After spending a year going through pro-ISIS groups on social media, researchers have come up with an algorithm they think might predict attacks. If only it were that easy.
Researchers analysed pro-ISIS posts and groups on VKontakte, the so-called Russian Facebook. They chose VKontakte partly because its members are diverse, including Chechen members that ISIS wants to recruit, and partly because (to their credit) social networks stateside shut down ISIS material pretty quickly. That’s good for our peace of mind, bad for a bunch of physicists looking for mentions of beheadings and pledges of allegiance to ISIS in multiple languages.
The result, detailed in a study published Thursday in Science, is an algorithm the researchers claim could have been used to predict the 2014 attack on the Syrian town of Kobani. That seems exciting, but since Kobani is the only one of many ISIS attacks that the algorithm has predicted, Faiza Patel at New York University’s Brennan Center for Justice, had this to say: “Frankly, at this point, I don’t think they’re predicting anything.”
The algorithm is probably more interesting from an insider-baseball methodology perspective than it is actually useful (To their credit, the researchers mostly acknowledge this). They focused on on small groups of pro-ISIS members that formed online instead of aggregating huge amounts of social-media data or tracking individual people. The idea is that quashing these small groups early can prevent the members from joining the larger ISIS organisation later on. It also makes sense that there might be a spike in the number of small groups formed online just before an attack. But that still doesn’t mean that the algorithm is going to keep us from being taken by surprise by terrorist attacks.
These researchers are hardly the first one to try to use maths to fight ISIS. U.S. intelligence agencies have noted that suggesting Facebook “like attacks” to fight the murderous terrorist group.