Automating Body-Worn Camera Footage Review Through AI: Preliminary Findings from a Multi-Site Randomized Control Trial

• In Event: Experimental Methods in Criminology (Sponsored by the Division of Experimental Criminology)

Abstract

Policing has been and still is largely invisible. Though body-worn cameras (BWCs) have addressed this problem to some extent, most of the footage recorded by BWCs (about 95%) is never reviewed or seen by anyone. Departments’ inability to review footage undercuts several of the perceived benefits of BWCs, from enhanced transparency to performance evaluation. One potential solution involves AI-driven analytics that process and categorize vast amounts of footage in near-real time. One example is Truleo, which uses natural language processing to analyze the audio of body-worn camera footage. For the current study, we describe three ongoing randomized controlled trials testing the implementation and impact of Truleo in the Apache Junction and Casa Grande Police Departments, as well as the Arizona Department of Public Safety. We describe our design, data, and preliminary findings related to implementation and officer attitudes towards this technology.

Authors

• Michael White, Arizona State University

• Aili Malm, California State University, Long Beach

• Seth Watts, Arizona State University