Skip to content

“The Billy Beane of murder”?

John Hall points to this news article in Businessweek by by Robert Kolker, “Serial Killers Should Fear This Algorithm,” and writes:

I couldn’t help but think that you should get some grad students working on the data set mentioned in the article below.

Meanwhile this story got picked up by the New Yorker, although without any reference to the earlier Businessweek article. This seems weird to me—wouldn’t the New Yorker writer have done a google search on Thomas Hargrove (the subject of the article) and found this earlier story?

In academic writing you’re always supposed to cite what came before and discuss how your work advances the field. I guess journalism is different: there it’s standard practice not to acknowledge related articles published elsewhere.

Anyway, I’d forgotten about this story but then I learned about the New Yorker article, and Michael Maltz and I looked into the story. We’ll publish something on it soon for Slate, or maybe our article has already appeared. (Yes, it did appear.)

In any case, yes, regarding the query above: This does seem like a good project for students to look into.


  1. Joe says:

    I don’t think it’s “standard practice not to acknowledge related articles” in journalism. Rather, there’s a different standard for what qualifies for a “citation” in journalistic circles. There’s just a different standard for what receives a “citation”.

    The basic goal of journalistic “citation” is to identify, for the reader, the source of information. If you reported something independently, you want to accurately convey that to the reader. On the other hand, when reporting consists of confirming something first reported elsewhere, you acknowledge it (“X has confirmed a report by Y…”; “As first reported by X…”).

    That is – the journalistic goal is to provide an honest intellectual history of the piece, so that the reader understands its origins and credibility.

    The academic goal is a different — you’re trying to fit your piece into the “literature”, rather than display the intellectual history of your own work. Thus, you cite works that you didn’t encounter until your own research was complete, and (for better or for worse) academics often cite articles that they have not read (which would be a cardinal sin for a journalist).

    • someone says:

      I have never ever read, seen or heard any journalists saying “As first reported by X” unless X is themselves or other journalists in the same company. There’s no way they’d give a shout-out to a rival news group.

  2. jrkrideau says:

    It does seem like a good project for a student but the data cleaning component sounds horrendous.

  3. Jonathan says:

    Detroit just announced that reviewing some of the vast backlog of rape kits showed 86 apparently active serial rapists. They mean something like 10 or more rapes. The issue is less ‘algorithm’ than that collecting data requires people, time and money and analyzing that data requires people, time and money. I don’t know what world you live in or your experiences, but spend a few months in the bowels of the police and judicial systems, which are not the same thing, and you see the gulf between what’s written about and what actually happens. Nearly everyone is caught in the act or they’re not caught at all until they’re caught in the act some other time. This is even true in my low crime part of almost Boston: we have a data specialist – she’s great – and she coordinates with Boston and Cambridge when there’s a rash of break-ins or other crimes so the police can catch the person in the act or just after (and every now and then just before). This works off simple incident reports collected by patrol officers and detectives.

    For some reason, the attraction is connecting events that aren’t obviously related. Like identifying that people who no one knows are dead are actually dead and by the same person. This is actually very, very marginal to police work because they know they’re unlikely to uncover anything until and unless the guy is caught in the act or near to it. Do you have any idea of the numbers of people who aren’t tracked in the system? I put it that way because the ‘algorithm’ approach implies you have rational data about missing people when no one tracks people generally – except for Google and various other apps and advertisers – and no one tracks missing people on the margins because people go missing all the time. I mean like all the time. And for a lot of reasons, like they’re addicts or they’re running out on some family or they just feel like leaving or whatever. Most lives are really fucked up and difficult. Life is not TV. There are very few ‘leads’ to follow. On TV, when family and friends don’t say anything, that becomes a way to get into what really happened but in real life it usually means the investigation stalls. So when you read that a girl from a ‘good’ family was reported missing years ago and it turns out she was killed by a guy who also killed homeless women and prostitutes, it isn’t that the police ‘should have done more’ but that they really can’t because we don’t have a way of pulling up all the cam footage across America to identify someone’s lost daughter.

    It’s not weird to me that ‘solve’ rates and so on have dropped because crime rates have dropped and most crimes are really easy to solve because the guy has the murder weapon on his person, is covered with cuts, has blood on his clothes and often the dead person is lying on the floor or is on the couch or bed. Like the guy who stabbed his girlfriend 15 times with a screwdriver, hit her with a club 50 or more times, and then ran over her head because she changed the TV channel and he snapped for reasons no one will ever know. (That’s real and I understated the actual violence.) Not a hard case to crack. People get turned in by neighbors. They get turned in by family, often because the family are scared to death of him and fear they’ll be next.

    For reasons hard to explain, crime rates dropped, murder rates have largely dropped too – unless you believe in vast numbers of missing person killed by serial killers moving around the country – and that means a big drop in cases where the victim is lying on the kitchen floor and the murder weapon is a big kitchen knife which may still be sticking out of the body. Is it rational for there to be a significant drop in closure rates? Is it rational that easy cases would drop more than hard ones? I don’t know but it doesn’t surprise me that this has happened because it appears to have happened.

  4. Steve Sailer says:

    The LAPD announced a few years ago that there were at least five serial killers active in L.A. around 1990 that nobody had noticed at the time, what will all the crack war killings.

    A useful computer tool would be one that could rule out different individuals as serial killers because they were locked up at the time.

Leave a Reply