AP Photo
Could a computer model that predicts insurgent attacks help to keep our nation's soldiers out of harm's way?
updated 1/7/2010 12:08:56 PM ET 2010-01-07T17:08:56

War is hell, but scientists are learning more about the devil that governs major conflicts.

Attacks and casualties as a result of modern warfare can now be predicted using a mathematical model developed by scientists in the United States, the United Kingdom and Colombia. The research could save the lives of soldiers in the field or prevent future attacks.

"When you hear about these attacks on the news at night, they just sound haphazard," said Neil Johnson, a scientist at the University of Miami in Florida and a co-author on a recent paper in the journal Nature. "But when we looked at them, we found that there were universal relationships across conflicts."

To find these relationships, the scientists first spent years gathering data about nearly 55,000 separate acts of violence from nine different conflicts, from the deserts of Iraq to the mountains of Colombia.

The information included the date of the attack, how long the attack lasted, the number of casualties and other data gleaned from government and press reports. The violence ranged from a single assassination in Colombia to the deaths of more than 1,000 people during a stampede in Iraq.

No matter where the incidents took place or what motivated the attacks, when the data was plugged into a graph, the casualties resulting from each conflict aligned on or near the same line, known as the power law.

The power law governs everything from rhythms of stock market volatility to predicting earthquakes. These models do not provide a firm date and time that a stock will change (i.e. IBM stock will rise half a point on Dec. 15th), but rather give a more general idea of how a stock will perform (i.e. IBM stock will rise more than one point over the next month).

Applied to warfare, the power law roughly predicts the number of people that will be killed during a certain time period.

The researchers start with a simple assumption: It's easier for enemy combatants to kill 10 people than 100. In fact, their model predicts that it's 316 times easier; therefore, smaller attacks are more likely than larger ones because they are less difficult to execute.

Like earthquakes and stock volatility, violent attacks also typically occur in "bursts," said Johnson, such as multiple bombings in one city on a single day, and even at a certain time.

"We can't predict what will happen tomorrow," said Johnson. "But we can say what, statistically, will happen over a month."

To many strategists and commanders, the timing of bursts of attacks could suggest a coordinated effort.

To Johnson and his colleagues, however, this suggests that enemy combatants saw a good opportunity and seized on it at the same time with little or no communication among rival groups.

The team's model relies on three major variables: occasional long distance communication among groups, the hierarchy of leadership and organization, and competition among insurgent groups.

Using these three variables, the scientists simulated more than 10,000 virtual wars in the computer. Each war had its own battles with virtual soldiers and civilians ending up on virtual causality lists. Each battle was based on the power law model created by the scientists, which itself is based on real wars.

Once the computer finished creating its casualty list, the scientists compared their results with the documented casualties resulting from modern conflicts.

The simulated casualty lists from the virtual wars matched the actual numbers.

The model doesn't hold for all wars. Nine wars the scientists examined were insurgent or guerilla warfare. Two additional wars, the American Civil War and the Spanish Civil War, did not fit into the power law relationship.

The team's computer model has limitations, but it could save lives. "This won't necessarily help a commander in the field deal with the day-to-day," said Chris Danforth, a scientist at the University of Vermont who also studies insurgent behavior.

However, if the model predicts that a large attack is likely to happen soon, governments or military commanders could take steps to prevent its occurrence by more closely monitoring communications of enemy combatants. Officials may also potentially lessen the impact of an attack by moving civilians and soldiers away from likely targets.