Dean Knox (UPenn): Data & Democracy Speaker Series

Please join us for a lecture hosted by DSI’s Data & Democracy Research Initiative.

Monday, May 4
4:00pm – 5:00pm
Data Science Institute, Room 105
5460 S University Ave

Title: Body-Worn Camera Data in the Study of Police Use of Force: Measuring Unreported and Hidden Uses of Force

Abstract: Analyses of police behavior have relied on self-reported officer accounts which may contain omissions, errors, or falsehoods. Body-worn camera (BWC) footage promised a more objective record, but resource and technological constraints leave most BWC footage unseen. In this paper, we draw on more than 700,000 BWC videos (~800 TB), and associated police administrative records from a large police department, to assess the accuracy and completeness of traditional use-of-force (UoF) reports. Using an original computational classifier in concert with human annotators, we estimate that between five and seven uses of force are captured on video for every one instance documented by officers. However, reporting rates vary sharply by force type: weapon use is well documented in traditional administrative files, while most undocumented force involves tackling and/or manually restraining civilians. Using a modified “capture-recapture” statistical technique common in ecology, we also provide the first lower bound on force incidents which are “hidden”—captured in neither paper nor video records—which we estimate is at least 13.2% of all force incidents. Taken together, we find that the data most used by supervisors and outside observers to monitor UoF—the written record—captures roughly 200 force incidents per year in this agency, but misses at least 1,300. We also show that BWC footage remains difficult to analyze computationally even with leading vision analysis techniques. To facilitate progress, we discuss possible policy changes and technological advances that would improve the usefulness of these data in the near term. We also provide the first ever human-annotated testbed of public BWC footage against which other researchers and industry leaders can objectively evaluate vision model performance.

Bio: Dean Knox, Assistant Professor, Operations, Information, and Decisions, Wharton School, University of Pennsylvania. Knox is a computational social scientist developing new methods for the study of complex and high-dimensional data. His research includes policing, speech analysis, ethnic politics, and political communication. His work has appeared or is forthcoming in Science, the Journal of the American Statistical Association, the Proceedings of the National Academy of Sciences, and the American Political Science Review. It has received the Gosnell Prize for excellence in political methodology, the John T. Williams dissertation prize, and the best poster award by the Society for Political Methodology.