Law enforcement agencies nationwide are using predictive policing software, yet almost none of their users, past or present, have clear measures for how effective or accurate they are.

An internal audit of the Los Angeles Police Department’s Chronic Offender Program and PredPol usage found few accurate ways to measure accountability in data-driven policing.

The AI Now Institute is calling for checks on the datasets used by predictive policing systems because of concerns that the technology can perpetuate, rather than address, “dirty” policing practices.

An Illinois Freedom of Information Act request filed by the Chicago-based Lucy Parsons Labs has led to the release of a “Best Practices and Training Guide” for the PredPol predictive policing software. The manual offers a rare look at how private companies are encouraging police to embrace algorithms as a means of seeking out crime before it happens.