Developing an Integrated Machine Learning Algorithm for Criminal Justice Practitioners: A Superior Method for Resource Allocation?

• In Event: Machine Learning and Causal Inference

Abstract

How should parole/probation officers prioritize their caseload? How should police decide where to patrol? And how should prison guards decide which inmates to monitor closely? Most would argue that these criminal justice actors should focus on offenders/inmates who are most at risk of later misconduct. But how do they know who is most likely to engage in offending in the future? More generally, how can we know, more precisely, whether one offender is at a higher risk of recidivism compared to another? In an attempt to provide empirical answers to these questions, we will adopt an integrated analytical approach that combines the “brute force” of machine learning techniques with the intuitive and logically structured Elo ranking algorithm. Our goal is to find an efficient and accurate way to identify cases that pose the highest levels of risk, such that limited resources can better be allocated to those most in need.

Authors

• Shahin Tasharrofi, University of Cincinnati

• J.C. Barnes, University of Cincinnati