{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,10,16]],"date-time":"2025-10-16T00:46:31Z","timestamp":1760575591747,"version":"build-2065373602"},"reference-count":0,"publisher":"Association for the Advancement of Artificial Intelligence (AAAI)","issue":"1","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["AIES"],"abstract":"<jats:p>As machine learning (ML) algorithms are increasingly used\nin social domains to make predictions about humans, there\nis a growing concern that these algorithms may exhibit\nbiases against certain social groups. Numerous notions of\nfairness have been proposed in the literature to measure\nthe unfairness of ML. Among them, one class that receives\nthe most attention is parity-based, i.e., achieving\nfairness by equalizing treatment or outcomes for different\nsocial groups. However, achieving parity-based fairness\noften comes at the cost of lowering model accuracy and is\nundesirable for many high-stakes domains like healthcare.\nTo avoid inferior accuracy, a line of research focuses on\npreference-based fairness, under which any group of\nindividuals would experience the highest accuracy and\ncollectively prefer the ML outcomes assigned to them if\nthey were given the choice between various sets of\noutcomes. However, these works assume individual\ndemographic information is known and fully accessible\nduring training. In this paper, we relax this requirement\nand propose a novel demographic-agnostic fairness without\nharm (DAFH) optimization algorithm, which jointly learns a\ngroup classifier that partitions the population into\nmultiple groups and a set of decoupled classifiers\nassociated with these groups. Theoretically, we conduct\nsample complexity analysis and show that our method can\noutperform the baselines when demographic information is\nknown and used to train decoupled classifiers. Experiments\non both synthetic and real data validate the proposed\nmethod.<\/jats:p>","DOI":"10.1609\/aies.v8i1.36565","type":"journal-article","created":{"date-parts":[[2025,10,15]],"date-time":"2025-10-15T13:18:41Z","timestamp":1760534321000},"page":"486-497","source":"Crossref","is-referenced-by-count":0,"title":["Demographic-Agnostic Fairness Without Harm"],"prefix":"10.1609","volume":"8","author":[{"given":"Zhongteng","family":"Cai","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Mohammad Mahdi","family":"Khalili","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Xueru","family":"Zhang","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"9382","published-online":{"date-parts":[[2025,10,15]]},"container-title":["Proceedings of the AAAI\/ACM Conference on AI, Ethics, and Society"],"original-title":[],"link":[{"URL":"https:\/\/ojs.aaai.org\/index.php\/AIES\/article\/download\/36565\/38703","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/ojs.aaai.org\/index.php\/AIES\/article\/download\/36565\/38703","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,10,15]],"date-time":"2025-10-15T13:18:41Z","timestamp":1760534321000},"score":1,"resource":{"primary":{"URL":"https:\/\/ojs.aaai.org\/index.php\/AIES\/article\/view\/36565"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,15]]},"references-count":0,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2025,10,15]]}},"URL":"https:\/\/doi.org\/10.1609\/aies.v8i1.36565","relation":{},"ISSN":["3065-8365"],"issn-type":[{"value":"3065-8365","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,10,15]]}}}