2019/10/22 by Hanchen Wang, Wang, Hanchen, Nina Grgić-Hlača +7 · 2 citations
Medicine · Neuroscience · Social Sciences · #Artificial Intelligence (cs.AI) #Artificial Intelligence in Healthcare and Education #Computers and Society (cs.CY) #Ethics and Social Impacts of AI #Ethics in Clinical Research #FOS: Computer and information sciences #Machine Learning (cs.LG) #Psychology of Moral and Emotional Judgment
paper · pdf · doi:10.48550/arxiv.1910.10255
openalex publication_date 2019/10/22 · openalex created_date 2022/07/28 · openalex updated_date 2026/07/28
The notion of individual fairness requires that similar people receive\nsimilar treatment. However, this is hard to achieve in practice since it is\ndifficult to specify the appropriate similarity metric. In this work, we\nattempt to learn such similarity metric from human annotated data. We gather a\nnew dataset of human judgments on a criminal recidivism prediction (COMPAS)\ntask. By assuming the human supervision obeys the principle of individual\nfairness, we leverage prior work on metric learning, evaluate the performance\nof several metric learning methods on our dataset, and show that the learned\nmetrics outperform the Euclidean and Precision metric under various criteria.\nWe do not provide a way to directly learn a similarity metric satisfying the\nindividual fairness, but to provide an empirical study on how to derive the\nsimilarity metric from human supervisors, then future work can use this as a\ntool to understand human supervision.\n