OK. So I wrote some article about predictive policing. And it had lots of links, and it turns out sometimes editors see your links and are like ¯_(ツ)_/¯ to keeping those links when they publish the article. This is probably in part my own fault because I should have included those links as footnotes, and because working in print vs working online is Different, etc etc--the point is I feel really rotten that lots of really good reporting wasn't given due credit. Putting this list together does not totally fix it but will hopefully help people interested in the topic.
Also: there is a part at the end of this gist that explains my mixed feelings about the piece in general. The citations are probably more important, but if you want to watch me tableflip a little scroll down.
-
Matt Stroud's The minority report: Chicago's new police computer predicts crimes, but is it racist? This is pretty much the story everyone goes to when they cite the existence of predictive policing, and that's because it's really, really good work, don't let me go on about Matt Stroud's reporting we'll be here all day.
-
Darwin BondGraham and Ali Winston also did some of the best reporting on PredPol in particular, also generally recommend them as reporters, human beings. Here are 2 stories that they did that provide a lot of good context:
-
Also here is a thing Ali Winston wrote about a pending Arizona bill to fund predictive policing.
-
A lot predictive policing experiments and PredPol clients have been/are in the United States, but just for fun reference, here is a story about how the London Metropolitan Police worked with Accenture (also: of course it was Accenture) to develop a predictive model sort of in line with the work that happened in Chicago.
-
The Problem With Some of the Most Powerful Numbers in Modern Policing by Christopher Moraff has some good context for how no one actually knows how to demonstrate that predictive policing "works."
-
RAND's Evaluation of the Shreveport Predictive Policing Experiment is a little dry for some readers, but it gives a clearer picture of how varied the predictive policing tech landscape is and how unclear its outcomes are. Like, they were just using SPSS and paying cops overtime and that was "predictive policing," and surprisingly the cops found being able to take more time to talk to people was a bigger value-add than being told by math to go someplace.
-
This other RAND overview on predictive policing generally also offers some more context.
-
This is the Police Chief Magazine article that uses Wal-Mart and Amazon as reference points for the value of predictive policing.
This is hard to say, and it is not an excuse, but merely a fact. The article I wrote was originally much longer and had a lot more relevant case study examples, and I hope was a little more clearly nuanced. While I completely understand the need to fit space in print, I wasn't really expecting the piece I wrote to be edited into essentially a hot take. I was really close to asking to have the story pulled or not published under my own name, but frankly, I never discussed a kill fee and I needed the money. Ultimately, the pay was nowhere near enough to justify the source I burned and, based on social media feedback, it mostly being read with the point being kind of missed.
This isn't to blame the editors entirely--I had a moment to withdraw and I didn't take it, and I take full responsibility for that. It bears repeating that the conditions also weren't in their favor--print is hard, working on deadline is hard. But it would have been cool to not spend 2 months working on something and have it that embarrassingly reduced and misinterpreted. Oh well.