Computer Program to Predict Civil War

WASHINGTON, Oct 26, 2001 (United Press International via COMTEX) -- Two U.S. researchers have devised a system they claim could predict civil wars and other political disasters by analyzing and categorizing news stories from the world's trouble spots.

"We still are working out (the system) but essentially we use a computer technique known as structural linguistic frame parsing, which analyzes complex forms of grammatical patterns," Craig Jenkins, a political sociologist at Ohio State University told United Press International in a telephone interview. "In effect, we use this as a device to code local newspapers or other news sources," he said.

Jenkins, who developed the method along with Doug Bond of Harvard University in Cambridge, Mass., told UPI the program lets researchers analyze events, such as outbreaks of civil unrest, protests and repressive actions by governments.

The "conflict barometer" attempts to give a weekly measure of unrest. Jenkins believes the system could have predicted past civil wars and could prove useful in helping officials decide how long to pursue Operation Enduring Freedom in Afghanistan.

It works by sorting and categorizing news stories, which are fed into an equation, which in turn produces a score known as a "conflict carrying capacity" or CCC. A score of zero means a country is in chaos. A 100 signifies stability.

Jenkins said nations tend to be in trouble when their CCC falls below 85 for about six to nine months.

Jenkins and Bond applied their system retroactively and found it capable of providing warnings six to nine months ahead of civil wars in Algeria and Sri Lanka.

"But we are trying to determine more precisely what that warning time frame is," he said. "Often times before a major event, there is a news lag, which could mean guerilla warriors are building an underground network, for example, or there is political bargaining behind the scenes."

Before the Sept. 11 attacks on the World Trade Center and Pentagon, the United States scored near 100 while Sri Lanka, for example, was scoring 60 or below.

The system's efficacy depends in large part on the quality and quantity of it receives. Often, as is the case in Afghanistan, reporters are not allowed into troubled areas and reports are gleaned from second-hand sources, such as refugees or traders.

One organization that has used the system, the Swiss Peace Foundation in Zurich, has augmented coverage by sending its own reporters to hot spots, Jenkins said.

Jenkins said the system shows Pakistan and Tajikistan are bordering on instability but he added applying the technique to Central Asian countries is generally problematic because news coverage is scarce.

While Jenkins emphasized the CCC's forecasting ability needs to be evaluated and honed through further studies, he thinks it could already help governments spot crumbling regimes.

Part of honing the system means tweaking its linguistic skills.

"The system has trouble with metaphors, for example, when a report refers to a business takeover as a coup," said Jenkins. "Also some writers don't use certain diction."

"People have been doing political risk analysis for some time and on the face of it this doesn't sound very different," said Rob Mortimer, a political scientist at Haverford, Pa. "The perennial question of whether "the close observer" wouldn't normally do as well as the computer formula seems worth asking. But all in all, the system is counting pertinent data and there's nothing inherently unworkable about it in my view."

By Kelly Hearn

Copyright 2001 by United Press International.

© 1996-2008 Virtual Research Associates, Inc. All rights reserved.