YAABIP (Yet Another Agent Based Intuition Pump :-)

[Tom Carter]

All -

   Here's another little NetLogo riff . . .  The question I got to  
wondering about is, when does it pay to specialize?  This wonderment  
was partially stimulated by a recent proposal by our new University  
President, who wants to break our big (23 departments) College of  
Arts Letters and Sciences into 4 smaller colleges, each much more  
specialized and more focused (i.e., Natural Sciences, Fine and  
Performing Arts, etc.).  I also recently caught up on this article in  
a recent issue of Science:

Phenotypic Diversity, Population Growth, and Information in  
Fluctuating Environments, Edo Kussell and Stanislas Leibler
  Science 23 September 2005; 309: 2075-2078; published online 25  
August 2005
http://www.sciencemag.org/cgi/reprint/309/5743/2075.pdf

(Cosma Shalizi has brief comments on this article, and some related  
information, here:
http://www.cscs.umich.edu/~crshalizi/weblog/2005/10/04/   )

   Anyway, suppose you live in a (relatively) rapidly changing  
environment?  When does it pay to specialize, and when might it be  
better to just "litter the fitness landscape" with diverse agents, so  
that there will always be at least a few agents who can take  
"immediate" advantage of the new environment?  One might think  
species, one "smart" and one "diverse," or groups of (somewhat  
similar) species.  One might think business entities, focusing on a  
single product niche, or broadly diverse to catch whatever the  
current "best market" might be.  One might think a broad,  
interdisciplinary researcher, or a highly specialized, focused  
researcher.

   Anyway, however you think about it, here's an outline of the  
model.  The environment is characterized by a single number (say,  
from 1 to 70).  There are two "species."  Each individual of each  
species has a (geno? pheno?)type, that may match more or less well  
with the environment.  The "Red" are "specialists" (or, the  
"learners").  During each time step, they adjust their type to be  
closer to the environment.  When their "type" gets close to that of  
the environment, they reproduce relatively rapidly.  (Their offspring  
may, or may not, be very similar to the parent.)  The "Blue" are  
diversifiers.  At any given time, there are Blue individuals across  
the whole spectrum of possible environments, so there are at least a  
few Blues ready to take greater advantage of whatever the new  
environment might be.  In this model, a Blue can always take at least  
some advantage of the environment, but reproduces faster the more  
closely it matches the environment.

   Anyway, have fun with it . . .

http://csustan.csustan.edu/~tom/SFI-CSSS/2005/Smart-Diverse/smart- 
diverse.html

tom

p.s.  I find it interesting that for the default set of parameters,  
it doesn't matter a whole lot whether the Red offspring closely match  
their parents or not . . . It's roughly a coin flip (:-) whether red  
or blue goes extinct for both "versions" of the reds ("Memes" on or  
off) . . .

[James Steiner]

On 11/10/05, Tom Carter <tom at astarte.csustan.edu> wrote:
>    Here's another little NetLogo riff . . .  The question I got to
> wondering about is, when does it pay to specialize? ....

Heh... this reminds me of role-playing character-developing games,
like Dungeon Siege or Dungeon Siege 2... Especially these two, since
the player control a party, rather than individual character, and the
characters have limited autonomy (game AI) to defend themselves and
select targets for attack. For example, in DS, one could succeed in
the game by developing a few characters (i.e. agents) with diverse
skills, like Archer + Sorcerer (combat-oriented/death magic) or
Warrior + Druid (nurturing-oriented/life magic). In the sequel, DS 2,
the skill-building system was altered, so that it hurts to diffuse
skills that way, and pays to specialize characters, to have a perfect
archer who can't even lift a sword, or a perfect magician who can't
even wear armor. Diversified characters can't be trusted to use the
most appropriate skill--a failing in the AI (or a deliberate effort to
provide the human player something to do)

~~James Steiner

P.S. Hi!