On the Record

Q&A With Joshua Epstein on Computational Modeling

Joshua M. Epstein

CSED Director and Economic Studies Senior Fellow Joshua Epstein explains his breakthrough computational modeling work, with a focus on how agent-based modeling can help explain human behavior as well as make strides in the public health field.

Q: Why should, for instance, a high school math teacher in Omaha, Nebraska be interested in the work of the Center on Social and Economic Dynamics?

Josh Epstein:  A high school math teacher in Omaha, Nebraska should be interested in the work at the Center on Social and Economic Dynamics because it makes the study of social dynamics fascinating and vibrant and interactive and user-friendly. And it’ll make kids thrilled with questions like, how do epidemics spread, how do fads start and stop, what are the spatial dynamics of disease, of violence, of all sorts of things kids would find fascinating and engaging.

Agent-based modeling is a new, computational technique for modeling social systems in which we populate landscapes with artificial people. We basically build artificial societies where people differ from one another…they can be connected in networks, but they’re very diverse. They can have partial or even bad information (what we call bounded rationality); they use simple local rules in deciding how to behave. They move around and interact with neighbors, and the basic idea is that if we’re interested in some social phenomenon – like an epidemic or distribution of wealth or a settlement pattern – we try to grow it in an artificial society composed of individual agents. They can be young ones, old ones, sick ones, healthy ones, rich ones, poor ones. We can make this society look as realistic as we like and try to generate from the bottom up the large-scale, macroscopic phenomena that we care about.

Q: How does this shed light on the obesity epidemic?

Josh Epstein: It sheds light on the obesity epidemic by bringing to the fore the role of norms, of networks, of differences in group attitudes about body type, ideal weight and so on…and it provides a dynamic, generative explanation of the actual, historical obesity epidemic in the United States and its distribution by sex and by race. There’s quite a lot of data on the current state of obesity, but there’s no good explanation for why obesity and overweight grew in the United States and has the racial and sex distribution that it does. And we think we can generate that in collaboration with experts on obesity and minority health, in particular, for the US data using an agent-based model.

Q: Explain your expertise on the pandemic flu?

Josh Epstein: My expertise on pandemic flu derives mostly from my work on the NIH project, MIDAS: Models of Infectious Disease Agent Study. This actually won the 2006 Secretary of Health and Human Services award for public service and largely because we’ve done so much work on pandemic flu. I direct the global epidemic modeling activity at MIDAS and we have literally modeled pandemic flu on a planetary basis and have published the results in the Public Library of Science, a very competitive scientific journal. But when you look at the model of pandemic flu that we built, you see the entire globe, every air flight to and from all the 150 largest cities on the planet, and we give a very credible projection of how a human-to-human variant would transmit, and then, using the model we design containment strategies to mitigate that.

Q: What are the center’s goals for 2008?

Josh Epstein: The goals for this year are to continue our work in public health, on pandemic flu, other contagious diseases, MRSA, drug-resistant TB, and so on. To continue our work in chronic public health areas like smoking and obesity, there are two projects, both funded by the Brookings President: one on civil violence and its dynamics, and another on the dynamics of urban decline, with special emphasis on Cleveland. So those are some particular efforts that’ll be underway. The overarching goal is to make the Center on Social and Economic Dynamics the leading center for agent-based computational modeling applied to public policy questions. So, we want to be “the” place for this and we feel that we’re well on the way.