FairMC Fair–Markov Chain Rank Aggregation Methods

Given a set of voters’ preferences expressed as rankings, rank aggregation approaches combine them into a unique consensus ranking. Even if some methods guarantee fair representation of single voters, they still overlook unfair biases potentially affecting individuals from marginalized groups in the original rankings. Rank aggregation is employed in many high-stakes decision-making processes, hence, due to the high societal influence, the development and study of fair rank aggregation approaches is essential. We introduce FairMC, a new fair rank aggregation approach based on Markov Chains to derive consensus rankings. By modifying the transition matrix and enforcing fairness in the transition probabilities across the groups, we obtain fairness in the representation of the items at the ranking level. The resulting ranking assures higher visibility for protected groups while being close to the original rankings.