>In the current (Sept. 2000) issue of Computational and Mathematical
>Organization Theory, I have a paper setting out an abstract, canonical
>framework relating the Carley models I think you have in mind to one of my
>models and the Stanford Virtual Design Team (VDT) model.  
>at http://www.cpm.mmu.ac.uk/cpmrep49.html)

I have read your paper and the Carley-Svoboda paper, and this is not one of
the Carley models I was referring to.  The Carley-Svoboda paper that you
reference, as well as your model, are individualist in that organizational
structure is not explicitly modelled, but is implicitly represented in each
agent's representations of its neighbors and relationships.  Carley has
developed other models that explicitly model organizational structure and
thus are "sociologically realist" in the (methodological) sense I described
in previous postings.  Good references here are Carley and Gasser's chapter
in the 1999 volume _Multiagent Systems_ or her introduction to the 1994
edited volume _Computational organization theory_.

>So if you see Carley as a social realist and me as a methodological
>individualist, either my paper is badly flawed or these are simply
>inappropriate distinctions for social simulation research as practiced by (at
>least) Carley and me.
>And this is actually the point I had in mind in my last posting:  The
>distinctions being made by Keith and others do not seem relevant to my
work or
>my understanding of the work of colleagues in the social simulation
community.

The problem is that your paper did not refer to any of the Carley models
that I was referring to, thus it is not germane to our discussion, except
as an exemplar of the methodologically individualist approach.  The type of
model that I am referring to--one that explicitly models social
structure--cannot be subsumed by yours nor the VDT model; such is the claim
of anti-methodological individualists.  If methodological individualism is
false, and structural phenomena cannot be reduced to individual-level
representations, then individualist MABS methodologies can have only
limited applicability to social modelling.  This is the "practical
implication" that you keep asking for, and that I claim sociology
potentially implies.

>Finally, I see social simulation as the computational investigation of the
>effects or consequences of interaction among social entities.  The social
>entities might for some purposes be whole organisations, for other purposes,
>departments or groups within organisations and for yet others representations
>of individual people might be appropriate.  

I agree completely with Edmonds' paper: my own position is not to claim
that social entities "exist" (I don't think that they do, and I agree that
for modelling purposes it is not important whether or not they "really" do)
but that our modelling has to proceed as if they do, due to the complexity
issues that Bruce raised in that paper.  Could you provide a reference to a
simulation that you have developed that explicitly models a social or
organizational entity?  If you could, you would convince me that you are
"pragmatically holist."  If you can't, then your own practice contradicts
your (and Bruce's) pragmatic claim.  My claim is that the great majority of
MABS are methodologically individualist: they do not explicitly model
social entities.  If you have a MABS simulation that does, you are in a
small minority.  I believe that MABS are individualist due to unquestioned
assumptions, not for pragmatic considerations of what would result in the
best model, and not as a result of studies of empirical sociological
research.  In this emphasis on pragmatic issues of methodology, and on
empirically validating models by reference to sociological data, I think
you and I and Bruce are on common ground.  

R. Keith Sawyer
Assistant Professor
Program in Social Thought and Analysis
Washington University
Campus Box 1183
St. Louis, MO  63130
314-935-8724



http://www.artsci.wustl.edu/~ksawyer



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%