On Purpose

[from Gary Cziko 930102.1800 GMT] (STILL typed "92" first!)

Rick Marken (930102.1200) warns Bill Powers:

If you don't write this book, Bill, I will! (Visions of a PBS
documentary are already starting to dance in my head). That was
a really incredible post.

No fair! I was planning to write such a book after finishing _Without
Miracles_. I was going to call it _On Purpose_ (I could then tell people
that I was writing a book "on purpose"). Bill's perspective of animism to
anti-animism to control gives a wonderful organizing framework for the
book. But since the current book is taking me so long (when my wife keeps
telling people "My husband has almost finished his book" they reply "He
must be a slow reader") I will let Bill and Rick take the first crack.
With the book should come a computer disk with Rick's computer demos and an
appendix with "my" collection of manual demonstrations of control (mostly
stolen from others).

There should also be a videotape companion to the book. It would include
both the computer and manual demonstrations of control. I've have just
viewed such a video which is a companion to John Koza's _Genetic
programming_ and it is a very effective way to demonstrate the algorithms
he describes in the book. And just about everybody (in North America at
least) seems to a videotape player these days. This would also give Rick a
start on preparing his PBS special (see what you've started, Ed?).--Gary

···

------------------------------------------------------------------
Gary A. Cziko Telephone: 217.333.8527
Educational Psychology FAX: 217.244.7620
University of Illinois E-mail: g-cziko@uiuc.edu
1310 S. Sixth Street Radio: N9MJZ
210 Education Building
Champaign, Illinois 61820-6990
USA
------------------------------------------------------------------

[From Chuck Tucker (08.20.2007 12.57 PM)]

Bill said (on 8/20):

"One of these days I'm going to design that experiment I've described before. Set up a two-person task in which both one person sees a display of movable objects, and the other sees the same display, but the objects are not movable. The "instructor" tries to tell the other person how to move the objects so as to match a configuration the instructor sees printed out on a sheet of paper.
Example "Put the red circle between the blue square and the yellow triangle."

Of course this could be done without a computer, using two sets of objects on tables with a screen between them. Maybe someone else will beat me to this."

This reminds me of several pieces of research that may be related to your issue.

The first was reported in the Frontline program several years ago (I think you wrote a piece on this to Bruce) on the issue of "Facilitated Communication" where the claim was made that autistic non-speaking children could write poetry and do math with the assistance of a communicator guiding their arm using a typewrite-like board. The study was set up with a board between the autistic child and a facilitator who was quite familiar with the child and they were shown pictures of objects (e.g., a shoe, a sandwich, a boat, etc.) that the child had seen and named before in the co-presence of the facilitator but in all these instances in this study the facilitator could not see what the child saw and visa-versa. In every instance (N=179) the child gave the name of the object that the facilitator saw and not what the child was shown. No statistics were necessary the results were 100%.

The other study is the famous study by Sherif (Clark took Social Psychology from him at OU) on the "autokinetic effect" (this study is also related to Ash's study [the teacher of Stanley Milgram] on the length of lines) where the participant was asked to estimate the movement of a light in a dark room. In several of the studies the experimenter was the only other participant while in others their were several other participants. In most of the instances (not clear on the numbers here) the participant stated that the light moved consistent with what he was told by the other participants(some of Sherif's work is described in Clark's book, pp. 62-70). Of course, in all of the studies the light was stationary. Someone replicated this study (I can't recall his name) where the participant were told the light did not move and they reported no movement.

I hope these may be helpful to you and others on CSGNET.

Chuck

[From Rick Marken (960911.0900)]

Purpose n 1. something set up as an object or end to be attained 2. an
action in course of execution (Webster's New Collegiate Dictionary, 1980).

Bruce Abbott (960910.2100 EST) --

I would prefer NOT to use "purpose" to refer to a reference signal; it seems
wrong to say that the purpose of my thermostat is 72 degrees Faherenheit.

