DATA-DRIVEN POLICE PROFILING

Abstract

Police departments increasingly rely on algorithms and other datadriven methods of identifying high-crime areas and people who are at high risk for involvement in crime. This Article examines several constitutional obstacles to this type of policing. First, to the extent that these algorithms rely on data entitled to privacy protection, they may violate the Fourth Amendment. Second, the steps police take in response to a "hot" place or person designation must also be subject to constitutional regulation. Further, the principle of legality should prohibit the police from acting on any risk designation, even one that is very likely accurate, in the absence of direct observation of risky conduct. For the same reason, and to combat the influence of racially based "dirty data," algorithm developers must finely tune both the inputs and outputs of their profiles. Finally, a failure to disclose the inner workings of a predictive algorithm may violate the Confrontation Clause. Combined, these legal concerns could well spell the demise of profile-driven policing.