Optimal Transport under Group Fairness Constraints
Researchers have developed a new framework for ensuring fairness in matching algorithms, specifically within the context of Optimal Transport (OT). Their work introduces a novel group fairness constraint that targets the probability of matches between individuals from different groups. The study proposes a modified Sinkhorn algorithm for efficient computation of perfectly fair plans and explores relaxation strategies to balance fairness with matching quality, including penalized OT and bilevel optimization. AI
IMPACT Introduces a novel algorithmic approach to fairness in matching, potentially impacting resource allocation systems.