Skip to Content

Predictive Policing Did Nothing to Prevent Violence in Chicago

Tom Cruise in Minority ReportTom Cruise in Minority Report
Tom Cruise in Minority ReportCourtesy of 20th Century Fox

New research led by RAND Corporation criminologist Jessica Saunders has concluded that a Chicago experiment in predictive policing was, at least in its early stages, a failure.

The program, launched in 2013, used data analytics to identify individuals, known as Strategic Subjects, likely to be involved in a shooting. The original premise was that social workers would engage with these at-risk individuals before they became embroiled in violence. Instead, the report finds that the list only served to help target those individuals for arrest after a crime had already been committed, making it little more than what one commentator called a “data-driven ‘most-wanted’ list.”

Get Data Sheet, Fortune’s technology newsletter.

Saunders and her co-authors found that identified Strategic Subjects were no more or less likely to be a victim of violence than other individuals. They were, however, more likely to be arrested for gun violence. The researchers further found that the program had no effect on Chicago’s broader crime patterns.

Speaking with the Verge, one of the report’s authors, John S. Hollywood, suggested that the program’s failure had more to do with implementation problems than with flaws in the idea of predictive policing. Amidst a number of other anti-crime initiatives, Hollywood says the Strategic Subjects program “just got lost”—until, that is, police needed to locate suspects.

That isn’t far from the dystopian worst-case scenario sketched way back in 1956 in Philip K. Dick’s “The Minority Report,” adapted into a Tom Cruise film in 2002. In that story, ‘Precrime’ investigations lead to the arrest of suspects before they’ve even broken the law.

For more on big data and policing, watch our video.

In response to the RAND findings, the Chicago Police have stressed that the results were based on an early version of the program, which has since “evolved greatly.” And most of the failures were in implementation of the new method, with little evaluation so far of the accuracy of the predictive model itself.

The CPD also stressed that the program made no use of race or socioeconomic data for identifying Strategic Subjects. That’s a vital point, since the program so obviously raises concerns that it would simply reinforce what a task force has described as “systemic and sanctioned” pattern of bias and abuse by the department.