RISK, RACE, AND RECIDIVISM: PREDICTIVE BIAS AND DISPARATE IMPACT*
2016/11/01 by JENNIFER L. SKEEM, Jennifer L. Skeem, CHRISTOPHER T. LOWENKAMP +1 · 271 citations
Psychology · Social Sciences · #Computer science #Computer security #Conviction #Crime Patterns and Interventions #Criminal Justice and Corrections Analysis #Criminal justice #Criminology #Law #Political science #Proxy (statistics) #Psychology #Psychopathy, Forensic Psychiatry, Sexual Offending #Race (biology) #Recidivism #Risk assessment #Sociology #Statistics
paper · doi:10.1111/1745-9125.12123
published in Criminology 54(4), 680-712 (Wiley)
openalex publication_date 2016/11/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/26
Abstract
One way to unwind mass incarceration without compromising public safety is to use risk assessment instruments in sentencing and corrections. Although these instruments figure prominently in current reforms, critics argue that benefits in crime control will be offset by an adverse effect on racial minorities. Based on a sample of 34,794 federal offenders, we examine the relationships among race, risk assessment [the Post Conviction Risk Assessment (PCRA)], and future arrest. First, application of well‐established principles of psychological science revealed little evidence of test bias for the PCRA—the instrument strongly predicts arrest for both Black and White offenders, and a given score has essentially the same meaning—that is, the same probability of recidivism—across groups. Second, Black offenders obtain higher average PCRA scores than do White offenders (d = .34; 13.5 percent nonoverlap in groups’ scores), so some applications could create disparate impact. Third, most (66 percent) of the racial difference in PCRA scores is attributable to criminal history—which is already embedded in sentencing guidelines. Finally, criminal history is not a proxy for race, but instead it mediates the relationship between race and future arrest. Data are more helpful than rhetoric if the goal is to improve practice at this opportune moment in history.
Cited by
- How Formerly Incarcerated People Envision Technologies for Prison Parole
- Removing biased data to improve fairness and accuracy
- Do risk assessment tools help manage and reduce risk of violence and reoffending? A systematic review.
- It's COMPASlicated: The Messy Relationship between RAI Datasets and Algorithmic Fairness Benchmarks
- Identifying Significant Predictive Bias in Classifiers
- Fair Prediction with Disparate Impact: A Study of Bias in Recidivism Prediction Instruments
- Unequal Uncertainty: Rethinking Algorithmic Interventions for Mitigating Discrimination from AI
- Fairness Evaluation in Presence of Biased Noisy Labels
- Emerging Technologies at Work: Policy Ideas to Address Negative Consequences for Work, Workers, and Society
- Algorithmic Fairness: Choices, Assumptions, and Definitions
- Empirical observation of negligible fairness-accuracy trade-offs in machine learning for public policy
- Framework, Standards, Applications and Best practices of Responsible AI : A Comprehensive Survey
- Big data hurdles in precision medicine and precision public health. [europepmc]
- Violence risk assessment for young adults receiving treatment for early psychosis. [europepmc]
- Feasibility and Utility of Different Approaches to Violence Risk Assessment for Young Adults Receiving Treatment for Early Psychosis. [europepmc]
- Defining AMIA's artificial intelligence principles. [europepmc]
- Towards a More Evidence-Based Risk Assessment for People in the Criminal Justice System: the Case of OxRec in the Netherlands. [europepmc]
- What's risk got to do with it: Judges' and probation officers' understanding and use of juvenile risk assessments in making residential placement decisions. [europepmc]
- Racial/ethnic disparities of the pact in predicting recidivism and court dispositions for justice-involved youth. [europepmc]
- Implementing a Machine Learning Screening Tool for Malnutrition: Insights From Qualitative Research Applicable to Other Machine Learning-Based Clinical Decision Support Systems. [europepmc]
- Purchaser, firearm, and retailer characteristics associated with crime gun recovery: a longitudinal analysis of firearms sold in California from 1996 to 2021. [europepmc]
- Enhancing the fairness of AI prediction models by Quasi-Pareto improvement among heterogeneous thyroid nodule population. [europepmc]
- Individual, firearm, and purchasing characteristics associated with risk of firearm-related violent crime arrest: a nested case-control study. [europepmc]
- The challenges of being imperfect: how do judges and prosecutors deal with sentencing disparity. [europepmc]
- Public perception of accuracy-fairness trade-offs in algorithmic decisions in the United States. [europepmc]
- Psychosocial determinants of recidivism risk among incarcerated individuals with a history of substance use: A systematic review. [europepmc]
- Examining the Reliability and Validity of the Criminal Sentiments Scale-Modified and the Pride in Delinquency Scale in Black and White Canadian Justice System-Involved Youth. [europepmc]
- bias goggles: Graph-Based Computation of the Bias of Web Domains Through the Eyes of Users [europepmc]
Related