Re: Rosen, Kauffman and compatibility

[Howard Pattee <***>]

Tim,
Kauffman interprets his NK model many ways. His first interpretation was
a genetic net with the nodes representing enzymatic Boolean switches
turning on and off other genes that make other enzymes, etc. The state
transition rule is determined by the Boolean functions of all the nodes
and all their connections (enzyme specificities) to other nodes, so I
would say the efficient cause arises entirely within the action and
configuration of the "closed" network.

The size of the state space and the number of possible state transition
rules is enormous (transcomputational in practice). What the model shows
is that starting from random states, random connections, and randomly
chosen Boolean functions, a network with low connectivity eventually
settles down to a relatively few, short, stable limit cycles. As
connectivity increases the behavior becomes pseudo-chaotic (not really
chaotic because it is a finite discrete system). 

It could be claimed that the state transition rule does not arise as an
explicit part of the model, but that is only because all the initial
conditions are random. I find the the model is really a long way from
real genes and molecules. That is one reason it is controversial. But in
any case, the model appears to me to be closed to efficient causation as
I understand it.

Howard

At 03:48 PM 1/27/05 -0500, you wrote:

Howard,
 
With regard to the Boolean networks and state-transition rules, I approach it from the view of Aristotelian questions - asking "why?". So, if I ask "why current state X?", I can answer:
* because prior state X₀. (material cause)
* because the particular type of transition dictated by the state-transition rule. (formal cause)
* because the transitioning effected by the state-transition rule. (efficient cause)
 
So, we have rule R effecting the change:       R: X₀ ® X
 
But if we ask "why R?", there is no answer within the above statement, aside from final cause ("because X"). But R will be composed of some formula which constitutes the rule. So, there is some formula R=f(a,b,c,...), or restated as something like f:(a,b,...) ® c. So, it will be possible to answer "why R?" with something like "because f" for efficient cause.
 
Now, either f represents something from the environment or it represents something from within the system under study. If it from the environment, then these state-transition rules describe impressed forces on the system; that is, if we ask "why f?", then the answer is "because the environment". If f is supposed to be from within the system, the question is: from where do they arise? The only answer in terms of efficient cause for f would be "because the simulator". Otherwise, f is unentailed (and therefore, so is R) within the model. There is a tacit interpretation that f, and thus R, have their origin somewhere within the system, but this tacit interpretation is extraneous to, and not part of, the model per se.
 
In my view, this situation reflects the Newtonian-style system-environment dualism, where a system is divided into an "inertial" portion and a separate "gravitational" portion which imposes forces on the inertial portion. In such a situation, the partitioning becomes problematic for answering " why?" questions as to how it is that such a system can remain a coherent and stable whole, because the partitioning makes the entailment picture incomplete. Further, that partitioning means that it also becomes not possible to determine if any such constructed model - whether it displays stable behavior (or "order") or not - is a model which is actually a model of a possible material realization into which both the inertial and gravitational aspects must be simultaneously instantiated.
 
Regards,
Tim
 

-----Original Message-----

From: ROSEN Forum [mailto:***]On Behalf Of Howard Pattee

Sent: Thursday, January 27, 2005 1:32 AM

To: ***

Subject: Re: Rosen, Kauffman and compatibility

Tim,

I agree and disagree. See my comments below.

At 08:54 PM 1/26/05 -0500, Tim wrote:

It seems to me that Kauffman's inquiries are at a different level of

analysis of the organism than Rosen's. Rosen's focus in "Life Itself" is on

organization of biological functions. Kauffman is focusing primarily on two

things: thermodynamics and chemistry.

HP: This is true for Investigations (a book with generally bad reviews). However, Kauffman's earlier work focuses on abstract organization at three levels, origin of life, evolution, and development. All these involve biological function. His basic NK network model was entirely physics-free  -- no thermodynamics or chemistry.

TIM: The idea of the collectively autocatalytic network is interesting, both as a

mechanism for organism's normal operation and as part of the origin-of-life

problem, although I simply don't know how widely or seriously the idea of

collectively autocatalytic networks occuring in organisms is taken.

HP: Almost all origin of life models depend on some type of catalytic cycles contained within some type of membrane (e.g., Eigen's, Hypercycles. See Steps Toward Life, Oxford UP, 1992)

TIM: It is not clear to me how they would map to Rosen's (M,R)-system (or the "closed

to efficient cause" facet of that model)

HP: Me neither. I do not yet see how the existence of "closure to efficient causation" can be empirically decided.

TIM: As to "order for free", which as best I can tell is based on his work with

Boolean networks, seems to be a far cry from Rosennean complexity.

HP: True, "order for free" is a loose metaphorical interpretation of network behavior, usually pre-biological. It does not reach the level of Rosen complexity.

TIM: The Boolean networks are entirely "Newtonian" in the sense that they consist of

"forceless" syntactic symbols pushed around by state-transition rules. The

regularities which arise and pass for "order" suffer from an inability to

sustain themselves; instead, the "forces" maintaining the "order" are

external to the network: they come from the simulator.

HP: On the contrary, the organization of limit cycles originates and maintains itself. There are no external forces necessary to sustain the order. Not only that, but the order can be stable to noise in the state transition rules, the node functions, and the node outputs.

TIM: Rosennean complex systems, on the other hand, will generically be rife with forces coming from within the system itself (hence the impredicativities).

HP: All the forces in Kauffman's NK models come from within the system. Since all the transition rules and initial conditions (and any noise) are random it is clear that the simulator has nothing to do with the stable order. Is randomness in states and state transitions a kind of impredicativity?

TIM: All those kinds of approaches which don't incorporate the dual nature of matter (what Rosen called the "inertial aspects" and the "gravitational aspects" in Essays) as

capable of both imposing forces as well as having forces imposed on them are

doomed when they finally have to ask: "but how will it stay together!!??".

HP: I would guess that the cycle structures stay together because they are closed to efficient causation. By that I only mean that they form and maintain an autonomous stable order that does not need any external cause.

TIM: This seems to be what lead to Kauffman's use of the "edge of chaos" notion,

and his concern with the "work cycle" as a means of avoiding a systemic

deterioration into thermodynamics equilibrium.

HP: The "edge of chaos" metaphor arose from later evolution experiments in a fitness landscape with coupled NK networks essentially competing in a game-theoretic model. This is physics-free and has nothing directly to do with work cycles or thermodynamics. This is a complex model. See Kauffman and Sonke Johnsen, "Coevolution to the edge of chaos: Coupled fitness landscapes, poised states, and coevolutionary avalanches." J. Theoretical Biology 149, 467-505, 1991.

Howard