Review of Cziko's "Things We Do"

CSGnet Archive CSG2002
1

[From Rick Marken (2002.04.15.1330)]

While looking for the next possible venue for my "Fielder's Choice" paper
(other than my file cabinet) I checked out _Acta Psychologica_. It probably
won't work for my paper but while at the site I noticed that the most recent
issue (volume 100 (1), 2002) has a review of Gary Cziko's "The Things We Do",
which is all about PCT in psychology and biology. The review is by someone
named S. Dewitte. They don't have _Acta Psychologica_ at my library but I would
really love to read the review. If anyone out there has access to _Acta
Psychologica_. perhaps they could scan the review and make a PDF version
available to interested parties on CSGNet.

Best regards

Rick

···

--
Richard S. Marken, Ph.D.
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org

2

[From Mike Acree (2002.04.15.1448 PDT)]

Rick Marken (2002.04.15.1330)--

While looking for the next possible venue for my "Fielder's Choice" paper
(other than my file cabinet) I checked out _Acta Psychologica_. It probably
won't work for my paper but while at the site I noticed that the most recent
issue (volume 100 (1), 2002) has a review of Gary Cziko's "The Things We Do",
which is all about PCT in psychology and biology. The review is by someone
named S. Dewitte. They don't have _Acta Psychologica_ at my library but I would
really love to read the review. If anyone out there has access to _Acta
Psychologica_. perhaps they could scan the review and make a PDF version
available to interested parties on CSGNet.

Thanks for the reference, Rick. I don't have access to a scanner, but I was able to print out a hard copy. I can mail that to you if you don't get a better offer. The volume, by the way (in case anybody else is looking) is 110, not 100.

It's an admittedly ambivalent review. Dewitte, who is on the "Faculty of Economics and Applied Economics and Fund for Scientific Research--Flanders," wants to argue, on the one hand, that there's nothing new in Gary's book (Carver and Scheier were talking about purpose 20 years ago!); on the other hand, he's willing to acknowledge that it's a contribution to put it all together.

Mike

3

[From Rick Marken (2002.04.19.0930)]

Mike Acree (2002.04.15.1448 PDT) --

It's an admittedly ambivalent review. Dewitte, who is on the "Faculty
of Economics and Applied Economics and Fund for Scientific
Research--Flanders," wants to argue, on the one hand, that there's
nothing new in Gary's book (Carver and Scheier were talking about
purpose 20 years ago!); on the other hand, he's willing to acknowledge
that it's a contribution to put it all together.-

I got the snail mail version, Mike. Thanks. Your review of it is _perfect_.
Dewitte does end by saying that the real test of Gary's "insights" will be
research based on the "novel ways of testing hypotheses" that Gary
describes (pp.242-243). And I heartilyu agree. He goes on tio suggests
waiting "a couple of years for papers that apply these insights to emerge"
-- a peculiar suggestion from one who believes that Carver and Scheier have
been doing research based on these insights all along. Of course, it's also
peculiar since _many_ papers that apply these isights have already emerged;
I think they many were even referred to in Gary's book.

The good news, however, is that Gary;s book was even _reviewed_. I hope
that gets it some readers.

On a related note, I attended a talk here yesterday on robotics. It was all
about path planning algorithms. The speaker, from Cal Tech, was clearly a
brilliant kid and the path planning algorithms were very clever. But, of
course, they are doomed to failure. The irony was that he called his talk
"Lessons from biology", suggesting that the study of biological locomotion
tells us that organisms use path planning algorithms (rather than control
of perception processes). I think the worship of output selection
algorithms is now a central tenet of the robotics "religion". It was
interesting to see, while watching this talk, that no matter how smart you
are, if you are using your smarts in the service of controlling for a
mythical higher level goal the ultimate result will be belief rather than
knowledge.

Best regards

Rick

···

--
Richard S. Marken, Ph.D.
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org

4

[From Bryan Thalhammer (2002.04.19.1300 CDT)]

Hi,

So, the problem you note below is that they made a model (albeit
inaccurate) of path planning, but then the *brilliant kid* suggested that
living organisms follow that model! It figures! That is how most
"religio-scientific" approaches go: Deductions based on false premises,
which are then used to drive the belief that the premises are correct.
Whew. Wrong kind of "looping structure."

Did the *brilliant kid*'s algorithms employ a closed continuous feedback
loop where action is fed back, or were they stepwise:
"move-calculate-move-calculate...." ? What about the problem of simulating
continous analog data vs. a digital approximation?

Was it Kuhn who said that people look at new phenoms through old paradigms,
and therefore cannot see the innovation? The AI critiques I have been
reading here and in other places suggest that AI research still can't shed
the notion of control of behavior.

Cheers,

Bryan

[Rick Marken (2002.04.19.0930)]

···

....
On a related note, I attended a talk here yesterday on robotics. It was all
about path planning algorithms. The speaker, from Cal Tech, was clearly a
brilliant kid and the path planning algorithms were very clever. But, of
course, they are doomed to failure. The irony was that he called his talk
"Lessons from biology", suggesting that the study of biological locomotion
tells us that organisms use path planning algorithms (rather than control
of perception processes). I think the worship of output selection
algorithms is now a central tenet of the robotics "religion". It was
interesting to see, while watching this talk, that no matter how smart you
are, if you are using your smarts in the service of controlling for a
mythical higher level goal the ultimate result will be belief rather than
knowledge.

Best regards

Rick

5

[From Rick Marken (2002.04.19.1650)]

Bryan Thalhammer (2002.04.19.1300 CDT)

So, the problem you note below is that they made a model (albeit
inaccurate) of path planning, but then the *brilliant kid* suggested that
living organisms follow that model!

Not quite. The lecture was about path planning algorithms. The algorithms work
fine (as far as I know) in a disturbance free environment. There was some talk
about "lessons from biology" but nothing specific.

Did the *brilliant kid*'s algorithms employ a closed continuous feedback
loop where action is fed back, or were they stepwise:

Path planning algorithms are designed to produce behavior in an "open loop"
manner. The algorithms may be revised periodically based on newly acquired data
-- so they do operate in a very slow control loop -- but the idea is to
calculate, in advance, the actions that will get you to your goal. This, of
course, is a whole different outlook on life compared to the "control of input"
view.

The tragedy, from my point of view, was knowing that this obviously very bright
person is going to do all this great work on output calculation algorithms in
an effort to get robots to behave like living systems when the behavior of real
living systems is based on a control of input rather than a control of output
architecture.

The reviews of my "Fielder's Choice" paper and Cziko's "Things We Do" and the
robotics talk (all in one week) show pretty clearly that people don't think
there is any significant difference between a _control of output_ and a
_control of input_ view of behavioral organization. What these people are doing
seems to works well enough for them and so they are not interested in
demonstrations that what they are doing can be done better. I think this is why
Bill's pendulum balancing model and my fly ball catching model, both of which
do things better and more simply than existing models, have been of little
interest to conventional behavioral scientists. Conventional behavioral science
models do things "well enough", from their developers' perspective anyway, so
there is no interest in change. I sure hope Abbott's robotic implementation of
Kennaway's computer bug model fares better (in terms of convincing behavioral
scientists of the value of a PCT model of behavior) than have the PCT demos to
date.

Best

Rick

···

--
Richard S. Marken, Ph.D.
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org