Dall’equità formale all’equità algoritmica : Per una capability-sensitive AI governance

The increasing use of algorithmic decision-making in important areas such as education, welfare and justice requires us to reconsider what fairness means. Reducing fairness to the formal impartiality of code can perpetuate structural inequalities that are divorced from real-world conditions. Building on the ideas of John Rawls and Amartya Sen, this paper puts forward a model of capability-sensitive algorithmic governance that integrates legal duties, adaptive technical architectures and distributed accountability. Rather than merely complying with neutral rules, algorithms should compensate for disparities in converting resources into real opportunities. Fairness must therefore emerge by design as an institutional, contestable and context-aware property. In this way, the algorithm becomes an ethical institution, embedding justifying constraints and assuming public responsibility within a decision-making infrastructure shaped by procedural and distributive justice, thereby redefining the normative status of algorithms as agents of institutional fairness.