Poetic justice in the age of (Generative) AI

AI systems are increasingly relied upon in the judiciary to carry out a variety of tasks (referred to as ‘judicial AI’) with the aim of improving judicial decision-making. This aim, however, presupposes an underlying normative ideal of what ‘good’ judicial decision-making is, which can be used to assess whether judicial AI indeed brings judges closer to the ideal. In this article, I seek to advance two goals. First, I articulate an ideal of judicial decision-making that draws on Martha Nussbaum’s conception of ‘poetic justice’ (section 2). This conception not only emphasizes knowledge about the law, but also the importance of some inherently human virtues, which dispose judges to acknowledge the individuality and dignity of the parties involved, rely on their emotions to attain a more comprehensive understanding of the case, and use their imagination. Second, I explore how judicial AI can advance or undermine the ideal of poetic justice (section 3). While traditional AI intuitively runs counter to this ideal, I argue that the responsible deployment of generative AI, which seems more prone to ‘poetic’ decision-making given its ability to mimic certain human qualities, could actually enhance judges’ poetic capabilities, and I offer several examples. At the same time, I also point out important shortcomings and describe how reliance on this technology can undermine poetic justice. My analysis reveals that the impact of generative AI on this ideal cannot conclusively be determined, as much depends on the design, purpose and deployment of the system. I therefore offer some suggestions of how judicial AI can be brought in line with poetic justice (section 4), which serve as pathways for future research, such as acknowledging the sui generis character of judicial judging, adopting a 'frictional' approach to the design and use of judicial AI, and introducing targeted regulation, before concluding (section 5).