The Islamic State, or ISIS, has recruited tens of thousands of people to its cause, mostly through social media. What has made the group’s online propaganda so successful, and what can be done to combat it?
To find answers, a team of researchers led by Columbia political scientist Tamar Mitts ’17GSAS recently turned to big data, analyzing the impact that some twenty-six thousand ISIS recruitment messages had on the attitudes of more than two hundred thousand of the group’s Twitter followers over a two-year period. Specifically, the researchers sought to identify characteristics of ISIS video and audio messages that succeeded in pushing followers further down the path of radicalization — as shown by the levels of enthusiasm people expressed for ISIS before and after encountering certain posts.
“Knowing what kinds of extremist propaganda resonate with people is crucial for designing effective counter-messaging,” says Mitts, an assistant professor at the School of International and Public Affairs and a member of Columbia’s Data Science Institute. “But until now, no one had examined this issue in a rigorous manner.”
Her team’s findings are powerful and surprising. The data set reveals, for example, that videos of beheadings and other atrocities were unpopular with all but the group’s most fanatical supporters. Instead, the vast majority of ISIS’s Twitter followers were inspired by propaganda emphasizing the personal benefits that people could supposedly enjoy by joining the group — benefits like getting a free home in the caliphate, finding a spouse, and feeling camaraderie with fellow fighters. Every time ISIS released messages extolling such “material, spiritual, and social” perks of jihadism, the researchers write, the Internet lit up with tweets declaring people’s intentions to join the group.
“You could consider this ISIS’s version of ‘positive messaging,’” Mitts says. “It spoke to people’s basic needs rather than any hunger for violence.”
Mitts says her research has implications for combating radical rhetoric both online and offline. She points out that while Internet companies have become more adept at rapidly identifying and removing violent content from their sites, they have struggled to detect other types of extremist propaganda in a timely manner. That is because the image-recognition software the companies use to spot dangerous content is better equipped to identify scenes of graphic violence featuring images of knives, guns, blood, flames, or military fatigues.
“The nonviolent propaganda gets taken down eventually, but not before lots of people have already seen it,” she says. “This is a big problem, because the nonviolent content is feeding ISIS’s pipeline of new recruits. Internet companies need to figure out a way to detect it faster, either by developing better AI programs or hiring more human censors.”
At the same time, Mitts says, people involved in community-based outreach efforts to combat extremism ought to speak frankly to vulnerable young people about the falsehoods ISIS peddles about the daily lives of recruits.
“Ample media reports show that life for ISIS members isn’t anything like what the group promises it will be,” she says. “Young people need to hear the truth so they don’t end up making irreparable mistakes out of desperation.”