Allen Newell and PCT

CSGnet Archive CSG1997
1

[From Fred Nickols (971227.1420 EST)]

Bill Powers (971227.0520 MST)--

Control is a phenomenon, not a theory. An adequate model of control can be
constructed to explain the phenomenon without bringing in all the possible
organizations that are NOT control systems. If we observe that human
behavior demonstrates the properties of control systems, then to explain
what we observe requires only that we explain control. All the other
possibilities become irrelevant until such time as we find some other
identifiable mode of operation.

Bill:

I pulled the snippet above from your "butting in" post to Martin Taylor
about his interactions with Tim Carey. I did so because I was leafing
through the late Allen Newell's book, "Unified Theories of Cognition,"
earlier today and came across the following remarks (pp.44-45) which I
presume you and/or Rick Marken might care to critique.

···

****************

"2.2 Knowledge Systems

How then should we describe systems? How should we describe their response
functions? To speak of mind as a controller suggests immediately the
language of control sytems--of feedback, gain, oscillation, damping, and so
on. It is a language that allows us to describe systems as purposive
(Rosenbluth, Weiner, & Bigelow, 1943). But we are interested in the full
range of human behavior and response--not only walking down a road or
tracking a flying bird, but reading bird books, planning the walk, taking
instructions to get to the place, identifying distinct species, counting the
new additions to the life list of birds seen, and holding conversations
about it all afterward. When the scope of behavior extends this broadly, it
becomes evident that the language of control systems is really locked to a
specific environment and class of tasks--to continuous motor movement with
the aim of pointing or following. For the rest it becomes metaphorical.

A way to describe the behavior of systems with wide-ranging capability is in
terms of their having knowledge and behaving in light of it."

Newell continues down the path he, J.C. Shaw and Herbert Simon followed for
years, namely, behavior as produced response functions.

My specific question is this: How would you (or Rick) respond to Newell's
assertion that the language of control systems is metaphorical for all
behavior but that of pointing or following?

Come to think of it, I have a second question. Are there PCT demos that
don't hinge on tracking behavior?

P.S. Your remarks to Martin reminded me of Jay Forrester's chiding of the
current crop of system dynamics gurus for insisting that only those who are
well grounded in advanced mathematics can truly understand, appreciate, or
do really good system dynamics modeling. In full forum at an MIT gathering,
Jay said, "That's simply not true. All you need is some basic math and a
little algebra."

P.P.S. I recall still from my fire control radar days of long ago the
infamous "flip-flop multivibrator" circuit (including the electronic kick in
the pants that was necessary to get it started) and I'm darned if I can see
how knowing about that kind of circuitry is essential to grasping control
theory.
Regards,

Fred Nickols
nickols@worldnet.att.net

2

[From Bill Powers (971228.1326 MST)]

Fred Nickols (971227.1420 EST)]

I was leafing
through the late Allen Newell's book, "Unified Theories of Cognition,"
earlier today and came across the following remarks (pp.44-45) which I
presume you and/or Rick Marken might care to critique.

Newell continues down the path he, J.C. Shaw and Herbert Simon followed for
years, namely, behavior as produced response functions.

My specific question is this: How would you (or Rick) respond to Newell's
assertion that the language of control systems is metaphorical for all
behavior but that of pointing or following?

I would say, "No, it isn't."

If he wanted to know more, I would tell him about the model. Having met him
a couple of times, I don't think he would have wanted to know more.

Come to think of it, I have a second question. Are there PCT demos that
don't hinge on tracking behavior?

Have you gone through the demo disks? There are half a dozen demos
involving behaviors other than tracking a target with a cursor. Ordinary
life is full of them: driving a car, making breakfast, getting an
education, solving an equation for an unknown, helping a child put together
a wooden train layout. I can't think of anything I've done for the last
week that wasn't a demo of control behavior. Have you done anything that
wasn't?

P.S. Your remarks to Martin reminded me of Jay Forrester's chiding of the
current crop of system dynamics gurus for insisting that only those who are
well grounded in advanced mathematics can truly understand, appreciate, or
do really good system dynamics modeling. In full forum at an MIT gathering,
Jay said, "That's simply not true. All you need is some basic math and a
little algebra."

I agree wholeheartedly. It's too easy to get lost in complex mathematical
problems and fail to understand the basics of the system you're modeling.

P.P.S. I recall still from my fire control radar days of long ago the
infamous "flip-flop multivibrator" circuit (including the electronic kick in
the pants that was necessary to get it started) and I'm darned if I can see
how knowing about that kind of circuitry is essential to grasping control
theory.

Thank you.

Best,

Bill P.

3

[From Rick Marken (971228.1715)]

Fred Nickols (971227.1420 EST)]

I was leafing through the late Allen Newell's book, "Unified
Theories of Cognition," earlier today and came across the
following remarks (pp.44-45) which I presume you and/or Rick
Marken might care to critique.

Here goes. Newell says:

But we are interested in the full range of human behavior and
response--not only walking down a road or tracking a flying
bird, but reading bird books, planning the walk, taking
instructions to get to the place, identifying distinct
species, counting the new additions to the life list of
birds seen, and holding conversations about it all afterward.
When the scope of behavior extends this broadly, it becomes
evident that the language of control systems is really locked
to a specific environment and class of tasks--to continuous
motor movement wit the aim of pointing or following. For the
rest it becomes metaphorical.

This statement simply reflects the fact that Newell didn't
know what control is. Each of the behaviors he names as an
example of a behavior that _can't_ be handled by control
theory ("the language of control systems") is, in fact,
an example of _control_. "Reading bird books" is a behavior
that involves control of many perceptions; the relationship
of book to eye, the location of text on the eye, the rate
at which new text is brought into view, etc; it also invloves
control of the imaginations elecited by the words being read.
"Planning the walk" involves control of perceptions and
imaginations such as duration of walk, strenuousness of walk,
distance to start of walk, map of walk area, etc. "Taking
instructions to get to the place" involves control of perceptions
of what instructions are being given and imaginations about
what these instructions imply. "Identifying distinct species"
is control of a category perception (type of bird). "Counting
the new additions to the life list of birds seen" involves
control of perceptions like category ("that was a golden crowned
kinglet") and relationship ("it's not currently on my list; now
it is"). "Holding conversations about it all afterward" involves
control of perceptions of things like "conversations",
"friendships" and "honesty" ("should I admit that I never really
_saw_ the golden crowned kinglet, just its egg").

My specific question is this: How would you (or Rick) respond
to Newell's assertion that the language of control systems is
metaphorical for all behavior but that of pointing or following?

I would say he's right if all (human and animal) behavior but
pointing and following _didn't_ involve control. In fact, the
only human and animal behavior I can think of (off-hand) that
doesn't involve control is falling (at 32 ft/sec/sec). Just
about everything else people do involves control of perception.
Ergo, control theory is no metaphor; it's the way behavior works.

By the way, there are four demos in my Java set that don't
really involve tracking: "Control of Perception", "Behavioral
Illusion", "Selection of Consequences" and "Hierarchy of
Perception and Control").

Best

Rick

···

--
Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates e-mail: rmarken@earthlink.net
http://home.earthlink.net/~rmarken/