The Telegraph
Since 1st March, 1999
Email This Page
Predicting violence
The tool allows scientists to describe collective behaviour by neglecting the behaviour of individuals

The trigger may have been a torched train, and police and political failure to protect innocent people may have fuelled the Gujarat riots of 2002. But with a bit of number juggling, physicist Yaneer Bar-Yam in Cambridge, Massachusetts, may have caught a glimpse of just how and why the trigger spawned and sustained a wave of violence that lasted weeks.

Bar-Yam of the New England Complex Systems Institute and his colleagues at the Massachusetts Institute of Technology (MIT) and Brandeis University — both in the Boston area — have developed a mathematical model to predict where ethnic violence is likely to erupt and where it is likely to be more widespread. It’s a tool to forecast how groups of people will behave when they encounter triggers for violence.

In a recent report in Science, the researchers claim they have used their model to test predictions of violence between ethnic or religious groups in India and the former Yugoslavia. They fed India’s census data — specifically, the distribution of ethno-cultural groups across the nation’s districts — into the mathematical model, basically a special computer program.

The model said that Kashmir, the northeast and Jharkhand were “conflict-prone” zones. The predicted zones overlapped well with actual sites of violence. “This study indicates which regions may run into trouble and how to avoid conflict,” said Bar-Yam, who studied physics at MIT before beginning to explore how mathematics can be used to predict collective behaviour of complex systems.

The mathematics is based on the assumption that violence does not arise in highly mixed regions or when groups are separate. In highly mixed regions, groups of the same type are not large enough to sway collective behaviour toward claiming any particular public space. On the other hand, well-segregated groups are protected by clear boundaries that identify their space.

Partial separation, with poorly defined boundaries, foster conflict, while well-defined borders help reduce ethnic tension. “Our research shows that violence takes place when an ethnic group is large enough to impose cultural norms on public spaces, but not large enough to prevent those norms from being broken,” said May Lim, a team member at Brandeis University in Waltham, Massachusetts.

The scientists concede that social sciences research has identified several factors that contribute to violence. They range from the oppression of minorities, economic grievances and weak social ties to national ethnic diversity, territorial claims and international influences. Bar-Yam and his colleagues have introduced a new element which, they believe, has a predictive power.

Their work suggests the risk of violence is low when ethnic or culturally diverse groups are clustered either in areas less than 10km or in areas larger than 100km, the figures representing a circle diameter.

When populations are very well mixed, the boundaries are irrelevant, and when populations are very well separated, the boundaries are sharp. But when there is poor mixing, the boundaries are hazy and risk of violence is greater, Bar-Yam said.

“But this is not a prescriptive study,” Bar-Yam cautions. The model does not put forward any specific solutions to ethnic violence in regions. All it does is suggest that either one of two distinct strategies — imposing greater mixing of populations or finding ways to geographically separate population groups with ethnic or religious differences may both lead to situations that lower the risk of violence.

Researchers point out that the idea that sharp boundaries between groups lower the risk of ethnic violence is not new. “The whole argument in defence of partitions is based on this hunch,” said Ashutosh Varshney, professor of political science at the University of Michigan, US, and author of Ethnic Conflict and Civic Life: Hindus and Muslims in India.

There has also been a feeling in some quarters that perhaps a degree of mixing makes peace more probable, not mixing per se. “But there is no clear theoretical argument let alone a formal model on these lines, Varshney said.

Bar-Yam cites the example of Singapore which has actively pursued policies of imposed mixing to avoid sectarianism. But there have been moments in history when groups of people were separated. Whether authorities advocate mixing or explore the possibilities of separating population groups would depend on specific situations, he said.

While Gujarat did not figure in their paper, the model did predict that the state would have a higher level of violence than other regions in India, Bar-Yam told KnowHow. “It is interesting that this coincides with the observation that Gujarat tends to have episodic violence triggered by individual events rather than ongoing violence found for the areas that are predicted to have higher levels,” he said.

The violence in Gujarat may have urban aspects of partially isolated populations that are just below the critical sizes the model predicted were susceptible to violence. The scientists believe the model would need to be refined for forecasts at the urban level. But it is possible that even within cities the risk of violence could be lowered by either clear boundaries or very well mixed populations.

It may appear odd that scientists are trying to harness mathematics — always bristling with logic and rationality — to predict human behaviour which can be irrational at times. But researchers say the central problem is not rational or irrational behaviour. “This is a tool that allows scientists to describe collective behaviour by neglecting the behaviour of individuals,” said Bar-Yam.

“We have many serious social problems that are not possible to address using conventional scientific tools.”

Email This Page