On Ian Lustick’s computer screen, masses of tiny multicolored squares represent thousands of people with different beliefs, economic status and ethnic backgrounds.
With a few deft keystrokes, Lustick makes events shift, causing the squares to change position. To the untrained eye, it looks like nothing more than green and red squares. It’s actually much more.
These shifting color blocks represent changes in people’s attitudes, actions or identities, and the process of square manipulation is known as “agent-based modeling.” The technique enables Lustick and his colleague, Dan Miodownik, a Ph.D. candidate in political science, not only to see what global political hotspots are likely to erupt in strife, but also to track and analyze the “hows” and “whys” of the process.
The Middle Eastern politics expert helped to develop the user-friendly program and uses it in three ways: In very abstract maps about general political mechanisms; in maps that blend several countries together to form a composite area; and in maps that incorporate ethnographic and public database information to duplicate a specific place with a specific political system, such as his program, “Virtual Pakistan.”
Q. What exactly is a political cascade?
A. A cascade is a pattern of sudden and widespread transformation. It’s a snowball effect, so it gathers steam as it moves. … A political cascade, as an example, would be revolutions in Eastern Europe in 1989 and 1990. Why, in a particular period of time or in a particular country, does a demonstration that starts on Monday [end up with] more people in the streets, until the government is toppled within a couple of months? That’s a political cascade. We’re interested in what determines the eruption of these cascades.
Q. How do you represent different interests and ethnic groups in programs such as Virtual Pakistan?
A. The colors represent particular identities. … It’s very complicated to make a map like this, but you get as much information as you can about how many of the workers in that rural area tend to have this but not that identity.
Q. How accurate are your predictions?
A. We’re living in a world that’s really one of many possible worlds. It’s not that we can live in any world, it’s not that anything could happen. When you make a prediction, you really can’t make a point prediction—which of the worlds that are possible will you be in. You can just make a prediction about which world situations are more likely. And that’s what we do.
Q. So how do you get to those predictions?
A. In order to get things as if history had been running for awhile we bias … what identity the world’s currently telling someone to be. For example, is it relatively good to be “red,” relatively bad to be “red?” Zero is neutral. We can say okay, that’s going to fluctuate randomly over time and every time we run it, there will be little differences. Those little differences could make a huge impact because of the complex way the world is organized. …[Pointing to concentrated yellow dots] It’s pretty well known that somewhere around there is where the nuclear materials are kept. …What we’re really interested in is why did we get all the breakdowns here [points to virtual nuclear explosion]. My initial hypothesis was that when fundamentalism was stronger—when this green color got very strong—that would be when you’d be more likely to [have nuclear meltdown]. But that turned out not to be true. There’s almost no relationship between fundamentalist strength and this kind of disastrous outcome.
Q. So what would make that kind of disastrous outcome happen?
A. What we found was very interesting. In the transition [from civilian control to military control], that’s where you get the high tension. A lot of people want to fix Pakistan—it’s a disaster waiting to happen—and they think the answer is a technocrat like [President of Pakistan Pervez] Musharraf who will expand the control of the civilian apparatus and democratize the state and replace the military. Pakistan’s best futures may actually be down that road. But what we’re suggesting is Pakistan’s worst futures are also likely to be down that road, because in the minority of cases when you actually do the things you need to do to fix Pakistan—which is take power from the military—you lose control of key facilities. So then the policymaker has to wonder: Should we just allow Pakistan to go on simmering but not melting down, or should we try to fix it but then run a slightly higher risk of really bad things happening?
Q. Is there a practical application for this yet?
A. I have received mailings from military people in different parts of the country who experiment on Virtual Pakistan and develop their own stuff with it. I have no idea what they do with it, really. But I do know that other work that I’ve done [on Middle East polity] did produce findings that went to decision-making levels [in the U.S.] and were combined with findings that were drawn from other kinds of sources. I can say that our predictions or advice was much more accurate than the advice or predictions that were chosen by these decision makers.
Q. What led you to this point?
A. In 1996, I became very interested in evolution and complexity theory, reading a book by Daniel Dennett called “Darwin’s Dangerous Idea.” It led me to believe that computers could be used to study problems in collective identity formation that were not able to be solved by normal methods. … I found an undergraduate in this area who ... worked with me and created a toy model in JAVA, which worked but crashed constantly. A graduate student here at Penn in math, Vladimir Dergachev, said, “This can be done much better.”… He was able to translate the theoretical requirements into a modeling system that does not require the user to be a programmer. And that distinguished what we’re doing from everyone else in the world.
Q. Will this change the field?
A. This is a technology that itself will cascade over the next 10 or 15 years because the number of problems that can’t be effectively addressed with available standard tools is vast.
Q. What else can you use this program for?
A. The opportunities for working with this are essentially limitless. When I did a course last spring, we had people from psychology studying weapons control, studying chimpanzees. We did a public piece on bioterrorism on smallpox, on secessionism. …
Q. Who else might theoretically be interested?
A. Advertisers would be very interested in this because it’s their dream to have a tip toward their particular product. Everybody just had to have one of the Cabbage Patch Dolls. There must be hundreds of these products that come out all the time, but once in a while, it cascades across the whole population of children. Why? It’s very difficult to answer that question because it’s hard to experiment with large populations. When it happens in one situation, and it doesn’t happen in another situation, you can usually find 14 things that are different. So how can you control all those variables? When you create a virtual world where agents are interacting with one another, you can control those variables so that only one at a time changes and see what makes the big difference in producing tipping or cascades.
For more information, visit Lustick’s web site, www.psych.upenn.edu/sacsec/abir/.
Originally published on November 18, 2004