This Algorithm Helps Find Serial Killers But Australia Won't Use it

Serial killers very rarely target people at random. Usually their victims are of a similar age, gender, and ethnicity, and are murdered in comparable ways. And while these correlations might seem obvious in retrospect, detectives are usually looking at crimes one at a time over periods of years. Which is how parallels can go unseen and serial killers can slip beneath the radar.

Computers, however, are immune to the obfuscation of time and bureaucracy. A former journalist by the name of Thomas Hargrove realised this in the early 2000s, and set about creating an algorithm that could comb FBI records for patterns in unsolved homicide cases. In 2010 he ran his algorithm over a series of US state records, predicting that Gary, Indiana, was the likely home of a serial killer. Police later caught and prosecuted a local man named Darren Vann, affirming Hargrove’s calculations.

Today Thomas Hargrove is the founder and chairman of a non-profit called the Murder Accountability Project, which tracks unsolved homicides using his method of crunching police records. We got in touch with him to hear how the algorithm works, and to learn why Australia’s opaque system of police data retention prevents us from using his algorithm here.

VICE: Hey Thomas, I want to start with something I've wondered. That is, right now, in 2019, does your algorithm suggest there are active serial killers in the US?
Thomas Hargrove: Yes, we’re convinced there’s a killer or a number of killers in Chicago that have gotten away with about 50 strangulations. We also think there are a few active serial killers in Cleveland, Ohio. If you go to the murder clusters tab [on the website] and select county clusters, you’ll see a map with two big bubbles over Chicago and Cleveland. So we’re in conversation with detectives from both cities at the moment.

Okay, let’s go back to the beginning. How did this project start out? I understand you became interested in serial killers via your work as a journalist?
That’s right. So back in the early 2000s I was in Washington, DC, and I got assigned a story about sex workers. Now, some cities in America rigorously enforce anti-prostitution laws, while others don’t. It’s a haphazard system so in order to study it I contacted the database library at the University of Missouri and purchased a copy of the Uniform Crime Report for the year 2002. And it came from a database I’d never heard of before. It was called the Supplemental Homicide Report, which was a database of individual murder cases. Details from every single murder that had been reported to the FBI that year were included. And the very first thought I had when I opened the record was: is it possible to teach a computer to spot homicidal series and patterns? I desperately wanted to figure it out.

While you're here, check out a VICE true crime documentary:

Tell me more about the Supplementary Homicide Report. We don’t have anything like that here in Australia.
The Supplementary Homicide Report has been around for 40 years, but it’s got tremendous holes in it. It’s a voluntary reporting program so some states simply don’t report. We’ve actually started using Freedom of Information laws to compel all states to report. Right now, the states of Alabama and Florida don’t report to the FBI but they are reporting to us.

Right, so what happened next?
Well from 2004 to 2010 I begged my editors to let me try to develop an algorithm to spot patterns in the data. After six years, the Bureau Chief in Washington said, “Okay Tom, I’ve giving you a year to do this.”

And how did you start?
During that year I worked to develop an algorithm—which is a series of mathematic steps to produce a solution—to spot patterns in the hundreds of thousands of murders reported to the FBI. When I thought it was working, I tested it out on one area where I knew serial killings had happened, which was Seattle. This was where in the 80s and 90s, Gary Ridgeway murdered 48 teenage girls and women. He was America’s worst serial killer at the time and I could see his victims clearly in the Supplementary Homicide Report. The question was could I teach a computer to spot the same victims? Like, could it link the unsolved murders together? And it could.

How? What sort of algorithm is this?
We ended up using a process called cluster analysis which looks for similar murders, in a similar geography, with a low solution rate. Quite often, those are the results of a killer who’s been killing multiple victims.

How then do you go from a cluster of data to identifying a person?
That comes down to police work. All we do is highlight clusters of murders that look suspiciously linked, and hand that over to the police in the area.

Right. So after you’d tested the algorithm on Seattle, what happened when you ran the numbers on other places?
The algorithm came back and said there were 15 strangulations of women in Gary, Indiana, within the last 15 years and not one of them had been solved. Not one. So we tried to have a conversation with the Gary Police Department.

And what was their response?
Back in 2010 when I first contacted them, their first response was “I checked with the homicide division, and we have no active serial cases in Gary, Indiana.” Then, when I contacted them again with the names of the recent victims the algorithm was picking up, they went into total radio silence. They just wouldn’t talk about those cases. They just refused to entertain the idea that they had a serial killer. But then they arrested Darren Vann in 2014—a whole four years later—and the first question that the Gary Police of Chief was asked was, “Were you aware that there was a serial killer in Gary, Indiana?” And of course he said, “We had absolutely no idea.” Bullshit.

So how did they catch Darren Vann in the end?
He just confessed. He literally said “you got me.” Like a scene from a movie! He’d been killing women in Gary since the 90s and he took police around Gary and recovered other victims that the police knew nothing about, bodies that he’d stashed over the last 20 or so years. Seven women died after I’d tried to tell Gary Police that they might have a serial killer.

How do you feel about that?
It’s infuriating. It’s been the most infuriating experience of my professional life.

Why hadn’t they seen a pattern between all these unsolved crimes?
It’s actually a phenomenon called linkage blindness. It’s just the nature of how murders are investigated. A single murder case is solved on its own, separate from other cases. Therefore, linkage is often only recognised years later.

Now, I initially contacted you to see if the algorithm would be able to spot serial killers in Australia… but it seems like it wouldn’t. Why is that?
You’re at least the third Australian journalist to call me on this exact point and so far nothing has come of it. You’d need a database similar to the Supplementary Homicide Report in Australia, but you guys don’t have one.

Yeah I looked into it, and no state police department is willing to give up such data.
Yeah well something ought to be done about that. I know that given the laws and the sheer size difference of the countries, there are less serial killings in Australia, but I don’t believe that’s a good enough reason for lack of transparency. I think the people of Australia deserve to know how they’re being murdered, and whether those murders are being solved, and whether or not they’re related. I don’t know that the motivation is behind Australia’s lack of transparency, but a journalist with a stubborn streak might be able to change that.

How do you think we change that?
It would be a political fight. The Murder Accountability Project would be happy to testify that the creation of a database similar to our Supplementary Homicide Report would be in the public’s interest. We would give you all assistance necessary. We’d be more than happy to help if you were ever to have some kind of lawsuit to create such a database.

What do you think the ultimate version of the Murder Accountability Project looks like?
My dream scenario would be that the entire world adopted a system like ours. We would be more aware of what’s actually happening in the world. It would provide us with transparency and perhaps even instil some trust in the big data system. We’d be living in a much safer world.

Thanks Thomas.

Interview by Laura Roscioli. Follow her on Instagram