DSI Professors Mine Social Media to Curb Gang Violence

Desmond Patton

Researchers from the Data Science Institute have developed a way to flag aggressive social media posts from gang mambers that could potentially trigger violence. The researchers collaborate with social workers who can intervene with gangs to defuse the aggressive posts before violence erupts on the streets.

And now, thanks to a $548,000 grant from The Defense Advanced Research Projects Agency (DARPA), the researchers will also study how ISIS uses social media to recruit members. The grant will allow the researchers to build an aggression-indicator for DARPA that can flag aggressive posts by ISIS that might signal or prompt terrorist acts. DARPA is the Department of Defense agency that develops emerging technologies for the military.

Desmond Patton, the lead on the research team, said the project is expected to have various applications, both for identifying indicators of ISIS recruitment on Twitter and curbing gang violence that’s initiated on social media.

"We’ll use an interdisciplinary approach that mixes qualitative analysis from the social sciences with latest techniques from computer science,” said Patton, an affiliate of the Data Science Institute and professor of social work at Columbia. “We’ll compare differences and similarities between how gangs and ISIS use social media to recruit members with the goal of assessing how qualitative differences correlate to quantitative metrics."

Patton has already collected millions of tweets from gang members in Chicago, and intends to collect a similar amount of posts from ISIS. To help him sift through the deluge of data, he has partnered with Kathleen McKeown, the founding director of the Data Science Institute and a renowned expert in natural language processing (NLP). A form of artificial intelligence, NLP trains computers to understand spoken and written language.  It can filter through mountains of text and find patterns and themes. For this project, McKeown, a computer science professor at Columbia Engineering, will use NLP to flag aggressive social media posts from gangs and ISIS.

J.M. Berger, another member of the research team, is a leading expert in social-media analysis. He will design metrics that measure how gang members use social media as well as how ISIS uses it to radicalize recruits. Berger is a fellow with George Washington University’s Project on Extremism and has co-authored “The ISIS Twitter Census.” 

Patton is also a recognized expert in the field. For the last four years, he has studied what he refers to as Internet Banging — how gangs use social media to communicate and recruit. And he's developed a method to identify indicators of gang violence called "the Digital Urban Violence Analysis Approach."  

He became interested in Internet Banging while working as a social worker in Chicago, when he happened upon posts by Gakirah Barnes, a teenaged gangster with a murderous reputation. Barnes had used Twitter to express a softer side, though, especially about a friend who was killed by police. Patton realized then that online expressions of grief by gang members can be preludes to violence. For soon after her posts, Barnes was murdered by a rival gang member. He also realized that had he flagged her posts, social workers might have intervened, saved Barnes's life and prevented a spiral of retaliatory gang violence. That’s when he began collecting Twitter posts from gang members in Chicago. He now has millions of their posts, and thanks to the new grant, a chance to mine the data and deter violence.

“We mostly see gang members as bloody killers,” said Patton. “But violence is complex; people are complex. Gakira was violent but had a sensitive and humane side. Her posts gave us an opportunity to intervene and prevent her from being murdered and more blood from being spilled. But this research project will help us develop the tools we need to detect and prevent violence from gangs and from ISIS."

-- By Robert Florida



550 W. 120th St., Northwest Corner 1401, New York, NY 10027    212-854-5660
©2017 Columbia University