[Date Prev][Date Next][Thread Prev][Thread Next]   [Date Index] [Thread Index] [Author Index] 

Re: anticipation

Howard,
Well, I'm thinking about this and reading RR's idea in Anticipatory
Systems.  At the moment I'm a little confused as to what Rosen was
saying. First, I havn't finished reading it all, so I'm probably missing
something. But what I have read seems to imply that a model is necessary
in order to relate phenotypic behavior to behavior one would count as
fit. These domains are related by selection, he says, but he  then
argues that the result is a model that the organism has for how to
behave in response to the environment, and that he calls anticipatory. I
have my doubts. I think you can get that from selection alright, but
that there need be no real-time model between behavior and the the
fitness landscape to get what most people consider Darwinian evolution.
Indeed I can see there is a model for behavior, and that model turns out
to be a good one for fitness too simply because it has been selected for
that correlation, but there is no modeling relation between phenotypic
behavior and fitness in what I read so far.

What I'm now looking for, and what would seem consistent with the rest
of his view, would be a statement about why, without a model relating
behavior to fitness, the adaptation/selection paradigm won't work, or
won't be at all effective or efficient. In that case, some real time
concept of fitness would be required.

It is possible that where he is going with this is that since there is
no real starting point for adaptation, prior selection indeed provides a
model of behavior that acts *as if* it were anticipatory. This
essentially gives the organism a concept of the future based on past
experience. Since it is a model of behavior, not the behavior itself, it
conforms with the modeling relation and thus is a complementarity
between model and realization. Novelty can thus be introduced into this
relationship from both sides, leading to functionally defined pathways
(the part I am interested in).

What disturbs me is that what I read so far is strictly mechanistic,
except that he states one mechanism, the behavioral dynamics, is not
reducible to the other, the genetic dynamics and that they must operate
and be modeled at different time scales, obvioiusly the genetic dynamics
being a slower process than the behavioral. So, it could be that this
non-reducibility between to mechanical systems makes their relationship
complex, but I'm still working through that.

Your question is difficult for me to answer clearly. I would like to
distinguish between a selected mechanism that only coincidently works in
future circumstances (by virtue of those circumstances being similar to
past selective ones) from an abstract concept of the future operative in
the organism. Where RR seems to take it is, first, that "abstraction" is
a model of reality and a simplification of it, not a broadening of
possibilities as might commonly be imagined. Hence a selected mechanism
is a model of past conditions and is an abstraction. Variability in that
model is tantamount to having various "thoughts" (my metaphorical
language now) about the past, which are applied to present behavior. All
that is fine, but what hangs me up is how that ever constitutes a
thought or abstraction about the future. It is a model for behavior
based only on the past, which in that sense I would say is syntactic
because its meaning (semantics) with regard to future events is not
input, or at least I havn't seen how yet. I think RR basically equates
semantics with one system inducing dynamics in another, i.e., the
relationship between systems. So, an obvious possibility is that the
semantics are input via selection, which seems to be what is being said,
but I'm not sure I find that satisfying because such meanings are still
not about the future or about the selective function that relates
behavior with fitness. What I am looking for is how fitness itself is
modeled by the organism, not just how it gets accidently represented.

I'm rambling. I'll comment after I've read and thought more.

Howard Pattee wrote:

John,

Would you say that anticipatory behavior is the same as predictive-model controlled behavior?

If so, would a purely syntactic model still allow anticipatory behavior?

Of course a purely syntactic (isolated) model would not be as versatile as an continuous input model.

Howard


From: John Kineman <***>
Date: 2004/03/30 Tue PM 05:58:37 EST
To: ***
Subject: Re: anticipation

Yes, the issue is how to argue the source of current anticipatory
"programs" in plants. I am much more comfortable talking about animals,
because not only do they have the hardcoded anticipatory behaviors
(female changes related with childbirth being perhaps an obvious
example) but they have also preserved the functionally abstract
abilities in the form of mental ability. My argument is generally that
its obvioius in the case of mental activity, its strongly indicated and
a major challenge in physics, hence it is most likely present in
systems  between these extremes, and it is thus most parsimonious to
assume some level of functional determination and then devise
experiements to test for it. Since it  would sharpen evolution and
explain how Darwin's gradualism can survive temporarily non-adaptive
intermediate stages (as Baldwin argued), it resolves a problem
(perceived by some) in the current theory. But most deny there is a
problem and we don't have good evidence collected to test that in most
domains.

JJK




--
© 2004 John J. Kineman
all rights reserved