{"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/mitigating-unwanted-biases-with-adversarial","title":"Mitigating Unwanted Biases with Adversarial Learning","arxiv_id":"1801.07593","date":"2018-01-22","proceeding":null,"authors":["Brian Hu Zhang","Blake Lemoine","Margaret Mitchell"],"abstract":"Machine learning is a tool for building models that accurately represent\ninput training data. When undesired biases concerning demographic groups are in\nthe training data, well-trained models will reflect those biases. We present a\nframework for mitigating such biases by including a variable for the group of\ninterest and simultaneously learning a predictor and an adversary. The input to\nthe network X, here text or census data, produces a prediction Y, such as an\nanalogy completion or income bracket, while the adversary tries to model a\nprotected variable Z, here gender or zip code.\n  The objective is to maximize the predictor's ability to predict Y while\nminimizing the adversary's ability to predict Z. Applied to analogy completion,\nthis method results in accurate predictions that exhibit less evidence of\nstereotyping Z. When applied to a classification task using the UCI Adult\n(Census) Dataset, it results in a predictive model that does not lose much\naccuracy while achieving very close to equality of odds (Hardt, et al., 2016).\nThe method is flexible and applicable to multiple definitions of fairness as\nwell as a wide range of gradient-based learning models, including both\nregression and classification tasks.","url_abs":"http://arxiv.org/abs/1801.07593v1","url_pdf":"http://arxiv.org/pdf/1801.07593v1.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":"mitigating-unwanted-biases-with-adversarial","repo_url":"https://github.com/FairUnlearn/detoxai","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"mitigating-unwanted-biases-with-adversarial","repo_url":"https://github.com/bspanjers/AIFBAREND","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"mitigating-unwanted-biases-with-adversarial","repo_url":"https://github.com/xmpuspus/parity-fairness","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"mitigating-unwanted-biases-with-adversarial","repo_url":"https://github.com/giandos200/Zhang-et-al-Mitigating-Unwanted-Biases-with-Adversarial-Learning","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"fairness","task_name":"Fairness"},{"task_slug":"classification","task_name":"General Classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1801.07593","atlas_url":"https://app.syntology.ai/?focus=1801.07593","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}