The HPCT reference signal seems to correspond exactly to the first dictionary
definition of "purpose"; the reference signal is a specification that is "set
up" (by another control system) as an object or end to be attained. By this
definition, the purpose of the thermostat is not a temperature (which is a
controlled varable); it's the reference specification (implemented as the
location of contacts) for the perceptual representation of the temperature
(size of the bimetallic strip).

The process of acting to bring the controlled variable to the reference level
can also be called "purpose"; this would be consistent with the second
dictionary definiton of the term; in this case "purpose" is the process - -
an "action in course of execution" -- that produces the end to be attained.

the bird can also give itself a bath by means of those wings, but if that
were the purpose of wings then surely something more efficient would have
developed by now to serve that purpose

So the "purpose" of wings is to allow flying rather than bathing because
wings are so perfectly suited to flying rather than bathing? The "purpose" of
an object, then, comes from how well suited it is to producing a particular
result. Yet you say: "I would reject the idea that the purpose of the Moon is
to cause tides" even though the Moon is well suited to producing this result.
So there is more to saying that an object has a "purpose" than the fact that
it is well suited to producing a particular result; it is also required that
the object be "selected (and natural selection is a form of selection) for
that purpose" -- ie. to produce that result.

So, according to you and Simon, it seems that an object has a "purpose" if it
is 1) well suited to producing a particular result and 2) this suitability
was selected. These seem like pretty subjective criteria to me. How do I test
to determine whether an object has purpose? Did the structures that we now
call wings have a purpose before they could function as wings? Do objects
have only one purpose at a time? Does the purpose of an object change
suddenly or gradually? Did the structures that were to become wings suddenly
have a new purpose as soon as they could be used to fly or did they gradually
gain this purpose, overlapping perhaps with their previous purpose (if they
had one), in evolutionary history? What is the scientific merit of your
opinion that the wing has only one real "purpose" -- flight?

In PCT, purpose is a testable phenomenon; we can objectively determine
whether a system has a purpose (a reference for the state for a particular
perceptual variable). The Test for the controlled variable is a test of
purpose. This is the kind of purpose -- the conventionally agreed on
referent of the word "purpose" -- that Simon and all other conventinoal
psychologist shows no evidence of understanding. PCT is about this kind of
purpose; that's why Simon has nothing to contribute to a PCT discussion of
purpose -- except Nobel quality noise.

RSM

[From Bruce Abbott (960911.1705 EST)]

Rick Marken (960911.0900) --

Purpose n 1. something set up as an object or end to be attained 2. an
action in course of execution (Webster's New Collegiate Dictionary, 1980).

The HPCT reference signal seems to correspond exactly to the first dictionary
definition of "purpose"; the reference signal is a specification that is "set
up" (by another control system) as an object or end to be attained. By this
definition, the purpose of the thermostat is not a temperature (which is a
controlled varable); it's the reference specification (implemented as the
location of contacts) for the perceptual representation of the temperature
(size of the bimetallic strip).

Yes, the purpose is the end _to be attained_, not the end in and of itself.
The purpose of the control system is not the value of the reference signal
(e.g., 72 degrees F) but the end to be attained (i.e., bring the room
temperature up to 72 degrees F and hold it there).

So, according to you and Simon, it seems that an object has a "purpose" if it
is 1) well suited to producing a particular result and 2) this suitability
was selected. These seem like pretty subjective criteria to me. How do I test
to determine whether an object has purpose? Did the structures that we now
call wings have a purpose before they could function as wings? Do objects
have only one purpose at a time? Does the purpose of an object change
suddenly or gradually? Did the structures that were to become wings suddenly
have a new purpose as soon as they could be used to fly or did they gradually
gain this purpose, overlapping perhaps with their previous purpose (if they
had one), in evolutionary history? What is the scientific merit of your
opinion that the wing has only one real "purpose" -- flight?

