[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-125638-en":3,"doc-seo-125638-105":30,"detail-sidebar-cat-0-en-105":91},{"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":4,"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},125638,962075006959,"Anda","https://ap-avatar.wpscdn.com/avatar/e0002397efbe92a78e?_k=1776741047341049297",8,"Research & Report","Causality-Aided Trade-off Analysis for Machine Learning Fairness","Machine learning fairness has attracted growing interest, yet current research lacks a systematic view of how trade-offs arise among factors inside the ML pipeline when fairness-improving methods are applied. Analyzing these trade-offs is difficult due to multiple interacting fairness parameters and competing metrics. This paper applies causality analysis to identify the causal relations driving trade-offs, offering domain optimizations for causal discovery and a unified interface for trade-off evaluation across real-world datasets.","Causality-Aided Trade-off Analysis for Machine  \nLearning Fairness  \nZhenlan Ji∗ , Pingchuan Ma∗‡, Shuai Wang∗‡, and Yanhui Li†  \n∗ The Hong Kong University of Science and Technology, {zjiae, pmaab, [shuaiw](shuaiw}@cse.ust.hk)[}](shuaiw}@cse.ust.hk)[@cse.ust.hk](shuaiw}@cse.ust.hk)[ ](shuaiw}@cse.ust.hk)†State Key Laboratory for Novel Software Technology, Nanjing University, [yanhuili@nju.edu.cn](yanhuili@nju.edu.cn)  \narXiv :2305 . 13057v2 [ cs .LG] 22 Aug 2023  \nAbstract—There has been an increasing interest in enhancing the fairness of machine learning (ML). Despite the growing number of fairness-improving methods, we lack a systematic understanding of the trade-offs among factors considered in the ML pipeline when fairness-improving methods are applied. This understanding is essential for developers to make informed decisions regarding the provision of fair ML services. Nonetheless, it is extremely difficult to analyze the trade-offs when there are multiple fairness parameters and other crucial metrics involved, coupled, and even in conflict with one another.  \nThis paper uses causality analysis as a principled method for analyzing trade-offs between fairness parameters and other crucial metrics in ML pipelines. To practically and effectively conduct causality analysis, we propose a set of domain-specific optimizations to facilitate accurate causal discovery and a unified, novel interface for trade-off analysis based on well-established causal inference methods. We conduct a comprehensive empirical study using three real-world datasets on a collection of widelyused fairness-improving techniques. Our study obtains actionable suggestions for users and developers of fair ML. We further demonstrate the versatile usage of our approach in selecting the optimal fairness-improving method, paving the way for more ethical and socially responsible AI technologies.  \nI. INTRODUCTION  \nMachine learning (ML) techniques are now essential for everyday applications in safety-critical domains like credit risk evaluation [1] and criminal justice [2] . However, ML models have exhibited inherent biases [3, 4], leading to realworld consequences such as discriminatory outcomes between privileged and underprivileged groups [5–7] . To address this, various fairness-improving methods have been proposed and studied by the software engineering (SE) community, including mitigating unfairness through data processing [4, 8, 9], model modification [10, 11], or prediction alteration [3] .  \nDespite the significant progress made, an important question arises: what are the trade-offs made by these fairness-improving methods in the ML pipeline? It is a widely held belief that there exists a trade-off between fairness and the functional quality properties of the ML pipeline, such as the ML performance and the ML model robustness. In general, empirical studies from the SE community and theoretical analyses from the ML community have demonstrated that optimizing for performance may come at the cost of fairness, and vice versa [8, 9, 12–14] . Furthermore, many metrics concentrating on fairness, such as group fairness and individual fairness, are inherently incompatible [13] . These trade-offs render additional complexity to the  \n‡ Corresponding authors  \nprocess of improving fairness in ML systems and are not well understood. As a result, the lack of transparent and manageable trade-off analyses makes it challenging for developers to make informed decisions within the ML pipeline.  \nTo understand trade-offs, it is essential to comprehend the interactions among fairness-improving methods as well as different metrics. More importantly, it is crucial to “disentangle” the true cause-effect relationships from the observed correlations. For example, a fairness-improving method may simultaneously affect the model’s fairness on both the training set and the test set. However, it is unclear how would training fairness affect test fairness due to the confounding influen","cbCairol7lZVqljp","https://ap.wps.com/l/cbCairol7lZVqljp","pdf",1474740,1,13,"English","en",105,"# Introduction\n## Fairness improvement methods and bias\n## Trade-offs in the ML pipeline\n## Causality analysis for disentangling causes\n## Challenges in causal discovery and inference","[{\"question\":\"Why is it difficult to analyze trade-offs in ML fairness improvement methods?\",\"answer\":\"Multiple fairness parameters and other metrics can interact, even conflict, making it hard to understand how improving one aspect affects others in the pipeline.\"},{\"question\":\"What approach does the paper propose for trade-off analysis?\",\"answer\":\"It uses causality analysis to study causal relationships between fairness parameters and other crucial metrics, rather than relying only on observed correlations.\"},{\"question\":\"How does the paper make causality analysis practical for ML pipelines?\",\"answer\":\"It introduces domain-specific optimizations for accurate causal discovery and a unified interface that reformulates trade-off analysis into standard causal inference queries.\"}]","Causality-Aided Trade-off Analysis for Machine Learning Fairness | PDF",1785900348,33,{"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":86,"head_meta":88,"extra_data":90,"updated_unix":28},"causality-aided-trade-off-analysis-for-machine-learning-fairness","",{"@graph":36,"@context":85},[37,54,68],{"@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/causality-aided-trade-off-analysis-for-machine-learning-fairness/125638/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-05",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why is it difficult to analyze trade-offs in ML fairness improvement methods?","Question",{"text":75,"@type":76},"Multiple fairness parameters and other metrics can interact, even conflict, making it hard to understand how improving one aspect affects others in the pipeline.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What approach does the paper propose for trade-off analysis?",{"text":80,"@type":76},"It uses causality analysis to study causal relationships between fairness parameters and other crucial metrics, rather than relying only on observed correlations.",{"name":82,"@type":73,"acceptedAnswer":83},"How does the paper make causality analysis practical for ML pipelines?",{"text":84,"@type":76},"It introduces domain-specific optimizations for accurate causal discovery and a unified interface that reformulates trade-off analysis into standard causal inference queries.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"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":106,"slug":138},19,"General","general"]