Abbott is not alone

CSGnet Archive CSG1997
1

[From Bill Powers (970803.1121 MDT)]

Rick Marken (970803.0900) --

Yesterday I discovered that Bruce Abbott is not the only one
who believes that Darwinian natural selection is a closed-loop
control process. While checking to see whether Yahoo had changed
the URL for "Mind Readings" (they haven't) I "tumbled" upon a
chapter by William Calvin, a biologist (I think) at the U. of
Washington, in which he describes Darwinian selection as being
equivalent to E. coli navigation (http://weber.u.washington.edu/
wcalvin/bk4/bk4ch2.htm). He never explicitly says that Darwinian
selection is closed loop control but he does imply that the
Darwinian/E. coli process is purposeful. I quote:

PURPOSE SEEMS SO DIFFERENT FROM CHANCE, but darwinism suggests
that you might be able to have your cake and eat it too: chance
plus selection, repeated for many rounds, can achieve much.

I then realized that the same thing happens in psychology:
reinforcent theorists see reinforcement theory as a negative
feedback control process in general and as equivalent to
Darwinian selection and E. coli navigation in particular.

Given the apparent ubiquitousness of this belief (that natural
selection and reinforcement are closed loop control processes)
it seems to me that it is rather useless and irrelevant to continue
to try to demonstrate that natural selection and reinforcement are
really _not_ control models. They simply _are_ closed loop control
models from the point of view of too many people.

That's kind of a downer. I've spent most of my life trying to persuade
psychologists that the control model ought to be considered, and now
they're "reinventing" it, or at least adopting the language.

Something like this happened to me in the 1950s. A friend of mine and I
invented a lawn sprinker with a limp rubber hose that thrashed around at
random spraying water in all directions very nicely. It was also pretty
funny. We took it to a large hardware chain and offered to sell the idea to
them. They turned it down. Six months later, Kirk Sattley came back to
Chicago from a trip East, and said "Guess what I saw in Grand Central
Station -- your water sprinkler thrashing away under a plastic dome." A
display of a new product by a large hardware company.

I know that lots of psychologists have become aware of my work over the
years. But they don't address it directly; instead, they think up
refutations of their understanding of my proposals without actually saying
they're aimed at my ideas -- my work is seldom mentioned. It's like seeing
a negative of a photograph, or the light place under the rectangle where a
picture used to hang. Do you suppose the time will ever come when a
historian of science comes right out and says, "Hey, Powers was saying all
this 20 years before anyone else, so where are the citations?" Well, in the
case of E. coli it was 14 years, but that's averaged out by the things I
was saying 40 years ago. I know that I pretend that I don't care about such
stuff, but I really do. It all seems terribly unfair.

Best,

Bill P.

2

[From Rick Marken (970803.0900)]

Yesterday I discovered that Bruce Abbott is not the only one
who believes that Darwinian natural selection is a closed-loop
control process. While checking to see whether Yahoo had changed
the URL for "Mind Readings" (they haven't) I "tumbled" upon a
chapter by William Calvin, a biologist (I think) at the U. of
Washington, in which he describes Darwinian selection as being
equivalent to E. coli navigation (http://weber.u.washington.edu/
wcalvin/bk4/bk4ch2.htm). He never explicitly says that Darwinian
selection is closed loop control but he does imply that the
Darwinian/E. coli process is purposeful. I quote:

PURPOSE SEEMS SO DIFFERENT FROM CHANCE, but darwinism suggests
that you might be able to have your cake and eat it too: chance
plus selection, repeated for many rounds, can achieve much.

I then realized that the same thing happens in psychology:
reinforcent theorists see reinforcement theory as a negative
feedback control process in general and as equivalent to
Darwinian selection and E. coli navigation in particular.

Given the apparent ubiquitousness of this belief (that natural
selection and reinforcement are closed loop control processes)
it seems to me that it is rather useless and irrelevant to continue
to try to demonstrate that natural selection and reinforcement are
really _not_ control models. They simply _are_ closed loop control
models from the point of view of too many people.

