{"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/calibration-for-the-computationally","title":"Calibration for the (Computationally-Identifiable) Masses","arxiv_id":"1711.08513","date":"2017-11-22","proceeding":null,"authors":["Úrsula Hébert-Johnson","Michael P. Kim","Omer Reingold","Guy N. Rothblum"],"abstract":"As algorithms increasingly inform and influence decisions made about\nindividuals, it becomes increasingly important to address concerns that these\nalgorithms might be discriminatory. The output of an algorithm can be\ndiscriminatory for many reasons, most notably: (1) the data used to train the\nalgorithm might be biased (in various ways) to favor certain populations over\nothers; (2) the analysis of this training data might inadvertently or\nmaliciously introduce biases that are not borne out in the data. This work\nfocuses on the latter concern.\n  We develop and study multicalbration -- a new measure of algorithmic fairness\nthat aims to mitigate concerns about discrimination that is introduced in the\nprocess of learning a predictor from data. Multicalibration guarantees accurate\n(calibrated) predictions for every subpopulation that can be identified within\na specified class of computations. We think of the class as being quite rich;\nin particular, it can contain many overlapping subgroups of a protected group.\n  We show that in many settings this strong notion of protection from\ndiscrimination is both attainable and aligned with the goal of obtaining\naccurate predictions. Along the way, we present new algorithms for learning a\nmulticalibrated predictor, study the computational complexity of this task, and\ndraw new connections to computational learning models such as agnostic\nlearning.","url_abs":"http://arxiv.org/abs/1711.08513v2","url_pdf":"http://arxiv.org/pdf/1711.08513v2.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":"calibration-for-the-computationally","repo_url":"https://github.com/cavalab/pmcboost","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"calibration-for-the-computationally","repo_url":"https://github.com/sid-devic/multicalibration","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"fairness","task_name":"Fairness"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1711.08513","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}