Subject: System Dynamics and Agent-Based Modeling

Posted by Dr. Niall Palfreyman on 7/25/2003
In Reply To:System Dynamics and Agent-Based Modeling Posted by David Koelle on 7/25/2003

Message:

The two software packages I use for class demonstrations are Stella and
Mathematica, because both are tools which are directly relevant in other ways
to the students I teach. What I'm saying here is that I wouldn't necessarily
recommend the use of Mathematica for agent-based simulation unless you have
other reasons for wanting to use it - it's a highly complex package which is
useful in many ways, but not worth it if you only want to use a small subset of
its functionality.

I use Stella for illustrating the way in which mathematical models arise from
real-life situations. Maths students often seem to think _solving_ equations is
all they need to be able to do, but Stella shows them how the _setting_up_ of
equations from real life also properly belongs to the skills of a
mathematician. Stella also of course provides immediate insight into the kinds
of solutions which one can expect from a set of dynamical equations.

What Stella often does not provide, however, is (gut-felt) insight into _what_
mechanisms interact to lead from the equations to the solutions. For this I
prefer to use spatial models which map these mechanisms in a way which can be
followed with excitement: "Will X happen? Or not? Or maybe even Y?" The
interaction of agents on a spatial board is powerfully exciting, and grabs the
attention of the students by the throat. For such simulations I use
Mathematica.

One example of such a simulation that I use is the Selfish/Altruistic model
reported by Uri Wilensky at the following website:
http://ccl.northwestern.edu/cm/EACH/ICCS.html
Here students become personally involved with the agents and their fates. For
several minutes it can look as though selfish agents will win the evolutionary
struggle, and then there's a roar of approval as the altruists finally come
into their own.

In summary, where I see a great advantage in agent-based models is in the way
they make explicit the developmental mechanisms of a dynamical system in a way
that appeals personally to students.

One final word about agent-based models in Mathematica. I have said that
Mathematica is a powerful system which possibly provides too much functionality
for some users. The flip side of this feature is that Mathematica offers
incredible flexibility in what your agents can do. In Mathematica there are
practically no constraints on the functionality of agents: they can be simple
or complex, and can develop or evolve even to the extent of making total
structural changes to themselves.

Niall Palfreyman.

Follow Ups:

System Dynamics and Agent-Based Modeling - Linda Booth Sweeney 7/25/2003