Wow!  I go out of the office for one day, and I have missed an exciting
discussion.  The issue that I connect  with in several of the postings,
first raised by Alan, is "what entities does your scientific practice
study, and thus assume exist?"  As Alan pointed out in a second posting,
this partly philosophical question is directly relevant to Bruce's emphasis
on the connection with practice.  Other than that thin thread of
connection, the following doesn't have much to say about the content of
Bruce's chapter (I sent my comments directly to Bruce last week).

MABS practice is based on (usually unstated) ontological assumptions, that
only individuals exist, and that macrosocial entities do not exist.  In
sociology, this stance is known as either "nominalism" or "individualism."
MABS goes further than this ontological individualism, and assumes the
stronger stance of methodological individualism as well: MABS explicitly
models AGENTS, but does not explicitly model social entities (groups,
institutions, norms/conventions/rules conceptualized as collective
entities, not simply shared psychological entities).  These collective
entities are not considered to be "real" and are not considered to be
capable of causal power: "epiphenomenal."  MABS methodology is
epiphenomenalist about the social.

This is not purely a philosophical issue, but is directly relevant to
current MABS practice.  Imagine if MABS accepted a different ontology:
sociological realism--collective phenomena are real.  If so, a MABS would
have to explicitly simulate not only the participating agents, but also the
collective phenomena as independently existing entities, with their own
causal powers.  (Macroeconomic models based on systems dynamics do
something like this.)  The fact that no MABS do this is evidence that the
field holds implicitly individualist ontological assumptions.  There are
some exceptions, as I noted in my prior posting: for example, some of
Carley's COT systems explicitly model organizations in addition to
individuals.  In fact, I believe that the simulations that are closest to
being used as models of naturally occuring social systems are the ones that
accept a degree of sociological realism.  We could argue about this...but
if so, this has implications for MABS methodology.


If we want to get into the history of the individualist assumptions of
MABS, we'd have to explore how MABS grew out of influence from economics
theory (the field is also individualist, in contrast to most sociology),
pre-existing methodologies of artificial intelligence and artificial life,
and culturally-ingrained European/American biases towards individualist
thinking.  

This brings me to Rosaria's emphasis on the connection with social science,
which I think is essential.  The issue of whether or not collective
phenomena are "real" (and whether they can have causal power over
individuals) has been central to 100+ years of sociology and economics.
The field of MABS, by taking a methodologically individualist approach, has
built several assumptions into its practice that are not widely accepted by
sociologists.  In fact, many sociologists would argue that their empirical
findings demonstrate that the individualist ontology/methodology of MABS
cannot succeed.  

Bhaskar's critical realism is a part of this ongoing debate.  Not only
critical realists, but all realists argue that abstract entities "exist"
and thus can have causal powers.  As Geoff notes, realism in this sense is
often opposed to positivism/empiricism, which in extreme forms claims that
causation itself doesn't "really exist" because all that we can observe are
regularities of succession.  But not many scientists are that extreme, even
those that deny the reality of one or another discipline's abstractions.
Psychologists may believe that individual mental states are real and can
cause individuals to take actions, even while they reject the reality of
sociological properties.  But a neurobiologist would, on the same grounds,
deny that mental states are real, and deny that they can cause anything.
The best contemporary arguments for "higher-level" realism are found in the
philosophy of mind (Fodor perhaps the most prominent).  Bhaskar's is only
one of many arguments for sociological realism.  The distinctive thing
about "critical" realism is the way that Bhaskar attempts to provide a
defense of sociological realism (using emergence arguments, among others).  


Rob's response about treating things as real if it is useful to do so: This
was the ontology of American pragmatism.  There is a methodological variant
of this, to argue that the categories of your science are not necessarily
"real," but it can nonetheless be useful to theorize as if they are real,
if it works as a useful shorthand.  This stance is called
"instrumentalism."  Note that MABS does not even do THIS with respect to
sociological entities, although doing so is consistent with an
individualist ontology (I think this is Carley's position).  In any case,
the problem here is that such entities are nothing more than analytic
tools, and as such they cannot have any causal power.  If
"institutionalized racism" or "the imperialist world system" or "being the
president of the world bank" is only a useful category for the analyst,
then it cannot exert any causal constraint over individuals.  If you want
to propose a theory that says that collective phenomena can constrain
individuals, you need a stronger, REALIST stance: that the entities really
exist.

To return, finally, to Bruce's focus on "actual practice": The everyday
scientific practice of MABS makes it clear that the field rejects any such
sociological realism; this has never been explicitly argued by MABS
researchers, but is one of the unquestioned assumptions that is built into
the practice of the field.  The ways to change such assumptions include (1)
to make them explicit and to question them, in a word, to "philosophize"
about them; (2) engage with research and theory in sociology that explores
the tension between sociological realism and methodological individualism;
(3) continue the "trial and error" of successive programming of
simulations, observing what fails and what works, and expanding the
methodology when it seems necessary.  

Number 3, although a necessary component of our practice, would give us a
rather weak discipline without being combined with (1) and (2).  There are
many hopeful signs that MABS researchers are increasingly beginning to
combine all three, including the recent postings to this list!

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



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