AI revolution in policing? An ethnography of digital technologies in law enforcement practice

Police agencies across the globe are collecting and processing an ever-growing variety and amount of data – not only have many agencies begun digitizing their records but video from body-worn cameras, CCTV recordings, automatic number plate readers, GPS traces from patrol cars, and ‘open source’ intelligence such as social media provide new sources of information. This poses both a challenge in accessing relevant data in practice and an opportunity for new search tools and new forms of data analysis to proliferate. Possibly the two most prominent examples of machine learning tools that address these issues are predictive policing (the use of data to suggest likely areas and individuals for police intervention with the goal of pre-empting crime) and facial recognition (used to either associate a name with the picture of a perpetrator in investigations or to identify persons of interests among passers-by in live-facial recognition). Technologies based on machine learning are set to transform the character of policing. Some technologies will fail while others will be adopted. Especially the public debate on predictive policing and facial recognition, but also the Black Lives Matter protests of recent months provide a unique moment for problematizing the role of the police and the technology they use. The values and assumptions coded into software products will be difficult to decipher once they have been closed into ‘black boxes’ and removed from a status as ‘matters of concern’. Only by exploring the grey areas of this technological practice beyond the current polarized debate, there is a chance to unpack the shifts in policing these technologies bring about and to shed light on what the use of such technologies does to questions of social inequality and to institutional attempts to re-orient policing.

My research aims to produce a multi-sited ethnography of the data practices in policing with a particular focus on predictive policing, tracing its contingencies from production to implementation, to fully grasp the changes that these technologies bring about. This project will contribute to AIming Toward the Future’s overall objective to understand the shifts in knowledge production brought about by data-driven technologies and their consequences for internal and external power relations of policing. I will look particularly at 1) the influence of data and risk scores on discretionary decision making, 2) the managerial use of data in officer supervision, 3) the use of new data sources in investigations, 4) the consequences for community interactions, 5) the role of civil society actors in shaping and restricting the adoption of these technologies, and 6) the ways private companies shape perceptions of risk.