{"publication":"2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition","citation":{"ieee":"D. Zietlow et al., “Leveling down in computer vision: Pareto inefficiencies in fair deep classifiers,” in 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, New Orleans, LA, United States, 2022, pp. 10400–10411.","ama":"Zietlow D, Lohaus M, Balakrishnan G, et al. Leveling down in computer vision: Pareto inefficiencies in fair deep classifiers. In: 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Institute of Electrical and Electronics Engineers; 2022:10400-10411. doi:10.1109/cvpr52688.2022.01016","mla":"Zietlow, Dominik, et al. “Leveling down in Computer Vision: Pareto Inefficiencies in Fair Deep Classifiers.” 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Institute of Electrical and Electronics Engineers, 2022, pp. 10400–11, doi:10.1109/cvpr52688.2022.01016.","short":"D. Zietlow, M. Lohaus, G. Balakrishnan, M. Kleindessner, F. Locatello, B. Scholkopf, C. Russell, in:, 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Institute of Electrical and Electronics Engineers, 2022, pp. 10400–10411.","ista":"Zietlow D, Lohaus M, Balakrishnan G, Kleindessner M, Locatello F, Scholkopf B, Russell C. 2022. Leveling down in computer vision: Pareto inefficiencies in fair deep classifiers. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. CVPR: Conference on Computer Vision and Pattern Recognition, 10400–10411.","apa":"Zietlow, D., Lohaus, M., Balakrishnan, G., Kleindessner, M., Locatello, F., Scholkopf, B., & Russell, C. (2022). Leveling down in computer vision: Pareto inefficiencies in fair deep classifiers. In 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 10400–10411). New Orleans, LA, United States: Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/cvpr52688.2022.01016","chicago":"Zietlow, Dominik, Michael Lohaus, Guha Balakrishnan, Matthaus Kleindessner, Francesco Locatello, Bernhard Scholkopf, and Chris Russell. “Leveling down in Computer Vision: Pareto Inefficiencies in Fair Deep Classifiers.” In 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 10400–411. Institute of Electrical and Electronics Engineers, 2022. https://doi.org/10.1109/cvpr52688.2022.01016."},"day":"01","title":"Leveling down in computer vision: Pareto inefficiencies in fair deep classifiers","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","doi":"10.1109/cvpr52688.2022.01016","_id":"14114","author":[{"full_name":"Zietlow, Dominik","last_name":"Zietlow","first_name":"Dominik"},{"last_name":"Lohaus","full_name":"Lohaus, Michael","first_name":"Michael"},{"first_name":"Guha","full_name":"Balakrishnan, Guha","last_name":"Balakrishnan"},{"last_name":"Kleindessner","full_name":"Kleindessner, Matthaus","first_name":"Matthaus"},{"last_name":"Locatello","id":"26cfd52f-2483-11ee-8040-88983bcc06d4","full_name":"Locatello, Francesco","first_name":"Francesco","orcid":"0000-0002-4850-0683"},{"first_name":"Bernhard","last_name":"Scholkopf","full_name":"Scholkopf, Bernhard"},{"last_name":"Russell","full_name":"Russell, Chris","first_name":"Chris"}],"article_processing_charge":"No","department":[{"_id":"FrLo"}],"publisher":"Institute of Electrical and Electronics Engineers","type":"conference","main_file_link":[{"open_access":"1","url":"https://arxiv.org/abs/2203.04913"}],"status":"public","date_created":"2023-08-21T12:18:00Z","year":"2022","page":"10400-10411","external_id":{"arxiv":["2203.04913"]},"publication_identifier":{"issn":["1063-6919"],"eissn":["2575-7075"],"isbn":["9781665469470"]},"oa":1,"language":[{"iso":"eng"}],"date_updated":"2023-09-11T09:19:14Z","conference":{"location":"New Orleans, LA, United States","start_date":"2022-06-18","end_date":"2022-06-24","name":"CVPR: Conference on Computer Vision and Pattern Recognition"},"date_published":"2022-07-01T00:00:00Z","month":"07","quality_controlled":"1","oa_version":"Preprint","extern":"1","abstract":[{"lang":"eng","text":"Algorithmic fairness is frequently motivated in terms of a trade-off in which overall performance is decreased so as to improve performance on disadvantaged groups where the algorithm would otherwise be less accurate. Contrary to this, we find that applying existing fairness approaches to computer vision improve fairness by degrading the performance of classifiers across all groups (with increased degradation on the best performing groups). Extending the bias-variance decomposition for classification to fairness, we theoretically explain why the majority of fairness methods designed for low capacity models should not be used in settings involving high-capacity models, a scenario common to computer vision. We corroborate this analysis with extensive experimental support that shows that many of the fairness heuristics used in computer vision also degrade performance on the most disadvantaged groups. Building on these insights, we propose an adaptive augmentation strategy that, uniquely, of all methods tested, improves performance for the disadvantaged groups."}],"publication_status":"published","scopus_import":"1"}