Algorithms As Prosecutors: Lowering Rearrest Rates Without Disparate Impacts and Identifying Defendant Characteristics Noisyy to Human Decision-Makers

Abstract

We investigate how machine learning might bring clarity to a human decisions made during the criminal justice process. We created a model that predicts a defendant’s risk of being rearrested after their charges are dropped. We used a database from the office of the Orleans Parish District Attorney that covers cases from 1988-1999. Applying strategies identified by past research that compared prediction models to judicial decision makers, our model selected higher risk individuals to prosecute than its human counterparts did. For a set charge rate, our model would reduce the rearrest rates between 5% and 9%. Developed further, such a model could have several important policy implications: it might identify defendant characteristics that are particularly ‘noisy’ to prosecutors; it could suggest ways of alleviating criminal caseloads without increasing crime rates; and it might provide important insights into how a prosecutor’s background relates to the quality and nature of their charging decisions.