[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-127721-en":3,"doc-seo-127721-105":30,"detail-sidebar-cat-0-en-105":92},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},127721,962084928432,"Emma Wilson","https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d",8,"Research & Report","Generalized Disparate Impact for Configurable Fairness Solutions in ML - read online","Generalized Disparate Impact (GeDI) proposes fairness indicators for machine learning with continuous protected attributes. It analyzes limits of the Hirschfeld-Gebelein-Renyi (HGR) indicator in semantics, interpretability, and robustness for finite samples, and its inability to capture certain scale effects. GeDI introduces a configurable family of interpretable, transparent, and sample-robust indicators complementary to HGR, supporting fine-grained constraints on allowed versus forbidden dependence.","Generalized Disparate Impact for Configurable Fairness Solutions in ML  \nLuca Giuliani * 1 Eleonora Misino * 1 Michele Lombardi 1  \nAbstract  \nWe make two contributions in the field of AI fairness over continuous protected attributes. First, we show that the Hirschfeld-Gebelein-Renyi (HGR) indicator (the only one currently available for such a case) is valuable but subject to a few crucial limitations regarding semantics, interpretability, and robustness. Second, we introduce a family of indicators that are: 1) complementary to HGR in terms of semantics; 2) fully interpretable and transparent; 3) robust over finite samples; 4) configurable to suit specific applications. Our approach also allows us to define fine-grained constraints to permit certain types of dependence and forbid others selectively. By expanding the available options for continuous protected attributes, our approach represents a significant contribution to the area of fair artificial intelligence.  \n1. Introduction  \nIn recent years, the social impact of data-driven AI systems and its ethical implications have become widely recognized. For example, models may discriminate over population groups (Julia et al., 2016 ; Gianfrancesco et al., 2018), spurring extensive research on AI fairness. Typical approaches in this area involve quantitative indicators defined over a “protected attribute”, which can be used to measure discrimination or enforce fairness constraints. On the one hand, such metrics are arguably the most viable solution for mitigating fairness issues; on the other hand, the nuances of ethics can hardly be reduced to simple rules. From this point of view, the availability of multiple and diverse metrics is a significant asset since it enables choosing the best indicator depending on the application.  \n*Equal contribution 1Department of Computer Science and Engineering, University of Bologna, Bologna, Italy. Correspondence to: Luca Giuliani \u003C[luca.giuliani13@unibo.it](luca.giuliani13@unibo.it) >, Eleonora Misino \u003C[eleonora.misino2@unibo.it](eleonora.misino2@unibo.it) > .  \nProceedings of the 40 th International Conference on Machine Learning, Honolulu, Hawaii, USA. PMLR 202, 2023 . Copyright 2023 by the author(s) .  \nRegarding available solutions, the case of categorical protected attributes is well covered by multiple indicators (see Section 2) . Conversely, a single approach works with continuous protected attributes at the moment of writing; this is the Hirschfeld-Gebelein-Renyi (HGR) correlation coefficient (Rnyi, 1959), which has two viable implementations for Machine Learning (ML) systems (Mary et al., 2019 ; Grari et al., 2020) .  \nWe view the lack of diverse techniques for continuous protected groups as a major issue. We contribute to this area by 1) identifying a few critical limitations in the HGR approach, and 2) introducing a family of indicators that complement HGR semantics and have technical advantages.  \nIn terms of limitations, we highlight how the theoretical HGR formulation is prone to pathological behavior for finite samples, leading to the oversized importance of implementation details and limited interpretability. Moreover, the generality of the HGR formulation makes the indicator unsuitable for exclusively measuring the functional dependency between the protected attribute and the target. Finally, the HGR indicator cannot account for scale effects on fairness since it is based on the scale-invariant Pearson’s correlation coefficient.  \nWe introduce the Generalized Disparate Impact (GeDI), a family of indicators inspired by the HGR approach and by the Disparate Impact Discrimination Index (Aghaei et al., 2019) . GeDI indicators measure the dependency based on how well a user-specified function of the protected attribute can approximate the target variable. Our indicators support both discrete and continuous protected attributes and 1) complement the HGR semantics, 2) are fully interpretable, 3) are robust for finite samples, and 4","cbCaiovzmVu64Rem","https://ap.wps.com/l/cbCaiovzmVu64Rem","pdf",821138,1,16,"English","en",105,"# Introduction\n## Fairness metrics and protected attributes\n## Limitations of HGR for continuous groups\n## GeDI contribution and configurable indicators\n# Background and Motivation\n## Pre-processing, postprocessing, and in-processing fairness\n## Disparate impact and ML extension\n## Continuous protected attributes and HGR-based methods","[{\"question\":\"What problem does the document address in AI fairness with continuous protected attributes?\",\"answer\":\"It addresses the lack of diverse, reliable fairness metrics for continuous protected groups, where current practice mainly relies on the HGR correlation coefficient.\"},{\"question\":\"What limitations are identified for the HGR indicator?\",\"answer\":\"The document highlights semantic and interpretability issues, pathological behavior for finite samples, limited suitability for measuring functional dependency, and insensitivity to scale effects due to its use of a scale-invariant Pearson correlation.\"},{\"question\":\"How does GeDI improve fairness evaluation and constraint design?\",\"answer\":\"GeDI introduces a configurable family of indicators that are complementary to HGR, fully interpretable, robust for finite samples, and able to enforce fine-grained constraints by allowing certain dependence types while forbidding others.\"}]","Generalized Disparate Impact for Configurable Fairness Solutions in ML - read online | PDF",1785941195,40,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":87,"head_meta":89,"extra_data":91,"updated_unix":28},"generalized-disparate-impact-for-configurable-fairness-solutions-in-ml-read-online","",{"@graph":36,"@context":86},[37,54,69],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/generalized-disparate-impact-for-configurable-fairness-solutions-in-ml-read-online/127721/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-22","2026-08-05",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72,78,82],{"name":73,"@type":74,"acceptedAnswer":75},"What problem does the document address in AI fairness with continuous protected attributes?","Question",{"text":76,"@type":77},"It addresses the lack of diverse, reliable fairness metrics for continuous protected groups, where current practice mainly relies on the HGR correlation coefficient.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"What limitations are identified for the HGR indicator?",{"text":81,"@type":77},"The document highlights semantic and interpretability issues, pathological behavior for finite samples, limited suitability for measuring functional dependency, and insensitivity to scale effects due to its use of a scale-invariant Pearson correlation.",{"name":83,"@type":74,"acceptedAnswer":84},"How does GeDI improve fairness evaluation and constraint design?",{"text":85,"@type":77},"GeDI introduces a configurable family of indicators that are complementary to HGR, fully interpretable, robust for finite samples, and able to enforce fine-grained constraints by allowing certain dependence types while forbidding others.","https://schema.org",{"og:url":52,"og:type":88,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":90,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":93},[94,98,102,106,111,116,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":103,"show_sort_weight":104,"slug":105},"Exam",70,"exam",{"id":107,"doc_module":4,"doc_module_name":46,"category_name":108,"show_sort_weight":109,"slug":110},5,"Comic",60,"comic",{"id":112,"doc_module":4,"doc_module_name":46,"category_name":113,"show_sort_weight":114,"slug":115},6,"Technology",50,"technology",{"id":117,"doc_module":4,"doc_module_name":46,"category_name":118,"show_sort_weight":29,"slug":119},7,"Healthcare","healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":107,"slug":138},19,"General","general"]