AI Risk Scoring Models in Parole Decisions

Abstract

Artificial intelligence (AI) risk scoring models are increasingly used in criminal justice systems to assist parole boards in evaluating offender rehabilitation, likelihood of recidivism, and public safety outcomes. These systems aim to support evidence-based, data-driven decision-making but raise concerns regarding algorithmic bias, fairness, transparency, ethical governance, and due process. This paper examines how AI-driven risk scores are constructed, how they influence parole outcomes, and what psychological, computational, and legal issues emerge. Using empirical research, comparative policy analysis, and evaluation of major tools such as COMPAS and LSI-R, the study proposes a Human-AI Hybrid Parole Decision Framework to balance predictive accuracy with accountability and ethical safeguards.