Uncertainty in criminal justice algorithms: Simulation studies of the Pennsylvania Additive Classification Tool
Swarup Dhar, Vanessa A. Massaro, Darakhshan Mir, Nathan C. Ryan · Proceedings of symposia in applied mathematics · 2025
Much attention has been paid to algorithms related to sentencing, the setting of bail, parole decisions and recidivism while less attention has been paid to carceral algorithms, those algorithms used to determine an incarcerated individual’s lived experience. In this paper we study one such algorithm: the Pennsylvania Additive Classification Tool (PACT) that assigns custody levels to incarcerated individuals. We analyze the PACT in ways that criminal justice algorithms are often analyzed: namely, we train an accurate machine learning model for the PACT; we study its fairness across sex, age and race; and we determine which features are most important. In addition to these conventional computations, we propose and carry out some new ways to study such algorithms. In particular, instead of focusing on the outcomes themselves, our approach shifts the attention to the variability in the outcomes. Many carceral algorithms are used repeatedly and there can be a propagation of uncertainty; we develop an approach to describe the propagation of uncertainty using simulation studies and sensitivity analyses. The results from our new approach and the conventional approaches shine light on problematic aspects of the PACT.