I wish you wouldn't put words in my mouth. "What is the scientific merit of
your opinion that the wing has only one real 'purpose' -- flight," you ask.
Now this assumes that somewhere in my writing I have asserted that the wing
has only one real purpose, doesn't it? What I did say is that the _main_
purposes of the wing are to provide lift, thrust, and control (during
flight), and this is hardly the same thing as asserting that these are the
_only_ purposes for which wings have evolved.

For most artifacts created by human beings, we can ask the designer or other
knowledgable person what the artifact is "for" (and there may be multiple
answers); we discover the intention of its designer. We learn that
screwdrivers are for driving screws and, if we are properly taught, that
they should NOT be used for chizeling as this tends to damage the
screwdriver and render it unsuitable for its intended purpose. But the
screwdriver has characteristics that may suit it to other purposes (I might
use a big one as a substitute tent stake, for example).

If there are no knowledgable people around to provide a hint, I may have
great difficulty discovering the purpose an artifact was designed to serve.
I recently visited a museum in which several rather mysterious implements
were on display and the visitor was invited to guess that they were for. I
can tell you that I didn't have a clue. Still, if you told me what the
purpose of each implement is, I could probably easily identify which
features of the implement contribute to its ability to serve that purpose.

Discovering the purpose from the characteristics of the artifact is much
more difficult: a matter of inference. Archeologists often have great
difficulty determining what purpose some artifact was designed to serve.
But the fact is that the intended purpose of an artifact often can be
inferred from its structure and organization. The job is even easier if you
can see the object "at work," doing what it was designed to do.

Now when we turn our attention to "natural" artifacts, I grant that there is
no entity out there with any intentions, no designer and thus no purpose in
the dictionary sense of intended use. Yet the structures that evolve still
have that same character, AS IF they had been designed for a purpose. The
reason they are as they are is that, over the course of evolution, these
were the "solutions" that worked: that aided in the survival of the
organism. If the bird's purpose (end to be attained) is to fly, it is able
to do so only because its wings have been moulded over the course of
evolution so as to exquisitely serve this purpose. They are not so well
suited to other purposes for which the bird may employ them, such as beating
off attackers.

Objects whose characteristics have been shaped in this way can be described
in ways that objects not so shaped cannot be. We can ask of them, what does
this or that feature contribute to the function it serves, given the
environment in which it must operate? In the control system, what is the
function of the sensor, the comparator, the reference signal, the output
mechanism? What is the function of the system as a whole? In objects not
so shaped, their characteristics are merely their characteristics. They are
as they are, and that's it. In biological organisms the characteristics
have been so shaped. We can ask, what is skin for, and why is the skin of
the seal backed up by a thick layer of blubber, whereas the skin of a dog is
not? What is the end these features help these animals to attain, the
purpose of these features?

If you're going to insist that there is no scientific way to discover the
answer, then you'll have to explain how biologists have been able to figure
out those purposes for so many biological structures.

Regards,

Bruce

[From Rick Marken (960921.1500)]

Bruce Abbott (960921.1145 EST) --

That was a wonderful quote from Kleiber; thanks for posting it. I think it
shows clearly why talk of purpose in evolution and behavior gets so
confused. Clearly, biologists don't know what purpose (control) is. For
example, Kleiber says:

To seek a goal, however, implies the expectation of future events and
implies further that present actions are based on this expectation.

PCT shows that this is incorrect. To seek a goal (carry out a purpose)
implies no expectation of future events. It implies a perceptual
representation of a current state of affairs, a (current) reference
representation of the intended (goal) state of affairs, the ability to
compare (measure the current difference between) these two variables and
act to reduce this difference. Everything in this "goal seeking" process --
goal (reference), comparison, action, and result (perception) -- is
happening simultaneously, in the present; the future is in the future.

Kleiber's belief that goal seeking involves the "expectation of future
events" seems to have led him, in the next sentence, to refer to goal
seeking as "planning";

Planning, in turn, requires a mind.