I can also see why Bruce Abbott thinks that the reinforcement
model in my "selection of consequences" demo is a "straw man".
It is a "straw man" if you believe that reinforcement theory
really is a control model. I can also see why the arguement about
natural selection will never end; natural selection simply _is_ a
control model from the point of view of its advocates.

I think the only thing we control theorists can do under these
circumstances is take the word of those who advocate these models:
natural selection is a closed loop control model; reinforcment is
a closed loop control model. Let's agree that these theories are,
indeed, closed-loop control models and that PCT is, therefore,
another _version_ of these theories.

If we can agree on that (that all these theories are closed loop
control theories) then we can ask the advocates of natural
selection theory and reinforcement theory and other control theories
to tell us what variables these models are controlling and what
research has been done to determine that the organisms modeled
are actually controlling these variables. That is, I suggest we
stop the debate about whether PCT is like (or "nothing but")
any particular theory (whether theories other than PCT are
really closed loop control theories) and start looking at data
on control.

Let's agree that PCT is like natural selection theory or rein-
forcement theory. Then let's ask 1) what variables are controlled
by selection theory and reinforcement theory and 2) what research
has been done to determine whether or not these variables are,
indeed, controlled, as specified by the theory.

If natural selection theories and reinforcement theories are,
indeed, control theories then they must have been developed
to explain a control process; the theories themselves must
control some variable(s); and researchers must have tested to
determine whether these variables actually are under control.
I'd like to hear about all the controlled variables that have
been discovered by reinforcement theorists and evolutionary
biologists -- and about the research that was done to discover
them. I would really like to have it described on CSGNet; no
"see such and such paper" or "see the writing of person X". I'd
just like to hear a description of the variables controlled by
the models and how research has determined that those variables
are, indeed, controlled.

i.kurtzer (970802) --

well guys, the time is up and there have yet been no suggestions
excepting functional analysis..So what am i to draw from this lacuna?
That noone reads my posts

I read you posts. I didn't answer because I knew I'd win (or, at
least, tie) and I don't drink beer anymore;-)

Best

Rick

···

--

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

3

[From Bruce Abbott (970804.0810 EST)]

Rick Marken (970803.0900)

Yesterday I discovered that Bruce Abbott is not the only one
who believes that Darwinian natural selection is a closed-loop
control process.
. . .
I then realized that the same thing happens in psychology:
reinforcent theorists see reinforcement theory as a negative
feedback control process in general and as equivalent to
Darwinian selection and E. coli navigation in particular.

I think the only thing we control theorists can do under these
circumstances is take the word of those who advocate these models:
natural selection is a closed loop control model; reinforcment is
a closed loop control model. Let's agree that these theories are,
indeed, closed-loop control models and that PCT is, therefore,
another _version_ of these theories.

You use the word "believe" to describe how these folks view the Darwinian
and reinforcement processes, as if the natures of these models were matters
of faith. There is no need to "take the word" of anyone. If these are
closed-loop, negative feedback systems, then a simple analysis will show
them to be such.

But your approach has not been to submit these views to such an analysis.
Instead, you repeatedly assert that they are open-loop systems, as if by
mere repetition you could make it so.

Whether one "believes" that these models do or do not apply to real
organisms is another matter entirely, but presumably one that can be settled
by empirical test.

Regards,

Bruce

4

[From Rick Marken (970704.0800 PCT)]

Bruce Abbott (970804.0810 EST) --

You use the word "believe" to describe how these folks view the
Darwinian and reinforcement processes, as if the natures of these
models were matters of faith....If these are closed-loop, negative
feedback systems, then a simple analysis will show them to be such.

Not quite true. The reinforcement model, for example, is basically
a verbal model. My interpretation of the verbal model yields a
passive selection (sieve) model; responses followed by reinforcers
become more probable (get through the sieve); responses not
followed by reinforcers become less probable (don't make it through
the sieve).

But your approach has not been to submit these views to such an
analysis.

