Re: Rosennean "Cookbook"

[Tim Gwinn <***>]

Jack,

A paper by George Kampis, "A CAUSAL MODEL OF EVOLUTION", might be of
interest to you. It discusses some of the problems with evolutionary
simulations and their premises which limit their effectiveness. He then
proposes a general strategy for introducing a more realistic role for
causality into the process. This would seem to increase what I would call
"algorithmic novelty" into the simulation. It is painted in broad strokes,
but I thought it might be helpful:
http://www.jaist.ac.jp/%7Eg-kampis/SEAL02/A_Causal_Model_of_Evolution.htm

His homepage for this project is at:
http://www.jaist.ac.jp/%7Eg-kampis/

Regards,
Tim

> -----Original Message-----
> From: ROSEN Forum [mailto:*** Behalf Of Jack
> Park
> Sent: Monday, January 24, 2005 1:04 PM
> To: ***
> Subject: Re: Rosennean "Cookbook"
>
>
> I agree with the need to consider the impacts on the initial
> conditions.  I am wondering if that is, in fact, how my program works.
>
> Let me explain. Given an initial condition, consider that a node in a
> tree. As process rules fire on those conditions, new nodes are created.
> When several rules fire (several entailments assert their influence),
> branches form in the tree. Each new node becomes the equivalent of a new
> set of initial conditions. We call that an "envisionment".
>
> In some sense, the envisionment-building process -- I call it model
> building -- is not always practical. When Bob Trelease at UCLA was
> growing an envisionment of some behaviors under hyperbaric immunology
> contexts, the program ran all week, right up till the moment the janitor
> unplugged his computer to get a broom into a difficult space. Sigh...
>
> Nevertheless, in some sense, given that each node in a growing model
> represents a fresh set of initial conditions, the totality of the model
> reflects a histories (different histories for each branch in the tree)
> of the various entailments and their effects.
>
> It becomes necessary to inject metarules -- stopping rules -- which
> reflect a modeler's view of when any particular history has sufficiently
> reflected the needs of the model.
>
> When nodes are found to contain the same or similar information as other
> nodes somewhere in the graph, that branch is linked into the other
> (earlier) location in another branch and that history ends. This is
> usually a looping condition.  When we ran this program on the Krebs
> cycle, we discovered cycles which were slightly different from loops;
> the end of a cycle brings more information to the start than the initial
> loop started with. We documented that as cycles and spirals.
>
> Jack