So Kleiber seems to recognize that goal seeking (planning, purposeful
behavior, the achievement of future ends) requires something that at least
some animals (like himself) possess: a mind. So Kleiber must believe that
goal seeking can be done by animals with minds. Nevertheless, he goes on to
say:

In biology, however, teleological explanations are still used extensively.

Few biologists actually believe that animals and plants plan their own

future, yet some, like Krebs (1954), feel that our understanding of an
animal, like our understanding of a machine, is improved by knowing its

purpose.

So Kleiber seems to be contradicting himself -- at least, he is if he
accepts the fact that he is an example of _one_ kind of animal with a mind.
If he does accept this, then he must admit that teleological explanations
are appropriate in at least _some_ cases; they are appropriate in cases
where animals and plants actually _do_ plan (set a reference for) their own
future. So Kleiber should be agreeing with Krebs: our understanding of some
animals (the kind with a mind; the kind that seek goals) is, indeed,
improved by knowing their purpose (what goals they "plan" to
acheive).That's how PCT contributes to our understanding of animal
behavior; it lets us determine an animal's purposes.

Kleiber is sure right about one thing:

A design implies a designer, a planning mind.

The problem, of course, is to determine when a particular state of the
world is the result of design and when not. This is equivalent to knowing
when a teleological explanation is appropriate and when it is not. Kleiber
implies that teleological explanation is _always_ inappropriate. That
attitude works against PCT. PCT says that teleological explanations are
_sometimes_ appropriate; you just have to know _when_ they are appropriate.
They are appropriate when the animal is producing results _on purpose_. PCT
shows us how to tell when results are being produce on purpose. This is
done by Testing to see whether some state of the world is or is not under
control; whether a variable is in a particular state by accident
(non-teleologically) or by design (teleologically). If one believes, as a
_matter of principle_, that teleological explanation is always
inappropriate, then one will be strongly inclined to reject the notion that
PCT has anything worthwhile (let alone revolutionary) to contribute to a
scientific understanding of the behavior living systems.

There is no evidence that Darwinian "selection" is teleological; indeed,
there is considerable evidence that it is not . For example, Darwinian
selection seems to have made no effort to preserve the dinosaurs -- which
had been around for millions of years -- after their numbers were reduced
by a large but brief (in geological terms) disturbance. Similarly, there is
no evidence that the behavior of most of the physical universe is
teleological; cause-effect (in the form of Newton or Einstein's laws) seem
to work fine. But there is strong evidence that the behavior of plants and
animals is teleological; there is just no other way to explain the fact
that these systems act to produce consistent results (tree trunks, houses,
roads, cars, handshakes, kisses, etc) under highly variable circumstances.

Best

Rick

From mr. Remi Cote 220996.2150

Kleiber talk about goal, Mr Powers was so specific
in his theory to speak about reference...These
are no synonym.
Amybe don't have goal, they don't even have idea...
but they live!

···

***
I wanted to tell to people on the list that I received
some more biblio, on equilibrium, second law of thermodynamic
and all those nice theory that I always thought I was too
dump to study, and now that I almost finish my thesis I am
confronted with these rampant deterministic vilain...

So I will dig in this rich mud, try do digest a bit of it
and deliver my little baby on this list... It will be naive.

Right now I have some interesting lead:

Kugler and Turvey : _Information, natural law and the self-
assembly of rythmic movement_

Brooks: intelligence without reason 1991

Beer, Randall: A dynamical systems perspective on agent environement
interaction

Stuff by Port R. & Gelder, Tim van: Mind as motion: explorations
in the dynamics of cognition. Cambridge, Mass.: Bradford MIT

Thelen & Smith: _A dynamic system Approach to the development
of cognition and action (1994)

So I would like to thanks every one who help me in this. No it
is time to look at this material...If any of you have further
biblioindication, please send it to me:

coteremi@ere.umontreal.ca

I will report for a small review. All those Idea are close to HPCT
Some are very complex AI stuff, very technical, I won't report
that. I will mostly try to survey stuff concernig the upper layer
of HPCT, since that is where my modelisation is concerned with.

Thank's REMI