{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/delayed-impact-of-fair-machine-learning","title":"Delayed Impact of Fair Machine Learning","arxiv_id":"1803.04383","date":"2018-03-12","proceeding":"ICML 2018 7","authors":["Lydia T. Liu","Sarah Dean","Esther Rolf","Max Simchowitz","Moritz Hardt"],"abstract":"Fairness in machine learning has predominantly been studied in static\nclassification settings without concern for how decisions change the underlying\npopulation over time. Conventional wisdom suggests that fairness criteria\npromote the long-term well-being of those groups they aim to protect.\n  We study how static fairness criteria interact with temporal indicators of\nwell-being, such as long-term improvement, stagnation, and decline in a\nvariable of interest. We demonstrate that even in a one-step feedback model,\ncommon fairness criteria in general do not promote improvement over time, and\nmay in fact cause harm in cases where an unconstrained objective would not.\n  We completely characterize the delayed impact of three standard criteria,\ncontrasting the regimes in which these exhibit qualitatively different\nbehavior. In addition, we find that a natural form of measurement error\nbroadens the regime in which fairness criteria perform favorably.\n  Our results highlight the importance of measurement and temporal modeling in\nthe evaluation of fairness criteria, suggesting a range of new challenges and\ntrade-offs.","url_abs":"http://arxiv.org/abs/1803.04383v2","url_pdf":"http://arxiv.org/pdf/1803.04383v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"delayed-impact-of-fair-machine-learning","repo_url":"https://github.com/lydiatliu/delayedimpact","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null},{"paper_slug":"delayed-impact-of-fair-machine-learning","repo_url":"https://github.com/mrtzh/whynot","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"delayed-impact-of-fair-machine-learning","repo_url":"https://github.com/zykls/whynot","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"machine-learning","task_name":"BIG-bench Machine Learning"},{"task_slug":"fairness","task_name":"Fairness"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1803.04383","atlas_url":"https://app.syntology.ai/?focus=1803.04383","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}