Autor: Romero Maria Noelia*, Anauati María Victoria**, Baraldi Lucia*, Sosa Escudero Walter *, Tommasi Mariano*
Institución: (*)UdeSA - CEDH, (**)UdeSA - CEDH - CIAS
Año: 2025
JEL: K4, C5
Resumen:
Recidivism is a persistent challenge for criminal justice systems worldwide, yet evidence from Latin America remains scarce. This study addresses that gap through three contributions. First, we review the individual, institutional, and environmental determinants of recidivism with attention to Latin American contexts. Second, drawing on two decades of Argentine prison data, we characterize recidivism patterns and apply six machine learning models to predict reoffending. Economic offenses and age at incarceration emerge as the strongest predictors, while geographic indicators also matter, given the clustering of repeat offenders in certain prisons. We show that prison-level information, often collected but underused, enables reasonably accurate risk prediction and can guide more effective rehabilitation and prison management. Third, we discuss how AI-based prediction tools could be applied by judges, correctional authorities, and policymakers, as well as the institutional, data, and ethical challenges such implementation entails.