Trustworthy AI in courts: A legitimacy analysis

Abstract

The use of AI in courts is highly contested, with positions ranging from outright opposition to unmitigated enthusiasm. However, normative assessments of its deployment often rely on standards or analogies that are unclear in their pedigree, coherence and scope conditions. This article explores how theories of legitimate authority can offer some guidance. Legitimacy theories serve to assess institutions and how they use their authority. The framework allows us to identify central issues and dilemmas, order different normative standards and identify relevant differences and similarities across time and jurisdictions. To undertake such a legitimacy analysis of AI in courts, we identify six forms of delegation to AI in courts (servant, interlocutor, advisor, partner, agent and trustee) and analyse the legitimacy implications for six aspects of adjudication (legal process, inventorising, problem-solving, reasoned judgment, legitimation and discretion). In some areas, AI may enhance both the efficiency and quality of the judicial process, while we argue that the delegation of judicial power to AI in aspects which require extensive human discretion and human presence is difficult to justify.