Technical fairness and social legitimacy: a conceptual mapping of algorithmic fairness, procedural justice, and explainable AI

Algorithmic fairness that conforms to mathematical criteria can sometimes be considered unfair by affected communities, undermining legitimacy in critical domains. This study explains this “perception gap” through the proposed procedural deficit hypothesis, in which technical fairness approaches often emphasize distributive outcomes while giving less attention to procedural justice, including transparency, contestability, and correctability. It develops a conceptual mapping that links statistical fairness metrics to perceived fairness dimensions and identifies scope conditions, such as ground-truth legitimacy, under which these links are likely to hold or break down. To mitigate these risks, this study introduces a fairness-by-design framework that integrates stakeholder involvement and explainability into the development lifecycle. It operationalizes this framework through the Fairness Process Card, a practical tool for documenting procedural justice mechanisms. The framework identifies procedural considerations that may complement statistical fairness assessment in AI system design.