior of the observable variables) that are almost
identical to those produced by the real system, under the same circumstances.
We have answered the question "why do these variables behave as they do" by
providing a model that says "because the system is organized as a control
system". Now you want to go "deeper" which, I presume, means asking
questions like "why is the system organized as a control system"? This seems
to me like asking "why is f=ma"? I don't think these are the kinds of
questions that can be answered by modelling (ie. by science). I think they
are not so much "deep" as they are "unanswerable". But many people do think
that the understanding provided by "modeling" (ie. by science) is not "deep"
enough. These are usually people who don't want to take the trouble to try
to understand the mdoels, but they are also the "The Tao of Physics" types
who have at least some grasp of the "shallow" models. I personally find such
"deep" approaches to science pretty unsatisfying. For some reason, just
understanding how the PCT model actually works gives me enough satisfaction
and the sense that I have learned a LOT about purposeful behavior.

The problem, for me, with the kind of "deep" understanding you seek is
that it is often pursued before there is a firm grasp of the essentials
of the "shallow" model. This is what SEEMS to be happening in the IT
discussion where some of the "deep" understandings provided by IT often
contradict what we already know from our "shallow" understanding of how a
control system works. For example, you have said that what would be really
interesting (I presume you mean "deep") to find out about control is
"...how the uncertainty of the disturbance given the perceptual signal
might affect the usable gain in the loop, the need for, and effectiveness
of, higher levels of control, and the like". Assuming that "uncertainty"
refers to the instantaneous probability distribution of d given p
(P(d|p) over d) then we already know (from the PCT model) that estimates
of P(d|p) are not involved in its operation; and for good reason -- there
is no information about d in p so P(d|p) is alway the same, 1/range(d), no
matter what the value of p.

I do not want to interfere with anyone's quest for "deep" understanding;
but I can't help throwing in my 2 cents when some premise on which this
"deep" undertanding is to be built is already understood (an the basis of
our shallow models) to be false. I might be out of my depth with IT but
I'm very familiar with the shallow waters of PCT.

Best

Rick