Procedural Justice in Algorithmic Fairness: Leveraging Transparency and Outcome Control for Fair Algorithmic Mediation

Algorithms increasingly mediate groups in society. Algorithms enable efficient, data-driven, scalable decisions, but cannot account for important social values and contexts accounted for in human decision-making. How can we leverage the strengths of both human and algorithmic decision-making in order to promote fair decisions? To investigate this question, we draw from procedural justice theory and present a framework for procedurally fair algorithmic decision-making. In order to evaluate the framework, we built an interface that leveraged two key elements of the framework and evaluated in the context of a fair-division application in which an algorithm computes allocations according to economic fairness properties. Our interface first explained how the algorithm worked (standards clarity), explained algorithmic outcomes (outcome explanation), then allowed people to discuss and interactively adjust the algorithmic allocations as a group (outcome control). We report qualitative results from our within-subjects laboratory study and implications for algorithmic fairness, transparency, and mediation.