Re: Modeling the Emergence of Political Parties

[Darren Schreiber]

Yeah... I'm glad I'm not trying to get a dissertation yet.  The main point
of this paper is possibility.  It is possible to unify some of the classic
results of political science into one model.  Currently, the four results I
am trying to demonstrate are handled separately in political science.  No
single mathematical proof deals with more than one of these issues.

Of course, I have a number of extensions to the model, that I do think will
be truly new in the discipline.  The limitations of formal and game theory
are just too serious to allow a lot of the things we can do as agent-based
modelers.  But, the formal and game theorists are the Old Guard right now
and my strategy with this paper is to show that agent based modeling has
something to offer.  My thinking (a result of Lars-Erik Cederman's
tutelage) is that my model stands a better chance in the discipline
tactically if I show my chops in the classics and then start to improvise.

As to "proving" the results of the model, this I think is the critical
issue for the agent-based modeling paradigm.  This model only uses
randomness for the initial distribution of voter preferences.  If we wanted
to tediously implement the algorithm by hand (the life of a game theorist)
we could demonstrate the specific end of one run mathematically.  Each run
of the model is strictly deductive.

The interesting thing about the agent based modeling approach is that we
can broaden the set of initial assumptions and deduce the results
repeatedly.  The inductive argument from a wide set of parameter sweeps of
the individual deductions is one way to make an agent-based modeling claim.
Political scientists are used to deductive arguments (formal theory, game
theory) and inductive arguments (statistics, case study), the task of a
political scientist using agent-based modeling is to show how this
epistemology rests on familiar assumptions about the nature of proof.

I enjoy having a devil's advocate, reminds of my former career.  I think
Sven has asked some great questions that the modeling community needs to be
hashing internally so our answers to the outside world are coherent.



(On a less esoteric note)
The network stuff involves lots of recursive algorithms.  The coalitions
and voters are both subclasses of a generic agent type and can share the
majority of methods.  The coalition agents all have a list of their members
and can call to them.  The graphics were pretty tricky initially, but with
some perserverence, turned out to be a lot easier than using a raster.

Thanks for your thoughtful questions.

        Darren



>At 06:56 AM 10/21/99 -0700, Glen Ropella wrote:
>
>>Cool model, Darren!
>><snip>
>>Any chance you'll publish the source?
>>
>>glen
>
>Second the motion! I'd definitely like a look at how you coded the network
>aspect of this.
>
>And I'll chime in with a nasty little question for you (and for the rest of
>the list, for that matter): I notice you start out saying that your model
>'demonstrates' some classic results. This is an excellent beginning, but
>(most likely) not enough to get a Ph.D.
>
>Let me play devil's advocate here: Let's assume your model next is used to
>ask a novel question, and you get an interesting result. Now, what were you
>planning to say by way of justifying why the world should take your result
>seriously? 'The world' so far has become used to accepting only
>mathematical results that can be proven (while much of the justification
>for using agent based models is precisely that we can relax assumptions
>until the model is too complicated to ever prove results mathematically!)
>
>[I could take a somewhat cynical stance here and say that proper
>justifications don't matter, that what matters is whatever the current high
>mucky-mucks in the respective disciplines accept, so that all we Swarmers
>have to do is to keep churning out models that are interesting to *us*
>while we wait for the Old Guard to die off so *we* can become the high
>mucky-mucks. But, let's not be quite that cynical ...]
>
>Sven
>
>
>
>                  ==================================
>   Swarm-Modelling is for discussion of Simulation and Modelling techniques
>   esp. using Swarm.  For list administration needs (esp. [un]subscribing),
>   please send a message to <address@hidden> with "help" in the
>   body of the message.
>                  ==================================


_____________________________________________

                 Darren Schreiber
                  Attorney at Law
                 Graduate Student
             Political Science, UCLA
                address@hidden


                  ==================================
   Swarm-Modelling is for discussion of Simulation and Modelling techniques
   esp. using Swarm.  For list administration needs (esp. [un]subscribing),
   please send a message to <address@hidden> with "help" in the
   body of the message.
                  ==================================