A fairness-aware extension of Stochastic Multicriteria Acceptability Analysis for ranking
Researchers have introduced SMAA-Fair, an extension of Stochastic Multicriteria Acceptability Analysis (SMAA) designed to incorporate fairness into ranking problems. This new framework reweights rankings based on group fairness metrics, giving greater importance to fairer outcomes. SMAA-Fair is adaptable to various aggregation models and can utilize different fairness metrics, including Statistical Parity and Kullback-Leibler divergence variants. Experiments demonstrate its ability to improve the representation of protected groups in favorable ranking positions while maintaining robustness against preference uncertainty. AI
IMPACT Introduces a novel method to improve fairness in AI-driven ranking systems.