Harcourt, Bernard. 2007. Against Prediction: Profiling, Policing, and Punishing in an Actuarial Age. Chicago: University of Chicago Press
This is really two books in one. The first part of the book is a completely fascinating history of the use of prediction (a.k.a. statistical treatment rules, a.k.a. profiling) in a criminal justice context in the United States. Perhaps surprisingly, this history dates back to eccentric sociologists in the early 20th century. The history has interest in its own right, but also serves as yet one more reminder that using statistical models applied to what were at the time "big" data to generate actionable predictions is no new thing. The methods have improved, and our standards of bigness have grown, but the idea and an understanding of many of the basic issues go way back.
The second half of the book critiques the more recent literature and policy practice in the criminal justice domain. This critique has three parts: The first makes the point that individuals most likely to engage in some behavior may not be those most likely to respond to some intervention. This is not a new point - it appears in Black, Smith, Berger and Noel (2003) in the context of UI profiling and it was not new there either - but it is one that many people really have trouble understanding. Much of this discussion in the book centers on the literature on police profiling in roadside stops. The second criticism concerns a sort of "ratchet" effect whereby the focus of criminal justice resources on one particular group as a result of the use of prediction models leads to a misguided change in public perceptions of underlying differences in criminal behavior across groups. The third criticism concerns changes in our underlying philosophical notions of justice that might result from the growing dominance of prediction in how we implement the justice system in practice.
I found the first criticism both the best realized in the book and the most compelling. The others are provocative and, in the case of the ratchet effect, have a sort of face validity, but, I felt, a bit underdeveloped in the text.
I enjoyed reading the book and learned a lot from it. Recommended if you are into such things.
Amazon book page
Barnes and Noble book page
Hat tip: Shawn Bushway, for recommending it to me several years ago.
Who was my favorite student this term?
5 years ago