A GA-based algorithm meets the fair ranking problem
作者：
Highlights：
• Presenting FARGO which is a new evolutionary-based algorithm aiming at tackling the fair ranking problem. The proposed algorithm not only effectively supports one protected group but also achieves near-optimal solutions for any number of protected groups.
• Introducing EGR, abbreviated Expected Gain Ratio, which is a novel evaluation metric for measuring the quality of the fair ranking problem solutions.
• Introducing a sophisticated objective function that computes total fairness concerns in a reasonable time.
• Running plenty of experimental analysis in order to evaluate the performance of the proposed algorithm and compare it with its counterpart.
摘要
•Presenting FARGO which is a new evolutionary-based algorithm aiming at tackling the fair ranking problem. The proposed algorithm not only effectively supports one protected group but also achieves near-optimal solutions for any number of protected groups.•Introducing EGR, abbreviated Expected Gain Ratio, which is a novel evaluation metric for measuring the quality of the fair ranking problem solutions.•Introducing a sophisticated objective function that computes total fairness concerns in a reasonable time.•Running plenty of experimental analysis in order to evaluate the performance of the proposed algorithm and compare it with its counterpart.
论文关键词：Ranking,Fair ranking,Bias prevention,Genetic algorithm,Fitness function,Simulated annealing
论文评审过程：Received 2 December 2020, Revised 19 July 2021, Accepted 27 July 2021, Available online 13 September 2021, Version of Record 13 September 2021.
论文官网地址：https://doi.org/10.1016/j.ipm.2021.102711
