2021/02/01 by Kristian Lum, David B. Dunson, Lum, Kristian +3 · 1 citation
Computer Science · Mathematics · Social Sciences · #Applications (stat.AP) #Bayesian Modeling and Causal Inference #Crime Patterns and Interventions #FOS: Computer and information sciences #Statistical Methods and Bayesian Inference
paper · pdf · doi:10.48550/arxiv.2102.01135
openalex publication_date 2021/02/01 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
Risk assessment instruments are used across the criminal justice system to\nestimate the probability of some future behavior given covariates. The\nestimated probabilities are then used in making decisions at the individual\nlevel. In the past, there has been controversy about whether the probabilities\nderived from group-level calculations can meaningfully be applied to\nindividuals. Using Bayesian hierarchical models applied to a large longitudinal\ndataset from the court system in the state of Kentucky, we analyze variation in\nindividual-level probabilities of failing to appear for court and the extent to\nwhich it is captured by covariates. We find that individuals within the same\nrisk group vary widely in their probability of the outcome. In practice, this\nmeans that allocating individuals to risk groups based on standard approaches\nto risk assessment, in large part, results in creating distinctions among\nindividuals who are not meaningfully different in terms of their likelihood of\nthe outcome. This is because uncertainty about the probability that any\nparticular individual will fail to appear is large relative to the difference\nin average probabilities among any reasonable set of risk groups.\n