Predictive policing

The Pasco County Sheriff’s Office permanently ended its ‘Intelligence Led Policing’ predictive policing program and paid $105,000 to four plaintiffs to settle a federal civil rights lawsuit. The settlement agreement acknowledged the program exceeded officers’ implied license to knock on doors and perform offender checks, interfering with the plaintiffs’ First, Fourth, and Fourteenth Amendment rights. Plaintiff Darlene Deegan had been harassed for three years after her son was flagged by the program. The program used algorithms to flag ‘prolific offenders,’ as revealed by a 2020 Tampa Bay Times investigation. The sheriff’s office also agreed not to create a similar program.

On January 23, 2020, the Chicago Office of Inspector General (OIG) released an advisory on the Chicago Police Department’s (CPD) predictive risk models, the Strategic Subject List and Crime and Victimization Risk Model. CPD received $3.8 million in federal grants and spent ten years developing the models, which scored individuals on likelihood of being a party to violence. The department decommissioned the program in November 2019 after grant funding ended. OIG raised concerns about unreliable risk scores, improperly trained personnel, and lack of access controls. Inspector General Joe Ferguson said, “We reiterate our hope that the Department will apply these lessons going forward.”

The Chicago Police Department decommissioned its ‘party to violence’ risk-model program — the Strategic Subject List and its successor, the Crime and Victimization Risk Model — on November 1, 2019, after telling the Office of Inspector General in August 2019 it intended to do so. On January 23, 2020 the OIG published an advisory reporting that CPD had received $3.8 million in federal grants and spent ten years developing the sequential models, which assigned scores predicting the likelihood a person would be the victim or offender in a shooting. The advisory’s areas of concern included the unreliability of risk scores and tiers, improperly trained sworn personnel, a lack of controls on internal and external access, and interventions that may have attached negative consequences to arrests that never produced convictions. The algorithm was built by the Illinois Institute of Technology under a Department of Justice Bureau of Justice Assistance grant. The databases held data on more than 300,000 people.

On March 8, 2019 the LAPD Office of the Inspector General released an audit of the department’s data-driven policing programs — PredPol, Operation LASER and the ELUCD survey platform — and presented it to the Police Commission on March 12. The audit found significant inconsistencies in how individuals were selected and retained in LASER’s chronic-offender program, found the programs lacked oversight and often strayed from stated goals, and concluded there was insufficient data to determine whether PredPol reduced crime; the department had used PredPol for roughly eight years. LAPD suspended its Chronic Offender program and database in August 2018 without announcing it. A memo dated April 15, 2020 said the department would discontinue PredPol immediately, citing the pandemic-driven freeze on new contracts.