Kweku Kwegyir-Aggrey, Naveen Durvasula, Jennifer Wang, Suresh Venkatasubramanian
{"title":"当种族无法观测时,观测背景可改进差异估计","authors":"Kweku Kwegyir-Aggrey, Naveen Durvasula, Jennifer Wang, Suresh Venkatasubramanian","doi":"arxiv-2409.01984","DOIUrl":null,"url":null,"abstract":"In many domains, it is difficult to obtain the race data that is required to\nestimate racial disparity. To address this problem, practitioners have adopted\nthe use of proxy methods which predict race using non-protected covariates.\nHowever, these proxies often yield biased estimates, especially for minority\ngroups, limiting their real-world utility. In this paper, we introduce two new\ncontextual proxy models that advance existing methods by incorporating\ncontextual features in order to improve race estimates. We show that these\nalgorithms demonstrate significant performance improvements in estimating\ndisparities on real-world home loan and voter data. We establish that achieving\nunbiased disparity estimates with contextual proxies relies on\nmean-consistency, a calibration-like condition.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"34 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2024-09-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Observing Context Improves Disparity Estimation when Race is Unobserved\",\"authors\":\"Kweku Kwegyir-Aggrey, Naveen Durvasula, Jennifer Wang, Suresh Venkatasubramanian\",\"doi\":\"arxiv-2409.01984\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"In many domains, it is difficult to obtain the race data that is required to\\nestimate racial disparity. To address this problem, practitioners have adopted\\nthe use of proxy methods which predict race using non-protected covariates.\\nHowever, these proxies often yield biased estimates, especially for minority\\ngroups, limiting their real-world utility. In this paper, we introduce two new\\ncontextual proxy models that advance existing methods by incorporating\\ncontextual features in order to improve race estimates. We show that these\\nalgorithms demonstrate significant performance improvements in estimating\\ndisparities on real-world home loan and voter data. We establish that achieving\\nunbiased disparity estimates with contextual proxies relies on\\nmean-consistency, a calibration-like condition.\",\"PeriodicalId\":501112,\"journal\":{\"name\":\"arXiv - CS - Computers and Society\",\"volume\":\"34 1\",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-09-03\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"arXiv - CS - Computers and Society\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/arxiv-2409.01984\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - CS - Computers and Society","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2409.01984","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Observing Context Improves Disparity Estimation when Race is Unobserved
In many domains, it is difficult to obtain the race data that is required to
estimate racial disparity. To address this problem, practitioners have adopted
the use of proxy methods which predict race using non-protected covariates.
However, these proxies often yield biased estimates, especially for minority
groups, limiting their real-world utility. In this paper, we introduce two new
contextual proxy models that advance existing methods by incorporating
contextual features in order to improve race estimates. We show that these
algorithms demonstrate significant performance improvements in estimating
disparities on real-world home loan and voter data. We establish that achieving
unbiased disparity estimates with contextual proxies relies on
mean-consistency, a calibration-like condition.