Mining Data to Predict Attacks in Iraq

Physicist Sean Gourley thinks he may be able to model and predict violence in Iraq.

And it’s not just Iraq; Gourley, who works for the San Fransisco-based startup YouNoodle, has used his military side project to map the distribution of attacks in wars in Afghanistan, Colombia and Senegal as well. His finding: the casualties in all four of those conflicts, despite the chaos, fall into a precise mathematical distribution.

Using data from 130 publicly-available sources like newspapers, cable news, and NGO reports, Gourley and his team think they may have found the nature of war. By plotting the deaths in each conflict by time, place and frequency, Gourley discovered that the casualties fell along a well-clustered line of best fit, with a common slope of 2.5. That, he believes, means that his model could be used to predict the probability of attack in a given place.

The secret lies in the group dynamics of the enemy. At a certain point in a war, Gouley argues, insurgent forces must adapt one of two specific models, or they collapse; they either become weaker and more numerous, or stronger and more consolidated. The degree to which they achieve one extreme or another corresponds to the slope of the line Gouley has graphed. To find out what his data says about the success of the recent surge in Iraq, watch the TED talk below.