An Assessment of Artificial Intelligence's Competence in Exercising Judicial Discretion

Abstract

This article examines whether artificial intelligence systems can exercise judicial discretion and function as normative agents within legal adjudication. The study first analyses the meaning of judicial discretion through the perspectives of Hans Kelsen, H.L.A. Hart, Lon L. Fuller, and Ronald Dworkin, arguing that judicial decision-making involves not merely the application of legal rules but also norm-creation, value assessment, and practical reasoning. Building on this framework, the article evaluates whether contemporary generative artificial intelligence systems possess the capacities required for normative agency. Particular attention is given to concepts such as autonomy, responsibility, practical reasoning, and authority, drawing on Joseph Raz’s theory of normativity. The study argues that current artificial intelligence systems operate primarily through statistical pattern recognition and lack the capacity for genuine normative evaluation. Furthermore, the article examines the issue through Robert Brandom’s conception of the space of reasons and the practice of giving and asking for reasons. It concludes that, although artificial intelligence systems may generate outputs with normative implications, they cannot presently be regarded as genuine participants in the space of reasons or as normative agents capable of exercising judicial discretion and creating law.