Kudos (Vancouver paper)

[from Jeff Vancouver 000114.1000]

Yesterday I got feedback from the action editor on the paper to which Rick
refers to in his Kudos post. It had been through two revise and resubmits
at the journal and only one issue seems to remain. One of the reviewers
thought that I had "cooked" the systems dynamics model to match the shown
participant's data. That was easy to counter, particularly if he was able
to download the model (which because of a server problem, he had not been
able to do before). Indeed, the point was quickly dismissed, with the
reviewer admitting his/her error.

Unfortunately, the actual model raised more flags for the reviewer. The
biggest being that the model was trivial. This prompted the action editor
to suggest dropping the model portion from the paper (something the reviewer
also suggested) given that the test for the controlled variable (TCV) and
other analysis (HLM) of the "experimental" data was sufficiently interesting
to merit publishing the paper without the model. He also listed options
regarding standing my ground, but after speaking with him, I am inclined to
drop the model.

The reasons for dropping are essentially two. First, the model IS trivial.
The point of the model was to introduce the modeling method to the study of
goal-striving (i.e., the study of human behavior). I wanted to keep it
simple so as not to loose the uninitiated. Of course the initiated (i.e.,
the reviewer) was completely unimpressed, and the uninitiated (the action
editor) did not see the value added over a verbal description. It seems to
me that what is needed to articulate the value added is substantial verbal
persuasion or a more complicated model that reveals and explains some
phenomenon difficult to understand without the model. In fact, it is this
latter situation that is the value of the modeling and my students and I are
working on studies like that. We have also almost finished a paper that
does the first thing (provides persuasive arguments for the value of the
modeling by explicitly distinguishing it from the type of modeling - and
control theory from the type of theorizing - typical in my literature).
This is the second reason for dropping the model from the current paper - I
will fight that battle elsewhere (my plan is to send it to a different
journal) for I still believe the uninitiated need a simple exemplar.

I would like to put a spin on this turn of events that might be non-obvious.
I think that the reviewer and editor thought the model was trivial because
they have no problem with the control theory explanation of goal-striving.
Hence, it showed what we already know. This is an argument I have been
making on this net for some time (which may explain my obviously biased and
wrongheaded interpretation of this turn of events). Anyway, that is my
take.

Finally, I could use some help on a few things. Because we have this second
paper that focuses on the model, model issues are still relevant. Besides
the triviality issue, the reviewer had some trouble with the units in the
model. He/she did not understand the neural units as a unit for
within-system signals (I know this was part of debate between Bill P. and
the Vensim guy with, last I saw, no resolution). I am inclined to stick to
my guns and, like Bill, continue to use neural units (although this time I
am going to head it off in the paper rather than hoping no one notices).
The second unit problem (I actually don't need help on the first) is that I
have a dimensionless level variable. I am just not clever enough to get
around this eloquently and was hoping a modeler could help. The model can
be downloaded from my web site
(http://oak.cats.ohiou.edu/~vancouve/research.html) if anyone was so
inclined and could help (Fred? Marc?).

The other thing I am wondering about is the Pressing article (1999,
Psychological Review). Recall, this is the article Mary Powers pointed to
as an example of the wrong-headedness of "cognitive" psychologists. Anyway,
I am about halfway through it. It is much more in Martin Taylor's (and
Rick's) areas of expertise (i.e., motor control) than mine, but I get the
impression it is not nearly as wrongheaded as Mary implied. In particular,
she did not understand Figure 1. When she brought it up I found it, looked
at Figure 1 and scratched my head as well, seeing Mary's point. However,
now that I have read more of it, I realize that Figure 1 is simply a general
depiction of the model Bill P. tested in his 1992 piece on cognitive control
systems (a chapter in Dynamic Psychological Systems). In particular, Figure
1 depicts an exogenous input to the system via a sensor and to the
comparator AS A GOAL. It turns out that it is simply depicting what Bill
did when he has "instructions" come from the environment to determine the
goal level.

The other thing Pressing does in Figure 1 is label the "controlled variable"
the "controlled system" with the external observer's signal of that labeled
"regulated variable(s)". Finally, he emphasizes aspects of the model that
are different from what Bill and Rick emphasize. But to me, through my
apparently fuzzy lens, what Pressing talks about is amazingly similar to
Bill's model, only Pressing uses a completely different citation list
(except Ashby). So has anyone else read the article? (If I missed this in
a previous thread, I apologize. Just point me in the right direction).

