Print

Print


*****  To join INSNA, visit http://www.insna.org  *****

Hi Moses and other Socnetters,

I tend to be a little shy about posting messages to
(very) large groups of people of whom I don't know,
but this time I thought maybe I might speak up...

Like you, Moses, I too am a fan of Liggett and Kemeny.
Last year at Sunbelt, Laszlo Gulyas and I
presented a paper where we have done just as you
propose below: namely couple computational simulations
with analytical mathematical results.
Skye Bender-de Moll commented how much he liked our
dual approach and suggested that we submit a paper
for JASSS on this. Hopefully Gulya and I will find
time to do this very soon (although both of us are
a little distracted with the end stages of PhD
dissertations these days). BTW to follow onto the
original thread of these mails: the multi-agent
simulation is coded in RePast.

That said, and here speaking as a mathematician,
I find there are indeed limits to what can be achieved
with an elegant analytical approach. For one, if you
want to consider something as simple as truly heterogeneous
agents, which can be an obvious thing to want to consider
with social models, an analytical approach can
quickly become untractable. In an extreme case, you
might have, say, a system of as many partial differential
equations to solve as the number of heterogenous agents times
the number of behaviors. To solve this analytically
simplifications would often need to be made. This is
what is done in the presentation last year at Sunbelt,
however, at best this can only be a benchmark.

It is still important, I agree with you wholly, because
it gives you some idea of the "ballpark," however
a system can behave quite differently with heterogeneity
as Gulya and I have found in our computational results.

As long as I am writing, I should probably also add a comment
to Marshall van Alstyne's recent post: Joan Walker and I have
done some work on separating out true state dependence
(agents directly influencing other agents' behavior)
from "spurious" state dependence (agents behaving the
same way because they happen to be influenced by the
same factors unobserved by the researcher). We will
present this at Sunbelt this year, so please stay tuned!

To add a comment to Ting Li's recent post: All of the above,
have an empirical application to transportation mode choice.
Last year's sunbelt presentation was also presented at
the 10th World Conference on Transportation Research
in a slightly different version, and this year's prospective
one will also be presented with a more domain-oriented
"take" at the Annual Transportation Research Board Meeting.

Finally a note regarding Barry Wellman's post on "networks
and geographical distance": Because the application is
transportation mode choice, and geography is so fundamentally
inherent to many questions regarding transportation, plus
there is a long tradition of dealing with geography in
transportation, the network considered is certainly a
"socio-spatial" one. People at the Sunbelt presentation
last year would remember the GIS graphics and the abstraction
of that in Payek, with many kind thanks to Vlado & Andrej!

Best regards,
Elenna Dugundji


| Elenna R. Dugundji
| AMIDSt - Amsterdam Institute for Metropolitan & International
Development Studies
| Department of Geography and Planning
| Faculty of Social and Behavioral Sciences
| University of Amsterdam
| Nieuwe Prinsengracht 130, 1018 VZ Amsterdam, Netherlands
| Tel: +31 20 525 4581 (work)
| Fax: +31 20 692 5813 (home)
| Email: [log in to unmask]





-----Original Message-----
From: Social Networks Discussion Forum [mailto:[log in to unmask]] On
Behalf Of Moses Boudourides
Sent: Saturday, December 18, 2004 11:41 AM
To: [log in to unmask]
Subject: Re: Social Networks Simulation Software


*****  To join INSNA, visit http://www.insna.org  *****

Yes indeed, Swarm, Repast and Jung (among a few others) are excellent
toolkits but, as a mathematician, I am tempted to say that computer
experimentations need to be accompanied with serious mathematical work,
in particular related to existence (convergence, stability,
path-dependence, ergodicity etc.) theorems of the ivariant measures or
the absorbing states or of any sort of fluctuations observed in the
corresponding computer simulations. From this point of view, there is
plenty of work in the area of networks and game theory (for instance,
Bala & Goyal, Skyrms & Pemantle, Jackson & coworkers etc.) but not so
much in the proper mathematical theory of social networks. Of course, a
striking exception is the recent work of Bonacich & Liggett (on exchange
networks) and Liggett & coworkers (on interacting particle systems -
during the last 30 years). My point is that social networks have been
always an amazing paradigm of applied mathematics research and their
study needs to dissemenate further inside the international mathematical
community in order to get more advances both in formal sociological
theories and related mathematical fields. The roads in social research
opened up by such brilliant mathematicians as Harary, Kemeny, Johnsen
(just to mention a few names) should not be neglected to be explored in
the present so-called information age.

--Moses Boudourides


On Fri, 17 Dec 2004, Meredith Rolfe wrote:

> *****  To join INSNA, visit http://www.insna.org  *****
>
> David is quite right about repast. It is worth noting that if you can
> use Repast and JUNG together, you get even more functionality in terms

> of intermediate analysis of network properties. The analysis that
> Repast provides alone is relatively limited right now (although nets
> can be exported to UCInet or another preferred program) -- although I
> believe that improvements are planned for the near future (check out
> the list at repast.sourceforge.net).
>
> There is no built in method within Repast for linking node attributes
> to the creation or deletion of social ties during simulations, if you
> do decide to pursue this sort of simulation -- please stay in touch --

> I'm working on something similar right now.
>
> -- Meredith
>
> Quoting David Lazer <[log in to unmask]>:
>
> > *****  To join INSNA, visit http://www.insna.org  *****
> >
> > The Repast agent-based simulation package includes a network module
> > that, I believe, has all of these functionalities.
> >
> > david
> >
> >
> >
> >                       chatterjee.ariji
> >                       [log in to unmask]              To:
> > [log in to unmask]
> >                       Sent by:                 cc:
> >                       [log in to unmask]         Subject: [SOCNET]
Social
> > Networks Simulation Software
> >                       .EDU
> >
> >
> >                       12/17/2004 08:35
> >                       PM
> >                       Please respond
> >                       to
> >                       chatterjee.ariji
> >                       t
> >
> >
> >
> >
> >
> >
> > *****  To join INSNA, visit http://www.insna.org  *****
> >
> > I am looking for a user-friendly network-level simulation software
> > which:
> >
> > 1. allows random addition and deletion of nodes and ties
> > 2. can set rules for interaction, rewiring or node attributes 3.
> > depicts pictorially the evolution of the 'whole network' (not
> > attributes or relations) at different iteration levels 4. provides
> > easy access to the intermediate matrices at any iteration level in
> > order to analyze them
> >
> > any help in this regard is appreciated
> >
> > thank you
> >
> > arijit
> >
> >
> >
> > --
> > Arijit Chatterjee
> > Graduate Student
> > Department of Management and Organizations
> > The Pennsylvania State University
> >
> > Phone: 814.865.1263
> >
> > ____________________________________________________________________

_____________________________________________________________________
SOCNET is a service of INSNA, the professional association for social
network researchers (http://www.insna.org). To unsubscribe, send
an email message to [log in to unmask] containing the line
UNSUBSCRIBE SOCNET in the body of the message.