Rosen cf. Kauffman

[Howard Pattee <***>]

At 11:23 AM 1/24/05 +0200, Ayten wrote:
[snip]
> After having written these passages, I asked myself the question of how
> close Kauffmann and Rosen in their views on this question of What is Life? I
> also further asked myself if the main actors of this list as Tim, John K,
> and Judith consider making a comparison of these two minds and show for us
> similarities and dissimilarities .  .  .

HP: Here are some of my own comparisons between the minds of Rosen and 
Stuart Kauffman. I have known them both personally and professionally for 
many years. But first, since in the past Tim and Judith have misinterpreted 
some of my critical opinions as a personal attack, I must emphasize that 
any critical comments are only my academic opinion and do not in any way 
detract from my close personal friendship and high regard for them both.

John M's intuition is correct. Here are two different types of scientific 
minds that start from different backgrounds and work from divergent points 
of view, ending up with completely different types of models. Kauffman 
began as an MD but was attracted to experimental developmental biology. 
Then, largely because of a single computer program, and with no significant 
mathematical background, he morphed into a theoretical biologist focusing 
on self-organization, origin of life, non-Darwinian evolution, and 
development.

Rosen began studying wet biology in high school, but his instincts and 
abilities were for mathematics rather than experimental biology. After 
studying math and physics, it was largely because of the influence of 
Rashevsky's relational models that he finally found that his deepest 
interest was in how our epistemology and our assumptions about scientific 
models influence our concept of life. As members of this list know, his 
basic conclusion, developed as an analogy with the creative open-ended 
"corpus of mathematics," was that (as in mathematics) formal (computable, 
purely syntactic) models are too "impoverished" to capture the creative 
open-ended novelties of life.

Kauffman's basic computer program (an elaboration of previous work by Ashby 
and Walker in England) is remarkable for three things, its simplicity, its 
initial randomness, and the richness of the interpretations he gives its 
results. Kauffman originally discovered these computed results entirely 
empirically and largely unexpectedly. The basic model is a randomly 
connected network of random (boolean function) nodes started with random 
initial conditions -- i.e., maximum disorder. The discrete states of the 
network are just the current values of the nodes, and the next state is 
completely deterministic (unless externally mutated). The well-known (and 
now mathematically understood) result for low connectivity is that the 
paths in the exponentially enormous state-space follow transient paths 
collapsing into relatively few, short, stable limit cycles. Higher 
connectivities lead to a finite analog of chaotic dynamics.

This behavior with various elaborations has been interpreted by Kauffman in 
many ways: 1) as self-organization from complete disorder, 2) as a first 
step in the origin of life, 3) as epigenetic canalization in development, 
4) as the self-organizing structures on which Darwinian natural selection 
can operate, but cannot override, and 5) by allowing variation in 
connectivities and by coupling nets, a model of meta-evolution where 
natural selection (coevolution) chooses the "edge of chaos" as the most 
adaptive condition.

It is fair to say that Rosen and Kauffman never formed more than a civil 
relationship that is usual at professional meetings. Kauffman has much more 
personal ambition than did Rosen and achieved a notoriety that in my 
opinion (and Rosen's and many others) is not justified by hard evidence 
supporting his imposing claims for his models. Rosen saw the behavior of 
random nets as an obvious example of the generic behavior of discrete 
dynamical systems. He saw Kauffman's many interpretations as simply 
illustrating that the same formalism can be encoded any way you like.

On the other hand, Kauffman has suggested biological observables that would 
allow empirical tests of his models. His books include biological evidence, 
and are written with conventional language that biologists understand well 
enough to critically discuss. Kauffman criticizes Rosen (as have I and 
others) for his imposing claims written in a language that few biologists 
understand, and that have as yet suggested no biological observables that 
could allow a verifiable model.

I don't think any of these criticisms, while probably true, are 
scientifically relevant. My current view is that Rosen and Kauffman are 
thinking on different levels. Rosen's thinking is primarily on an epistemic 
model. One might call it a general principle that any biological model must 
satisfy to answer a "Why?" type of question. After all, the modeling 
relation itself is a model. It is based on the Hertz condition that is not 
itself empirically verified except that a model must satisfy it to give us 
the answers we want.  An analogy is the symmetry principles of physical 
models. These principles are not empirically testable. They are epistemic 
conditions we discover we must place on empirical models to give us the 
types of explanation or answer to the questions we ask.

Kauffman, on the other hand, has a simple computer simulation that behaves 
by various interpretations like developing organisms and like an evolving 
population. If I had to say it in < 20 words, Kauffman has empirical models 
without a specific epistemology; Rosen has an epistemology without (as yet) 
specific empirical models.

Howard