A matter of justice. The opacity of algorithmic decision-making and the trade-off between uniformity and discretion in legal applications of artificial intelligence

In the last few years, decisions about matters of distributive and retributive justice have been more and more outsourced to automated systems (A.I.), and unprecedented ethical challenges have progressively emerged. As compared to human adjudicators, A.I.-based systems present, or may present in the future, concrete advantages in terms of efficiency and uniformity of performance. However, striving for uniformity may also have some sizeable costs. This paper aims to focus on a specific challenge “ the difficult trade-off between uniformity and discretion in judicial applications of artificial intelligence “ against the backdrop of current debates in philosophy, cognitive science, and artificial intelligence. I will argue that sidestepping the peculiarities of human reasoning might have some detrimental effects on the fairness of justice administration. This is particularly the case when the emphasis on uniformity is conducive to the elimination of reasonable standards of discretion, including the ability to bend the rules when circumstances so require.