Computational Social Sciences

I've been spending a lot of time recently talking with economists--of which the University of Chicago has quite a few. We're running a "Disciplinary Deep Dive" (3-D) look at computational economics this quarter, with lectures and discussions on a wide range of relevant topics.

Why economics? Bob Rosner, director of Argonne National Laboratory, spoke recently of the opportunities:

Imagine a world in which subjects like sociology or economics turn into disciplines that are extremely computationally driven. Those disciplines differ hugely from the physical and biological sciences. There you explore the world by doing experiments, whereas in the social sciences, you’re not going to manipulate societies, or individuals within a society, and so your freedom to experiment in social disciplines is sharply circumscribed.

But at the scale that one can compute today, you can imagine a method to conduct thought experiments that go far beyond speculation. The Sims game offers a toy version of what social scientists could do in a serious way: posit how individuals interact, posit attributes of individuals, and ask questions about the collective behavior that emerges from the individuals’ interaction. Computation could let researchers look at how a given energy solution is deployed from technical points of view—physics, chemistry, engineering, and so on—and from social, political, and economic perspectives: What level of investment is needed to introduce the technology? How hard is it? What are the economic consequences? The social implications?

My colleagues in economics would hasten to add that there is much more to computational economics than agent-based simulation. For example, see the program of the Institute in Computational Economics held at Argonne each summer.

In addition, there are interesting opportunities in data analysis, and in  social sciences disciplines other than economics. For example, in a collaboration between Argonne and the University of Chicago's Oriental Institute, simulation tools are being used to study Bronze Age Mesopotamian settlement systems.

However, while such models are interesting, its not clear to me yet how you validate their output. Perhaps in the case of Mesopotamian settlement systems, you can make predictions to be verified via subsequent excavation? Such questions raise challenging intellectual issues.

Comments

[Tim Freeman]

"SimCity for real"

http://www.eurekalert.org/pub_releases/2006-11/rcu-sfr111306.php

A large agent-based modelling experiment using grid computing and real data of the entire UK population (collated from multiple agencies and formats).

[Dave Berry]

Just to follow up Tim Freeman's comment, I'd like to add that we demonstrated the MOSES project (to give it's proper name) on the UK e-Science stand at SC06 last week. It got a fair bit of interest, including some from the media.
http://distributed-thinking.blogspot.com/2006/11/sim-city-for-real.html