AI Risk Assessment Tools Exhibited Racial Bias in Prison Parole Decisions

Beginning in 2016, investigative reporting by ProPublica revealed that the COMPAS (Correctional Offender Management Profiling for Alternative Sanctions) algorithm, used by courts and corrections departments across the United States, exhibited significant racial bias in predicting recidivism risk. The analysis of over 7,000 people arrested in Broward County, Florida found that Black defendants were almost twice as likely to be incorrectly flagged as future criminals compared to white defendants (45% vs 23% false positive rate), while white defendants were more likely to be incorrectly flagged as low risk (48% vs 28% false negative rate). The COMPAS system, developed by Northpointe (later Equivant), was used by correctional agencies in multiple states including Florida, Wisconsin, California, and New York to inform parole decisions, sentencing recommendations, and custody classifications. Similar bias patterns were subsequently identified in other risk assessment tools including the Pennsylvania Risk Assessment Instrument (PRAI), the Level of Service Inventory-Revised (LSI-R), and the federal PATTERN (Prisoner Assessment Tool Targeting Estimated Risk and Needs) system implemented by the Bureau of Prisons in 2019. Research by the Brennan Center for Justice and ACLU found that these algorithmic tools systematically overestimated recidivism risk for Black defendants while underestimating risk for white defendants. The algorithms incorporated factors that served as proxies for race, including neighborhood characteristics, employment history, and family background - all areas where historical discrimination created disparate impacts. A 2018 study by researchers at Dartmouth College found that even when controlling for criminal history, Black defendants received higher risk scores than white defendants. The widespread deployment of these tools affected hundreds of thousands of incarcerated individuals across the United States. In Wisconsin alone, the COMPAS algorithm influenced decisions for over 15,000 defendants annually. The biased risk scores directly impacted parole decisions, with higher scores leading to longer periods of incarceration for Black prisoners who posed similar actual risk levels as white prisoners receiving lower scores. This perpetuated and amplified existing racial disparities in the criminal justice system, with Black individuals serving disproportionately longer sentences for equivalent crimes and risk profiles.