Judged by Robots: Preferences and Perceived Fairness of Algorithmic versus Human Punishments

Abstract Automated decision-making is increasingly prevalent, prompting discussions about AI replacing judges in court. This paper explores how machine-made sentencing decisions are perceived through an experimental study using a public good game with punishment. The study examines preferences for human versus automated punishers and the perceived fairness of penalties. Results indicate that rule violators prefer algorithmic punishment when penalty severity is uncertain and violations are significant. While human judges are typically reluctant to delegate, they are more likely to do this when they do not have discretion over the sanction level. Fairness perceptions are similar for both humans and algorithms, except when human judges choose a less severe penalty, which enhances perceived fairness.