This algorithm could help predict ISIS’ future moves – Alex Lockie/Business Insider
A computer scientist has pioneered an artificial intelligence-driven method of modeling the behaviors of militant groups, and the Department of Defense is interested.
In a paper titled “Mining for Causal Relationships: A Data-Driven Study of the Islamic State,” presented at the 2015 Conference on Knowledge Discovery and Data Mining, a team led by Paulo Shakarian of Arizona State University used 2,200 individual data points on ISIS-related incidents from the Institute for the Study of War to build a descriptive model — an algorithm that models ISIS’s behavior.
Shakarian, who is a former Army officer, DARPA military fellow, and assistant professor at West Point, applied principals of computer science to turn the raw reports from ISW into a database which he could then analyze.
As the paper notes, “ISIS has displayed a high level of sophistication and discipline in its military operations.” Even so, it’s possible to glimpse patterns in their operations, and the paper found strong correlations between certain tactics.
For instance, a spike in car bombs in Baghdad was very often followed by ISIS attacks in northern Iraqi cities. Based on this relationship, the paper suggests that ISIS uses the car bombs to draw Iraqi security forces away from ISIS infantry pursuing other targets.
In the paper, Shakarain and his research partners identified two targets that ISIS seems to value especially highly: Balad and Baiji. Baiji is home to a major oil refinery and Balad is near an important government air base.
Shakarain described the paper as a “proof of concept,” telling Business Insider that “it’s not a really big data set, but it’s still significant.” Based on limited data from the only the second half of 2014, the paper focused only on modeling the past behavior of the elusive terror group.
“We came up with a description of their behavior, not any predictions,” explained Shakarian.
Shakarian said that there were people in the Department of Defense who “found the study very interesting.” And he thinks that his kind of computer science-driven research methods will become more accepted inside the Pentagon.
“It’s been revolutionary for DoD to see what you can do with this much data,” Shakarian told Business Insider. The ASU algorithm, updated with real-time data from the Pentagon, could be a powerful analytical tool.
Shakarian is excited at the prospect of applying his machine-learning methods to complex security situations around the world. “People are going to the battlefield with computers and recording data. We all forget how new this stuff is.” Shakarian noted that “It’s only been happening in the last 10-15 years where you have this much high-resolution data.”