modelling theory!

[Sven N. Thommesen]

After all those nerdy programming questions, let me throw out for
discussion the following question relating to agent based modelling as such:

Epstein and Axtell suggest that we are engaged in 'generative science': we
are looking for micro-specifications (i.e. agent behaviors) that are
capable of generating a given macro behavior. If we are lucky enough to
have more than one micro specification that is generative in this sense,
they suggest we employ Occam's razor and go for the 'simplest' one.

Now, can we not in principle find a multitude of micro-specifications that
capable of generating a particular macro outcome? If so, what we have is a
'sufficient' model but definitely not a 'necessary' one. And we cannot say
that because a given micro specification is capable of generating a given
macro behavior this must be the micro behavior that generates that macro
behavior in the real world (e.g. in the economy).

So, should we not also require that our behavior rules at the micro level
be 'plausible' in some sense? (E.g. conforming to a priory theory, or
conforming to observed behavior.) And how does this question depend on
whether we are modelling an artificial chemistry or a stock market?

Have at it, all you repressed theorists out there!

--Sven