This is a really interesting issue -- the constraints put on social interaction when
considering modeling and simulation. Physical constraints play a major role in
defining social interaction. For example, consider queuing systems. People have
frequently modeled with queuing models and continue to do so. Why don't these models
contain detailed 'agent' information? The problem is that these patterns of human behavior
are often independent of what someone is "thinking" or how someone
reacts "socially". Queues/lines of people waiting
for service in banks, entertainment parks, or for ATM machines confirm this. In
this sense, anyone who does queuing models (where the packets are "people") is
doing agent-based modeling by default.
Consider, also, the case where we are concerned with reactions of people in
catastrophe situations. Groups of people in a building tend to follow "trained"
routes. Outliers can be modeled also, but there is little need to model cognition
since it may not play a significant role. The same holds for military situations,
where military procedure dictates how humans will act 90% of the time. Automobile
traffic modeling is another: modeling of traffic is done without considering
how an agent thinks. All of this is "agent based modeling" even though there is
are no cognitive or social effects being modeled. Modeling people interacting at
a party or at a meeting (where there are no significant constraints) is an entirely different matter.
It is not always clear when one needs to create a fully artificial human with simulated
cognition, and when it suffices to create only the usual queuing models and models
based on spatial, temporal and procedural (trained) constraints.
-paul
Bruce Edmonds wrote:
> Alan Penn wrote:
> > The problem with this example is that way that observed populations of real
> > people move through space has been found to be highly regular (the patterns
> > repeat from day to day) and predictable (they are dependent on a relatively
> > limited set of factors defining the physical configuration of the spatial
> > system being considered). Incidentally, the characteristics of spatial
> > variations in people movement are quite different to those for a gas, and
> > the analogy clearly does not hold. If anything people are more predictable
> > than gases where, for instance, turbulence is still on the edge of what can
> > be predicted.
>
> The point was not how easy the simulation problem was, but that it did
> *not* necessarily become easier with increasing constraints.
>
> > The main difference as I see it is that people are 'rational' at least to
> > some degree, where gas molecules are not. One piece of evidence to back up
> > this notion is that in spatial configurations that are highly unintelligble
> > (lacking in correlations between local and global configurational
> > parameters) the predictability of movement from configuration is also
> > reduced. In other words, as you reduce the potential for rational decisions
> > to be made by making the environment unintelligible the behaviour of the
> > population appears to become less predictable.
>
> It is not clear to me which side of the argument this 'fact' supports.
> Clearly when constraints are approach being total (i.e. only one action
> is possible) then this can increase the predictability (e.g. one
> signposted route to the lecture theatre), but that is not the general
> case.
>
> Prehaps a clearer case in support of my argument is Kaufmann's NK model
> of genetic search via bit-mutation, a hillclimber's route through the
> network will be far easier to predict with K (the number of mutual
> constraints on the bit string) small than when it is large.
>
> Regards.
>
> --------------------------------------------------
> Bruce Edmonds,
> Centre for Policy Modelling,
> Manchester Metropolitan University, Aytoun Bldg.,
> Aytoun St., Manchester, M1 3GH. UK.
> Tel: +44 161 247 6479 Fax: +44 161 247 6802
> http://www.cpm.mmu.ac.uk/~bruce
--
Dr. Paul A. Fishwick E-Mail: [log in to unmask]
Dept. of Computer & Info Phone & FAX: (352) 392-1414
Science and Engineering WWW: http://www.cise.ufl.edu/~fishwick
University of Florida (PGP Key available at above WWW address)
P. O. Box 116120
332 Bldg. CSE, Gainesville, FL 32611-6120
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|