Actuarial risk assessment instruments are used for various applications, including criminal justice, to assess the likelihood that a contextually relevant outcome will occur. However, several stakeholders have expressed concerns about the fairness and technical soundness of this approach. For example, some researchers believe that these models cannot correctly assess individuals due to large statistical uncertainty at the individual level. Lum et al. suggest that RAIs can only satisfy fairness requirements if the models are designed to represent individual-specific probabilities and their associated uncertainty intervals such that individuals who are not statistically distinguishable will not be treated differently by the model’s predictions. They propose an approach that uses a Bayesian hierarchical model with individual-level random effects and a large longitudinal dataset to learn the distribution of individual probabilities across the population.