Association for Public Policy Analysis & Management

Algorithmic risk assessments are being implemented in the criminal justice system to inform decisions made about pretrial detention, sentencing, and supervision. Their use even extends to the juvenile justice system. While technology is often viewed as a mechanism to move away from biases and create more uniformity, risk assessments have been criticized for their inclusion of, and heavy reliance on, factors that could create and perpetuate racial disparities within the criminal justice system. These factors include criminal history and socioeconomic characteristics. This paper examines validation studies that have been completed on various risk assessments used throughout the United States. The findings vary widely but most conclude that risk assessments cause higher rates of false positives for minority populations, including African Americans and Hispanics. Findings also suggest that risk assessments often use factors that are not significantly attributed to predicting risk for minority populations. Both issues could create and perpetuate racial disparities within the criminal justice system depending on their application. However, the use of algorithmic risk assessments should not be abandoned. Instead, universal standards of fairness and validity should be implemented, and all risk assessments should be validated for each racial group that they will be used for. Additionally, race-specific interactions with the factors used within risk assessments should be taken into consideration.