From Greg Williams (930507 - 2)
Tom Bourbon (930507.0940)
Agreed, on the role of the arm model, but on the original issue, you are
saying that the way to convince our skeptical physiologist is for us to do
everything ourselves. This person is not much of a skeptic, if some of the
ideas I mentioned above have not come to mind.
Actually, I suspect that "this person" is much MORE skeptical than
you think -- of nonHPCT models as well as of HPCT models. But because
nonHPCT models tend to be more local/ad hoc than HPCT models, the
person might APPEAR more skeptical of the former (since they are
bolder attempts to cover a wider range of data than do the latter).
Another thing that comes in is preconceived notions of the greater
difficulty of "global" modeling. The skeptic might not want to bother
about global modeling because he/she thinks it is too big a bite to
chew at present. If so, then, yes, they'll say, go ahead and do it
yourselves, and we'll continue with our local modeling. Originally,
you had framed this discussion in terms of getting the skeptics
"interested," but I think it is more a problem of the perceived
difficulty of making and testing global models, whether HPCT or
nonHPCT.
···
-----
Bill Powers (930507.0800 MDT)
Alternatives to HPCT (some perhaps hierarchical themselves, but
not structured exactly as Bill Powers postulates) exist which
DO behave to various extents (the Beer bug behaviors appear a
lot closer to reality than do the Rick spreadsheet behaviors,
in my opinion, and there are lots of models in BEHAVIORAL
CYBERNETICS with rich behaviors).
The numbers in the model
tell the whole story that matters.
They tell HALF of A story. The other half is told by real-world data
(both at the behavioral and neurophysiological levels). And then, THE
"story that matters" is told by the correlations of the numbers of
both halves. I think that is all granted by many of the authors of
B.C. articles. Also, I think that impressive output isn't important to
many B.C. authors -- many of the papers show nothing but numbers or
graphs of variables vs. each other or time, and gloss over the details
of computer graphics (some, in truth, do seem to be, shall we say,
"display-impressed").
Would it be worth my while to subscribe to BEHAVIORAL
CYBERNETICS? How expensive is it? Unfortunately, Fort Lewis'
library doesn't have this kind of journal.
It is ca. $100 per year, I think. Back numbers extend ca. 30 years.
Maybe you could talk Ft. Louis into dropping SCIENCE and replacing it
with B.C. (:>>)
If you just want behaving bugs, like Beer's, you can sit down and
design one that will do what you want.
What do you do if you just want model bugs organized like real ones?
Or model humans organized like real ones?
I remember that at one time you claimed that Beer wasn't just
doing circuit design, but was using the properties of neurons
that are actually there. In fact, all the neurons and circuits he
used were simply made up to do what he imagined is actually done
(see p. 49 ff in _Intelligence as adaptive behavior_).
Isn't that what ALL modeling is? There are, of course, models which
include representations of more of the organizations of (a
circumscribed part of) reality, and models which capture less. Beer
certainly incorporated properties of neurons (like the membrane time
constant). He didn't, for example, try to model the membrane gating
processes involved in action potential generation. He made lots of
other simplifications, and DE NOVO constructions to fill in gaps, too.
I think that ALL MODELERS MAKE SIMPLIFICATIONS AND DE NOVO
CONSTRUCTIONS. The trick is in keeping in the "important" (always as
seen BY A PERSON) features and ignoring the "unimportant" ones, while
not letting your imagination get too wild.
I see Beer's bug as a valuable step toward more realistic models of
bugs. I see your HPCT modeling efforts as valuable steps toward more
realistic models of humans. It isn't difficult to find unrealistic
aspects of Beer's OR your own models. That's good if it leads to more
realistic models; that's bad if it is used to unconstructively deride
models for their lack of agreement with reality. It's especially bad
if it convinces anybody that, in modeling, simplifications necessarily
result in less "good" models.
My feeling is that when you consider the behavior you're trying
to explain carefully enough, there won't be a "range" of models
that "explain" it.
Given that unlimited data is not available to us for any aspect of
behavior, what aspect(s) of behavior do we indeed have ENOUGH data on
NOW, in your opinion, to limit the viable kinds of models to just one,
HAVING THE LEVEL OF DETAIL OF HPCT MODELS?
I do, but the question is whether a skeptical physiologist
would count these as more than examples of HPCT NOT failing,
with the question of what other kinds of models ALSO might not
fail left open.
This makes me wonder whether I really care what a skeptical
physiologist would think. It's hard enough to come up with one
specific and compelling model that fits the facts. Having some
person hanging over my shoulder saying "Yeah, but I'll bet there
are lots of other models that would work, too" makes me wonder
how my cup of coffee would look all over the jerk's face. If he
thinks there are other possibilities, let him produce them.
Otherwise, stuff a rag in his mouth.
I have tried to say in my reply above to Tom why the physiologist
might not think it worthwhile to work on other possibilities.
Regardless of his/her excuses, I think that the distinction between
valuing a model because it is the only one around and valuing a model
because it does a better job than others is important. Sometimes the
champions of PCT sound as I imagine the champions of epicycle
astronomy must have sounded before elliptical astronomy came along:
"We've got the ONLY model that works well, so give us some respect."
Well, you DO get SOME respect (as did the epicyclists). You'd get a
lot more if you won a "Battle of the Models" that wasn't just a
"Battle of the Model."
Perhaps what we need to do is issue a challenge for an open
competition to model a particular kind of behavior?
That sounds good to me. If we get to pick the behaviors that are
modeled, and the conditions under which that behavior has to
work, it will produce a lot of blessed silence.
But "blessed silence" will just lead to more respect at the current
level. Is that what you want?
Why not find a nonHPCT model that already predicts some set of
behaviors (e.g., fly-landing, as published in B.C. and J. COMPARATIVE
PHYSIOLOGY; a paper about this ongoing work was also published in
BIOSCIENCE a year or two ago -- your college library should have that
journal) and then make HPCT models that do a better job of matching
the data?
As ever,
Greg