Transparency at Sentencing
Vincent Chiao
Abstract
Abstract A frequently heard criticism, particularly in high-stakes settings such as criminal sentencing, is that automated decision-making aids can be highly opaque. In this chapter, the author analyzes two conceptions of transparency: transparency as a norm of publicity, and transparency as a norm of intelligibility. Automated decision-making aids plausibly raise concerns along both of these dimensions. However, the relative strength of these ethical concerns is mitigated on an incrementalist, rather than idealized, ethical analysis. This is because existing modes of human decision-making also suffer from defects of secrecy and intelligibility. In some cases, this is by design, as when officials are prohibited from revealing the bases of their decisions, and in some cases this is a feature of human psychology, as decision makers possess limited and partial insight into the reasons that actually explain why they decide as they do. Consequently, the strength of ethical objections grounded in transparency as publicity, and transparency as intelligibility, remains ambiguous.