An AI tool developed by a Carnegie Mellon University graduate has helped law enforcement rescue hundreds of human trafficking victims across the United States and Canada. Emily Kennedy, who graduated in 2012, created Traffic Jam, software that scans online sex advertisements to identify victims, and founded her company Marinus Analytics in 2014.
Kennedy's interest in the issue began at 16 during a trip to Macedonia, where she saw children cleaning cars at a stoplight. A friend told her the children were trafficked by the Russian mob and would be punished if they didn't bring home enough money each day, she said Friday at the Women in the World event in New York City.
While studying humanities at Carnegie Mellon, Kennedy connected with machine learning experts and engineers who had designed technology for complex problems like epidemic detection and food safety. That collaboration, combined with her persistence, led to a tool that uses Amazon Web Services' facial recognition to match photos of suspected victims from missing person ads or social media with images in online sex ads. The data is messy, so Marinus Analytics has gathered online ads since 2011 to analyze characteristic misspellings and narrow down posts that may involve trafficking victims.
The software also includes image similarity search, which allows detectives to upload a photo of a victim taken in a hotel room and find other photos of victims in the same hotel or room. Speed is critical because traffickers move victims between cities to avoid detection.
Debunking Trafficking Myths
Traffic Jam has an 88 percent success rate, and its data has challenged common assumptions about trafficking. Kennedy said the software found that events like CES, the Oracle conference, and the Daytona 500 had more statistically significant increases in activity than the Super Bowl, contradicting the widely held belief that the Super Bowl drives more trafficking. She also criticized commonly used statistics from old studies that researchers themselves say should no longer be used.
Marinus Analytics partners with law enforcement, prosecutors' offices, and a few nonprofits, with Kennedy carefully vetting partners because the software is powerful and could be abused. The software now crawls data from 18 countries, with plans to expand to the United Kingdom and beyond. Kennedy and her team are also exploring applications for other types of predictive policing, such as the sale of guns and illegal drugs.
Kennedy, now 27, received the 2018 Mothers of Invention grant from Toyota, a sponsor of the Women in the World summit.
Editorial Notes