Re: [Swarm-Modelling] decision modelling

[Darren Schreiber]

I'm not aware of any work in this direction.

You should definitely describe more about what you have in mind in 
order to obtain more specific suggestions.

But, here is how I would approach the problem.  My understanding of 
bankruptcy (I actually have litigated a few bankruptcy cases and even 
written appellate briefs, but that was too long ago for me to remember 
much) is that the creditors have some competing interests in the 
company and are vying to get the court to reorganize the company in a 
way that is favorable for their interests.  So, I am thinking that you 
are trying to look at how model the competition among creditors?

Are you also looking to model the behavior and/or success of the 
resulting firm?

My advice on this problem (and it is relatively generic) is to stay as 
true to the traditional micro-models as possible (at least initially).  
In this case, your micro-agents will be the creditors.  What does your 
literature say about the decision calculus of the creditors?  What are 
its inputs and outputs?  What kinds of interactions are there among 
creditors?  How do decisions by one creditor effect the others?

While you can go ahead and create the micro-level model from scratch, I 
think it is tactically useful to be conservative and use existing 
results as much as possible.  This will allow you to justify your 
decision more easily and to justify departures from the traditional 
model if necessary.  It will also make it more familiar to reviewers 
and should make it easier to acquire data to validate the micro-level 
behaviors with.

You should also keep your model as simple as possible.  My tendency is 
to brainstorm the most complex and powerful version of a model that I 
can.  If I could make it super-flexible, what would it look like?  I 
spend a lot of time on big concept design so that I am going to have 
parameterizable, flexible code that will be useful for a number of 
development steps.

Then, with this huge Christmas list of features, I work the other 
direction.  What are the simplest features of the model that I could 
build?  If I had to show some progress tomorrow, what could I do in a 
day or two?  I write my development plan to go from zero to 0.00001 and 
then to go to 0.00002.  In contrast to my grand vision, this isn't 
ambitious, but it gets me lots of little victories and helps me to make 
sure my code is on track all along the way.  At each step, I test it 
and make sure I know what the code is doing and how it functions.

You will also want to think hard about why you want/need to use 
agent-based modeling.  I think that ABM is an incredibly useful 
paradigm, but I would urge you to make sure that you have the right 
tool for the job.  If I started playing opera music with a hammer, 
someone might be suspicious that I only know how to use a hammer and 
not much else.  What are the traditional ways of modeling creditor 
decisions?  What are they useful for?  What are their limitations?

You will need to describe more about why and what you want to model 
with ABM's for insight into whether it is a good tool for the project.  
But, I don't think there should any technical barrier here as far as 
the tool is concerned.

        Darren






On Tuesday, September 2, 2003, at 02:18  AM, 
<address@hidden> wrote:

> Hello,
>
> my name is Boris Heller. I am writing my dissertation at the 
> university of
> Essen, Germany.
> Recently I have been doing some research about agent-based-modelling. 
> I would
> like to use agent-based-modelling to simulate the decision of 
> creditors about
> the reorganization of a firm (e.g. "chapter 11"). Therefore I am very
> interested in further information about agent-based-modelling. Is 
> there any
> literature that I could use for this problem or do you know about any 
> similar
> projects in the past?
>
> In your opinion is it possible to create a programm for the purpose I
> discribed obove?
>
> I am looking forward to your answer.
>
> Sincerly yours
>
> Boris Heller
> _______________________________________________
> Modelling mailing list
> address@hidden
> http://www.swarm.org/mailman/listinfo/modelling