{"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/strategic-classification","title":"Strategic Classification","arxiv_id":"1506.06980","date":"2015-06-23","proceeding":null,"authors":["Moritz Hardt","Nimrod Megiddo","Christos Papadimitriou","Mary Wootters"],"abstract":"Machine learning relies on the assumption that unseen test instances of a\nclassification problem follow the same distribution as observed training data.\nHowever, this principle can break down when machine learning is used to make\nimportant decisions about the welfare (employment, education, health) of\nstrategic individuals. Knowing information about the classifier, such\nindividuals may manipulate their attributes in order to obtain a better\nclassification outcome. As a result of this behavior---often referred to as\ngaming---the performance of the classifier may deteriorate sharply. Indeed,\ngaming is a well-known obstacle for using machine learning methods in practice;\nin financial policy-making, the problem is widely known as Goodhart's law. In\nthis paper, we formalize the problem, and pursue algorithms for learning\nclassifiers that are robust to gaming.\n  We model classification as a sequential game between a player named \"Jury\"\nand a player named \"Contestant.\" Jury designs a classifier, and Contestant\nreceives an input to the classifier, which he may change at some cost. Jury's\ngoal is to achieve high classification accuracy with respect to Contestant's\noriginal input and some underlying target classification function. Contestant's\ngoal is to achieve a favorable classification outcome while taking into account\nthe cost of achieving it.\n  For a natural class of cost functions, we obtain computationally efficient\nlearning algorithms which are near-optimal. Surprisingly, our algorithms are\nefficient even on concept classes that are computationally hard to learn. For\ngeneral cost functions, designing an approximately optimal strategy-proof\nclassifier, for inverse-polynomial approximation, is NP-hard.","url_abs":"http://arxiv.org/abs/1506.06980v2","url_pdf":"http://arxiv.org/pdf/1506.06980v2.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":"strategic-classification","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":"strategic-classification","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":"classification-1","task_name":"Classification"},{"task_slug":"classification","task_name":"General Classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1506.06980","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}