If there are substantial differences, it seems that opens the opportunity to
pit the models. Pitting models is very publishable.

Thanks for your time,

Jeff

P.S. Rick, apparently I am controlling for you not thinking I am the real
deal.

[From Rick Marken (2000.01.15.0930)]

Jeff Vancouver (000114.1000) --

Unfortunately, the actual model raised more flags for the reviewer.
The biggest being that the model was trivial. This prompted the
action editor to suggest dropping the model portion from the paper...
I am inclined to drop the model.

If that's what you have to do to get the paper published then I
concur. But I don't think the model is trivial. Simple, yes. An
over-simplicication, surely. But trivial, no.

I would like to put a spin on this turn of events that might be
non-obvious. I think that the reviewer and editor thought the
model was trivial because they have no problem with the control
theory explanation of goal-striving.

I think it's true that many reviewers have no problem with the
control theory explanation of goal-striving. In recent reviews of
my own _Controlled Variables_ paper, one of the main criticisms
was "we already know and accept control theory". But I think the
reviewers see your model as trivial because they don't really
know what control theory (as applied to living systems) is about:
control of perceptual representations of the external environment.
They're just thinking "control theory = closed loop = negative
feedback"; that is, they are just thinking in words. They don't
understand what control (goal striving) _is_ and they certainly
don't know how it works (control of perception). The fact that
your model is controlling a perception of the overall cost of the
schedules _rather than_ other possible representations of the
schedules is just not of interest to them. But if you can get the
paper published then _someone_ reading the paper might notice
what is going on and say to themselves "Ah ha. It's control
of perception; I'm going to start looking into that" and bingo,
the PCT science train has left the station.

The second unit problem...is that I have a dimensionless level
variable. I am just not clever enough to get around this
eloquently and was hoping a modeler could help.

I don't know what the problem is but I'll suggest an answer to what
your problem _might_ be. The vertical and lateral velocity variables
controlled in my baseball model are dimensionless (they are changes
in the tangent of an angle). So I talk about, say, vertical velocity
in terms of "units/sec". "Units" is a word that describes a
dimensionless measure. Does this help?

The other thing I am wondering about is the Pressing article
(1999, Psychological Review)....It is much more in Martin
Taylor's (and Rick's) areas of expertise (i.e., motor control)
than mine, but I get the impression it is not nearly as
wrongheaded as Mary implied.

...Finally, he emphasizes aspects of the model that are different
from what Bill and Rick emphasize. But to me, through my apparently
fuzzy lens, what Pressing talks about is amazingly similar to
Bill's model, only Pressing uses a completely different citation
list (except Ashby). So has anyone else read the article? ...

If there are substantial differences, it seems that opens the
opportunity to pit the models. Pitting models is very publishable.

I couldn't make heads or tails of Pressing. I don't have a copy of
the article handy. The only thing I _thought_ might be wrong with
Pressing's model (based on what I _could_ understand -- it is very
mathematical) is that it seems to be a "control of output" (rather
than a control of input) model. But if the equations describe a
control model (which they probably do) then Pressing's model is
really a control of input model (regardless of what Pressing _says_
about it) and Pressing (like 50 years of cyberneticists before him)
doesn't understand how a control system really works. If Pressing's
model is _not_ a control model -- if it is actually an output
generation model -- then I'm sure you know how to quickly expose
it's shortcomings (hint: disturb the "output").

P.S. Rick, apparently I am controlling for you not thinking I
am the real deal.

And failing;-)

If you keep publishing papers that not only talk about _controlled
variables_ but also describe research that tests to determine
whether or not various variables are controlled (that is, papers
that demonstrate the TCV) then you are going to have to work pretty
hard to to convince me that you are _not_ the real deal (in my
eyes, anyway). Telling me that you think there is value in the
work of people who are clearly clueless about perceptual control
isn't going to change my mind. The only way you can convince me
that you are _not_ the real deal is to stop studying percpetual
control and return to studying the effect of independent variables
on the behavior of groups of people.

Best

Rick

ยทยทยท

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