WikiLeaks Can Teach the Pentagon How to Predict the Future

July 18, 2012, 10:20am

For all the griping the U.S. government has done about WikiLeaks, there are actually some useful applications for the treasure trove of data the whistleblower organization has gotten its hands on. The WikiLeaks dumps offer mountains of data that offer the potential to crowdsource some of our national security challenges, ones that the Pentagon has been working on for years with limited success behind closed doors. The war in Afghanistan, of course, is on the top of that list.

A team of researchers armed with 77,000 logs from the Afghan War Diary recently decided to take a crack at some data driven analysis of the war. Using the details from the military logs about all kinds of instances from “preplanned military activity” to “spontaneous stop-and-search events,” the team from the University of Edinburgh figured out a way to successfully predict insurgent attacks. (They’ve just published the details in Proceedings of the National Academy of Sciences.) The software that they wrote basically looks at underlying trends in the data which ranges from 2004 to 2009 and ended up with “deeper insight in the conflict dynamics than simple descriptive methods by providing a spatially resolved map of the growth and volatility of the conflict.” In layman’s terms, they were able to guess where insurgent attacks would happen.

Videos by VICE

The key to the researchers’ efforts to predict what the ever-elusive insurgents might do was understanding volatility. Events in the more volatile parts of the country were more difficult to predict, but fortunately the researchers built in an audit layer that revealed how accurate each prediction was. In other words, the software expressed the likelihood of an attack, as well as the likelihood that it was correct in making that prediction.

In the less-volatile parts of the country the predictions were pretty good. In the Baghlan province, for instance, the researchers predicted that attacks would rise from 100 incidents in 2009 to 228 incidents in 2010. In reality, there were 222 insurgent actions in 2010, according to data from the Afghanistan NGO Safety Office. Not bad, right? And while it was harder to predict attacks in more-volatile regions (which, of course, is an inherent product of higher volatility), the WikiLeaks-released data still provided a fresh look at predicting the war.

This is obviously pretty ironic from the Pentagon’s point of view. For one, the brass has long been trying to contain the effects of the WikiLeaks releases and prevent another leak from happening in the future. As Wired points out, the Pentagon has also been trying (and failing) for years to come up with its own prediction models. Then along comes this massive batch of leaked data and a team of eggheads from Scotland, and suddenly a successful model is produced.

The worst part of it all, though, is that the research suggests that we’re more or less screwed in Afghanistan. The insurgents have kept up with the American troop surge and are showing no signs of slowing down. “Our findings seem to prove that the insurgency is self-sustaining,” the study’s lead author Guido Sanguinetti told the Los Angeles Times. “You may throw a large military offensive, but this doesn’t seem to disturb the system.” But hey, at least we’ll have an idea of how badly our offensive will fail before we launch it.