I have simply tried to implement the verbal (and sometimes
mathematical) models as working programs. The result has been
a passive selection process. I am willing to admit that my
implementations of the reinforcement model may be wrong (I'm
getting pretty tired of hearing my efforts described as "straw
men" -- though I was very fond of the straw man in Wizard of Oz).

Instead, you repeatedly assert that they are open-loop systems,
as if by mere repetition you could make it so.

You're right. I give up. As I said in my earlier post [Rick Marken
(970803.0900)]:

I think the only thing we control theorists can do...is take
the word of those who advocate these models:...reinforcment
is a closed loop control model.

What I want to do now is hear about the research that has been
done to test these purportely close-loop control models of
reinforcement. I know that there has been a ton of research
done based on reinforcment theory. If reinforcement theories
do, indeed, translate into closed loop control models of behavior
then (as I'm sure you know) these models must make some assumptions
about the variable(s) that the organism is controlling. I want
to know 1) what variables are controlled by closed loop
reinforcement models and, more important, 2) what research
has been done to determine whether or not these variables are,
indeed, controlled.

Unless you can point me to a systematic line of reinforcement
theory research aimed at determining controlled variables,
I'm afraid I am going to be inclined to believe what I have
believed all along; that reinforcement theory (and it's parent,
natural selection theory) is a passive, open-loop seive theory
being promoted as an explanation of purposeive behavior by
swamp people;-)

Best

Rick

···

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

5

[From Bill Powers (970804.08723 MDT)]

Bruce Abbott (970804.0810 EST) (replying to Rick Marken) --

You use the word "believe" to describe how these folks view the Darwinian
and reinforcement processes, as if the natures of these models were matters
of faith. There is no need to "take the word" of anyone. If these are
closed-loop, negative feedback systems, then a simple analysis will show
them to be such.

Of course I agree with you about the last statement. But the requirements
for doing such a test are seldom met. I have periodically raised the point
that if reinforcers are controlled variables, they should prove resistant
to disturbance. However, if disturbances of the reinforcement rate were
introduced, and control really existed, we would quickly see that a given
reinforcement rate might go with either an increase or a decrease in
behavior rate, and the concept of reinforcement would quickly fall apart.
Reinforcement is said to increase the probability of the behavior that
produces the reinforcement. If we independently disturb the reinforcement
rate, however, we will find that a different behavior rate results, as
required to maintain the same reinforcement rate. Now we have (if control
is involved) either increases or decreases in behavior rates resulting from
the same reinforcement rate, so there is no one behavior rate to be
reinforced any more. In fact we could probably arrange for disturbances
that require different _kinds_ and _directions_ of behavior to produce the
same reinforcement, which would make the case for control and against
reinforcement theory even more strongly.

When reinforcement theory is defended in the absence of tests like this
(which are also equally severe tests of the concept of control), what could
be the justification for believing in reinforcement other than faith?
It's no test to set up a situation where the critical factors are held
constant, and then interpret the results as supporting reinforcement theory.

In your "test" of the idea that Aplysia's gill-withdrawal is a result of
natural selection, all you did (in thought) was prevent the gill withdrawal
and show that Aplysia's mortality rate went up. But that is not what is
being questioned. The question is whether blind variation and selective
retention of genome characteristics through changes in reproductive success
would be sufficiently effective to produce this particular neural circuit
in the time that elapsed between Aplysia's not having and having this
neural circuit. Your test doesn't answer this question at all -- except by
assuming, on faith, that this mode of evolution MUST HAVE BEEN effective
enough, since the neural circuit DID appear. This is exactly the mode of
argument that says there MUST HAVE BEEN a God who created the universe,
because the universe DOES exist. This kind of argument has been recognized
for two millenia or more as specious.

Giving examples that can be interpreted as supporting a theory is no test
of a theory. In fact, that's just how people argue in supporting their
faiths. I prayed, and my migraines went away. What more proof do you need
of the power of prayer?

Best,

Bill P.