Abstract
Four definitions of “cultural fairness” are examined and found to be not only mutually contradictory (for reasons which are explained), but all based on the false view that optimum treatment of cultural factors in test construction or test selection can be reduced to completely mechanical procedures. If a conflict arises between the two goals of maximizing a test's validity and minimizing the test's discrimination against certain cultural groups, then a subjective, policy‐level decision must be made concerning the relative importance of the two goals. The terms in which this judgment should be made are described, and methods are described for entering the result of this judgment into mechanical procedures for constructing a “culturally optimum” test. Such a test will not necessarily fit any of the four definitions of “cultural fairness.”
Cited by
168 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献
1. Fair Learning by Model Averaging;Risk and Decision Analysis;2025-03-02
2. Surrogate Modeling to Address the Absence of Protected Membership Attributes in Fairness Evaluation;ACM Transactions on Evolutionary Learning and Optimization;2024-10-11
3. When causality meets fairness: A survey;Journal of Logical and Algebraic Methods in Programming;2024-10
4. Your Neighbor Matters: Towards Fair Decisions Under Networked Interference;Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining;2024-08-24
5. Fairness in Machine Learning: A Survey;ACM Computing Surveys;2024-04-09