Hope, Hype, and Fear: The Promise and Potential Pitfalls of the Big Data Era in Criminal Justice

Over the past decade, algorithmic decision systems (ADSs) — applications of statistical or computational techniques designed to assist human-decision making processes — have moved from an obscure domain of statistics and computer science into the mainstream. Advocates of these “intelligence-led” or “evidence-based” policy approaches assume big data tools will allow government agencies to use objective data to overcome historical inequalities to better serve underrepresented groups. However, the assumption of objective data is flawed. All human behavior or social phenomenon that machine learning algorithms attempt to predict come from a data-generation process (DGP) which is comprised of trillions of complex interactions between the roughly seven billion people that inhabit our planet. If a statistical model — an abstraction of the DGP — assumes incorrectly about the underlying dynamics, the predictions and conclusions generated will be inaccurate and biased. In this paper, I highlight examples from a variety of fields to discuss why this is particularly true for algorithmic decision systems in criminal justice, then examine a specific application of predictive policing in Oakland, California, and then conclude with what police departments should consider in the deployment of policing technologies in the era of artificial intelligence and big data.