Re: how would we define "kin"?

[Mark P. Line]

address@hidden wrote:
> 
> I would think kinship would be a more meaningful concept if it meant
> genetic similarity (your later comments) instead of knowledge of kinship
> (your earlier comments).  Because, I assume we are interested in kinship
> because of the behavior that an agent displays and the closeness of that
> behavior to the behavior of another agent.  The issue of whether the
> agents have knowledge of each other does not seem especially important to
> me.

I guess I don't know what kind of real agents you're attempting to model
here, then -- maybe you're not thinking of real-world agents at all, of
course.

Maybe there's somebody here who is a better ethologist than I who can
shed some light on the question of whether or not we might expect (real)
animals to behave differentially on the basis of actual genetic
similarity as opposed to membership in a breeding group (which should
correlate pretty well with genetic similarity, but not perfectly). I
would certainly not expect humans to do so -- I would expect humans to
behave differentially on the basis of their beliefs (knowledge) about
membership in a socially-constructed kingroup, not on the basis of
actual genetic similarity.


> If we are using genetic algorithms to generate new agents and two agents
> are closely related genetically, might we get retarded agents from
> incestuous mating?  Obviously not, but why not?

Probably because most of our agents come into existence as mature
adults. If we were to model ontogenetic processes, then I can imagine
that some of the effects of inbreeding might show up.


> What does biological mating do that genetic algorithms do not?

Biological mating produces a cell with a new genotype. Genetic
algorithms (usually) produce a new adult with a new genotype and a
(usually) deterministically derived phenotype.


> Might we gain something by being
> more closely aligned to biology in our mating paradigms?

Eeeeeeyup. :)


-- Mark

(Mark P. Line  --  Bellevue, Washington  --  <address@hidden>)



                  ==================================
   Swarm-Modelling is for discussion of Simulation and Modelling techniques
   esp. using Swarm.  For list administration needs (esp. [un]subscribing),
   please send a message to <address@hidden> with "help" in the
   body of the message.
                  ==================================