Last year Emily Kennedy found herself sitting at an LAPD precinct. The 25-year-old wasn't in trouble—she was helping with a sting.
A group of girls sat in the office waiting to be questioned. Kennedy watched as a detective casually asked one a series of questions. It was clear this wasn't an interrogation; both the detective and the girl were just sitting and waiting. "Are you in high school?" the detective asked. "What's your favorite class? Do you like sports?" The girl answered, clearly scared and aware of the power asymmetry.
Kennedy watched from a desk as this transpired. This operation was working to find and rescue underage prostitutes, and the detective sitting there was trying to learn the truth about this girl's conditions.
The Los Angeles Police Department had spent months trying to needle its way into the city's prostitution rings, keeping track of potential suspects and victims using a piece of analytic software. Kennedy had developed this software. Kennedy was at the LAPD precinct because of her job; she helps protect victims of sex trafficking by writing software.
Traffic Jam, as her program is called, helps authorities track prostitution rings by using publicly available data. The project started when Kennedy was an undergrad at Carnegie Mellon University. Since graduation, the operation transformed into an entire company — Marinus Analytics — of which Kennedy is the founder and CEO.
I recently talked to Kennedy about Marinus and her digital human trafficking crusade. "Human trafficking is the crime du jour," she said. But really, this is something that Kennedy has been thinking about for a long time. She first learned about the issue in junior high. A youth leader at her church mentioned that he was going Cambodia to rescue girls in the country's Red Light District. "Just the concept [of trafficking] floored me," she told me. "Not only thinking that someone is being repeatedly raped, but that someone else is benefiting and making money off that."
The church leader's horrifying tales stuck with Kennedy. Throughout high school Kennedy kept tabs on the state of trafficking by reading numerous articles — "self-education" she called it. In college, she decided that her senior project would focus specifically on it.
The concept [of trafficking] floored me. Not only thinking that someone is being repeatedly raped, but that someone else is benefiting and making money off that.
In Kennedy's eyes, a major issue in combatting human trafficking is that the criminals have the Internet on their side. "Technology has empowered the work that pimps do to give them a wider audience," she said. "To more easily transport victims, recruit people, etc." Law enforcement, however, has yet to create tools that utilize the vast and open digital networks that are connecting pimps, johns, and victims. According to Kennedy, most agencies' problem is that they just don't have the time to invent new things.
In its early stages Traffic Jam was hypothesized as a way to organize the endlessly available data the world has on prostitution. Back in the day, print classifieds were known for advertising pictures of a scantily clad woman offering unspoken of services. Online destinations have followed suit with similarly vague and likely illegal ends. Turn to Craigslist alone, and you'll see post after post for any sort of sex you want. The question for Kennedy was how she could translate this raw data into something that could be used to help potential victims.
To tackle this problem, she decided to become familiar with the raw data. She read hundreds — if not thousands — of online sex-related posts, and it became clear how she could use them to fight the problem. She surmised that the pimps were the ones writing these ads: While the personally identifiable information — phone numbers, names, etc. — changed frequently, the general form of these postings remained constant. There were tells in each ad about who was writing them and which girls they were potentially trafficking.
Kennedy wanted to build software that analyzed all this data and grouped similar posts together. This would, ostensibly, make it possible to keep tabs on which pimp was doing what. It would also keep up to date information about whereabouts. If an ad changed phone numbers, it likely meant the pimp bought a new burner phone and Traffic Jam would be automatically updated. If the ads posted in a different location, this too would be noted. The idea was to use the information already available to have constant up-to-date information about who's who, and ultimately which victims are being trafficked by which pimps. When I asked her how detectives differentiate Traffic Jam's data between trafficked victims and sex workers, she said that they rely on their intuition and knowledge of the community they protect. She added that under US law, "any minors working in the sex industry are considered trafficking victims (compelled by force, fraud, or coercion), so that is very clean cut."
In order to be able to do this, Kennedy needed to know sex ads. She described her early college years; Her eyes glued to the computer trawling online escort ads to see what popped out to her. "I would literally just spend hours on these websites, looking at ads, getting a sense for what was the norm," she said. She began to pick up the nuances of every post, understand how a template was made, and get a feel for the different voices behind these ads. Though an unnerving subject matter to dive into, this years-long research helped Kennedy understand what telltale signs the software should be looking out for.
Traffic Jam began as an undergraduate honors project, but it didn't really set sail until after Kennedy graduated. In 2012, the Carnegie Mellon-based Auton lab decided to bring her on as a researcher. Auton focuses on discovering ways to analyze and exploit data to discover true, hidden patterns. The team looks at large datasets and uses tools like artificial intelligence to decipher any hidden meaning. Kennedy knew that Auton was the perfect place for her project. Others did too. That same year Traffic Jam received financial support from both the federal Defense Advanced Research Projects Agency (DARPA) and the National Science Foundation.
Now, three years later, her senior project has transformed from research project to full-fledged business. As soon as agencies got their hands on the first version of Traffic Jam, they began asking for more and more features. Then, Traffic Jam was considered "research grade." For the program to continue developing it became clear it needed a different sort of support structure, explained Kennedy. Last year Auton and Traffic Jam parted ways and Kennedy founded Marinus Analytics. With that she transformed from academic researcher to tech entrepreneur.
Marinus is based in Pittsburgh, and Kennedy splits her time between there and Northern California. Around 50 organizations currently use the software. They range from state and local agencies all the way up to the feds. The current statistic the company touts is that it's helped to assist and recover more than 120 victims of human trafficking. While that number may seem like a drop in the bucket, it's only the beginning for these types of automated programs working to pinpoint this kind of crime.
All the while Kennedy remains a bit shocked that her senior project got her to this point. She's definitely an oddity in the scene. While woman are the focus of prostitution cases, males make up a great deal of the detectives working on the cases. Since 2012 Kennedy has been talking at conferences and helping train other professions. She remarked about going to her first detectives conference and how attendees didn't quite know what to make of her. "I was always a curiosity," she said. "I don't look like a detective."
It's true: She's not a detective. She's the technological brains behind a tool detectives so desperately need. And while she has people's attention she'll continue going to conferences and stings, and being that smart and unassumingly powerful person in the background.