The New Surveillance Discretion: Automated Suspicion, Big Data, and Policing

In a crime analytics bureau, a police officer logs in to see what alerts have been posted by social media software designed to spot potential threats within the billions of daily online tweets, pins, likes, and posts. On the street, a police officer uses his body-worn camera to scan a crowd; the feed is sent in real time back to the department where facial recognition and movement analysis software alerts the patrol officer as to whether furtive movements or people on watch lists have been identified. Police follow up on these alerts to identify people who should be immediately investigated. Other people are dismissed as not posing an immediate threat but are logged on watch lists for future reference. No police department has all of this technological ability today, but some will one day soon.1 There is no question that this version of big data policing is on the cusp of wider adoption,2 and it raises key questions about fundamental issues of police discretion and accountability. Whether the police identify a person and choose to investigate him for suspected criminal activity is a decision largely left up to the police. The decisional freedom3 to focus police attention on a particular person or persons rather than others—what I’ll call “surveillance discretion”—is a widely accepted means of investigation. Law enforcement would be unimaginable without it. This task of filtering—identifying suspects from the general population—exemplifies traditional police work. Police officers usually generate leads by focusing their attention on particular suspects through observation, questioning, and information conveyed by witnesses, victims, or other third parties. New technologies have altered surveillance discretion by lowering its costs and increasing the capabilities of the police to identify suspicious persons. Furthermore, soon it will be feasible and affordable for the government