“Visibly Fair Mechanisms”

Inácio Bo

University of Macau

Abstract: We study allocation problems where agents are assigned positions under a strict priority order (e.g., exam ranks). While traditional fairness---elimination of justified envy---leads to Serial Dictatorship (SD) outcomes with complete preference elicitation, SD outcomes might not satisfy policy goals. Motivated by reforms in the Indian Civil Services and U.S. Army, we introduce the concept of visible fairness and show that restricting message spaces can broaden the range of outcomes that are visibly fair. We provide characterizations of visibly fair mechanisms as generalizations of Serial Dictatorship accommodating partial preferences and provide conditions for these to be strategy-proof. Building on these insights, we propose mechanisms enforcing modular upper-bounds, which integrate distributional objectives via type-specific caps.