January 22, 1998
Basic Geometry May Explain
omputers and humans may not think alike, but is there a way for the Spock-like logic of a computer program to reveal the basis for human behavior?
By PETER WAYNER
In the past, models have revealed startling insight into the basis for racism, economic mobility, pollution, segregation and many large scale social phenomena. This Thursday and Friday, the Brookings Institution in Washington is running a colloquium aimed at examining the limits of what these mechanized models of societies can reveal and debating what can be learned in the future.
Illustration: Nicole Schooley / CyberTimes
The researchers are, in essence, building computer simulations that treat the people in society like windup toys with completely predictable behavior. Then they start the computers running and watch what happens when the windup toys bump into each other. Ideally, the patterns will give some insight into how people behave.
Eliminating the problem may be harder than anyone might have imagined. The simulated society becomes quickly segregated if people only move to homes that have at least 13 percent of the neighbors from the same group. The implications of this research may be wide-ranging. Many Federal, state and local programs devote large amounts of money to pushing for an integrated society where groups intermingle. While there have been some successes, there have also been persistent and frustrating failures. In some cities, for instance, schools are even more segregated than before the system began expensive and time-consuming experiments with projects liked forced busing.
Some, like economist Glenn Loury, the founding director of Boston University's Institute on Race and Social Division, are beginning to question whether the money could be better spent on teachers, books and infrastructure instead of pushing for a desegregated system. Associate Justice Clarence Thomas has wondered why society implicitly assumes that all-black institutions are somehow inferior.
The hurdles may be more difficult than previously imagined, according to a CyberTimes segregation simulator, and may help explain why American cities are so segregated after so much work and money has gone into desegregating them. The people in the simulation quickly end up segregated, even if they only require as little as 13 percent of their neighbors to be from the same group. The simulator initially places people from two groups (red and blue) randomly around a checkerboard. At each turn, several people move to a new square at random, but only settle down if there are enough neighbors of the same group. The simulation lets you control the threshold that makes the decision. On a checkerboard, each square has a maximum of eight neighbors. If the threshold is set to zero, then people will move anywhere without regard to the neighborhood. If it is set to two, then at least 25 percent of their new neighbors (two out of eight) must come from the same group.
Test how different comfort levels, highways and other barrers, and population density affect the rate and extent of segregation by using the segregation simulator.
Segregation Simulator FAQ
The finer points of assigning mathematical values to integration comfort levels are explained in the simulator FAQ.
Running the simulator shows that people will quickly end up bunched with members of their same group even when the threshold is as low as 25 percent or even 12.5 percent. This may seem surprising at first, but it is a consequence of geometry. If people demand even one neighbor from the same group, they must choose a spot next to another member of the same group. When this decision is repeated over time by many people, the groups grow into larger clusters. As the smaller clusters disappear, the only viable choice is to move to the outskirts of one of the large clusters. Once a society settles into two large clusters, people can't move outside. This simulation gives the user several options for controlling the size of the neighborhood. It can either be the immediate eight squares or the twenty-four that are closest. While this changes the threshold, it does not change the ultimate effect. If people worry about being too alone, segregation follows.
Thomas Schelling, a professor at the University of Maryland in College Park, experimented with many different versions of the system, in some cases using a checkerboard and others with a computer. He said that the experiments have taught several lessons including just how low the threshold can be.
Assessing the ultimate meaning of simulations like this may be difficult, and the lessons they offer may be different from the messages that people want to hear. Robert Axtell, a researcher at the Brookings Institution, said, "It doesn't mean that people are racist, just because a system ends up segregated. People can have only modest preferences for local segregation, but at a system level, it can have a systematic effect that has a largely segregated outcome."
He suggested that some of the most interesting research in the next several years will surround stabilizing systems. For instance, he said, "Is there a way to change the tension across the boundary [between groups]? For instance, can you subsidize people to move or stay?"
Still, simulation is an imperfect science. These models leave out questions of taste and feelings of history. It is not clear, however, whether extra details make a model more accurate or merely muddy the results. Axtell pointed out that one of the maxims of modeling is that "You're finished with your modeling work not when there's nothing left to be added, but when there's nothing left to be taken away."
The CyberTimes simulator shows one simple fact: If two groups of people are placed in a city and people have only a moderate requirement about finding a home with neighbors from the same group, it's only a matter of time before the society will become segregated. The unanswered question is how to use this fact to society's advantage.
Following are links to the external Web sites mentioned in this article. These sites are not part of The New York Times on the Web, and The Times has no control over their content or availability. When you have finished visiting any of these sites, you will be able to return to this page by clicking on your Web browser's "Back" button or icon until this page reappears.
Peter Wayner at [email protected] welcomes your comments and suggestions.