“Voiceless”: the procedural gap in algorithmic justice

Abstract

Abstract Conversations about algorithms and fairness can overlook the bifurcation of justice. Whilst distributive justice is integral to accountability for algorithmic bias and related harms, without procedural justice and ‘voice’, defined as consent, transparency, and a right to be heard, algorithmic decisions will fail to find legitimacy and acceptance. Fairness metrics respond better to distributive justice because algorithmic bias is primarily framed in such terms. The potential of AI Ethics Guidelines and formal regulation to address voice is compromised by ambiguous concepts of fairness, rigidity, and competing market demands. The scope of a right to a fair trial limits the potential of human rights impact assessments to address procedural rights. Jurisdictional issues of reviewability, interpretation, and applicability challenge judicial review. In examining the capacity of these accountability mechanisms to ensure ‘voice’, a ‘procedural gap’ emerges in algorithmic decision-making.