[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-272573-105":3,"detail-sidebar-cat-0-en-105":73,"doc-detail-272573-en":122},{"code":4,"msg":5,"data":6},0,"ok",{"site_id":7,"language":8,"slug":9,"title":10,"keywords":11,"description":12,"schema_data":13,"social_meta":66,"head_meta":68,"extra_data":70,"updated_unix":72},105,"en","algorithmic-fairness-verification-with-graphical-models-efficient-fairness-verifier-fvgm","Algorithmic Fairness Verification with Graphical Models - Efficient Fairness Verifier FVGM","","Machine learning models increasingly influence safety-critical and high-stakes decisions, making algorithmic fairness essential. The document studies bias in ML classifiers with respect to multiple fairness definitions and the probability distribution of input features, not just finite datasets. It highlights limitations of existing fairness verifiers in accuracy due to imprecise feature-correlation modeling and in scalability from restrictive SSAT/SMT encodings or sampling. It proposes FVGM, encoding correlations as a Bayesian network and using a stochastic subset-sum approach for linear classifiers.",{"@graph":14,"@context":65},[15,34,56],{"@type":16,"itemListElement":17},"BreadcrumbList",[18,23,27,31],{"item":19,"name":20,"@type":21,"position":22},"https://docshare.wps.com","Home","ListItem",1,{"item":24,"name":25,"@type":21,"position":26},"https://docshare.wps.com/document/","Document",2,{"item":28,"name":29,"@type":21,"position":30},"https://docshare.wps.com/document/research-report/","Research & Report",3,{"item":32,"name":10,"@type":21,"position":33},"https://docshare.wps.com/document/algorithmic-fairness-verification-with-graphical-models-efficient-fairness-verifier-fvgm/272573/",4,{"url":32,"name":10,"@type":35,"image":36,"author":41,"headline":10,"publisher":44,"fileFormat":47,"inLanguage":8,"description":12,"dateModified":48,"datePublished":49,"encodingFormat":47,"isAccessibleForFree":50,"interactionStatistic":51},"DigitalDocument",{"url":37,"@type":38,"width":39,"height":40},"https://docshare.wps.com/thumbnails/algorithmic-fairness-verification-with-graphical-models-efficient-fairness-verifier-fvgm/272573.png","ImageObject",300,407,{"name":42,"@type":43},"Jasmine","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-10-30","2026-09-15",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",10,{"@type":57,"mainEntity":58},"FAQPage",[59],{"name":60,"@type":61,"acceptedAnswer":62},"What is FVGM and how does it improve fairness verification?","Question",{"text":63,"@type":64},"FVGM is an efficient fairness verifier that encodes feature correlations using a Bayesian network. It verifies linear classifiers using a stochastic subset-sum approach, enabling accurate and scalable assessment across fairness metrics and fairness-enhancing algorithms or attacks.","Answer","https://schema.org",{"og:url":32,"og:type":67,"og:title":10,"og:site_name":45,"og:description":12},"article",{"robots":69,"canonical":32},"index,follow",{"doc_id":71,"site_id":7},272573,1789466160,{"code":4,"msg":74,"data":75},"success",[76,80,84,88,93,98,103,107,112,115,118],{"id":22,"doc_module":4,"doc_module_name":25,"category_name":77,"show_sort_weight":78,"slug":79},"Story & Novel",90,"story-novel",{"id":26,"doc_module":4,"doc_module_name":25,"category_name":81,"show_sort_weight":82,"slug":83},"Literature",80,"literature",{"id":33,"doc_module":4,"doc_module_name":25,"category_name":85,"show_sort_weight":86,"slug":87},"Exam",70,"exam",{"id":89,"doc_module":4,"doc_module_name":25,"category_name":90,"show_sort_weight":91,"slug":92},5,"Comic",60,"comic",{"id":94,"doc_module":4,"doc_module_name":25,"category_name":95,"show_sort_weight":96,"slug":97},6,"Technology",50,"technology",{"id":99,"doc_module":4,"doc_module_name":25,"category_name":100,"show_sort_weight":101,"slug":102},7,"Healthcare",40,"healthcare",{"id":104,"doc_module":4,"doc_module_name":25,"category_name":29,"show_sort_weight":105,"slug":106},8,30,"research-report",{"id":108,"doc_module":4,"doc_module_name":25,"category_name":109,"show_sort_weight":110,"slug":111},9,"Religion & Spirituality",20,"religion-spirituality",{"id":110,"doc_module":4,"doc_module_name":25,"category_name":113,"show_sort_weight":110,"slug":114},"World Cup","world-cup",{"id":55,"doc_module":4,"doc_module_name":25,"category_name":116,"show_sort_weight":55,"slug":117},"Lifestyle","lifestyle",{"id":119,"doc_module":4,"doc_module_name":25,"category_name":120,"show_sort_weight":89,"slug":121},19,"General","general",{"code":4,"msg":74,"data":123},{"doc_id":71,"user_id":124,"nickname":42,"user_avatar":125,"doc_module":4,"category_id":104,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":126,"file_id":127,"file_url":128,"file_type":129,"file_size":130,"view_count":55,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":131,"language":132,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":133,"faqs":134,"seo_title":135,"seo_description":12,"update_tm":72,"read_time":136,"preview_page_count":4},2336478487870,"https://ap-avatar.wpscdn.com/davatar_085a072bc5b1113ac321206ff7593b45","arXiv :2 109 .09447v2 [ cs .LG] 1 Jun 2022  \nAlgorithmic Fairness Veri􀀌cation with Graphical Models 􀀃  \nBishwamittra Ghosh  \nSchool of Computing  \nNational University of Singapore Singapore  \nDebabrota Basu  \n􀀓  \nEquipe Scool, Univ. Lille, Inria, UMR 9189 - CRIStAL, CNRS Centrale Lille, France  \nKuldeep S. Meel  \nSchool of Computing  \nNational University of Singapore Singapore  \nAbstract  \nIn recent years, machine learning (ML) algorithms have been deployed in safety-critical and high-stake decision-making, where the fairness of algorithms is of paramount importance.  \nFairness in ML centers on detecting bias towards certain demographic populations induced by an ML classi􀀌er and proposes algorithmic solutions to mitigate the bias with respect to di􀀋erent fairness de􀀌nitions. To this end, several fairness veri􀀌ers have been proposed that compute the bias in the prediction of an ML classi􀀌er|essentially beyond a 􀀌nite dataset|given the probability distribution of input features. In the context of verifying linear classi􀀌ers, existing fairness veri􀀌ers are limited by accuracy due to imprecise modeling of correlations among features and scalability due to restrictive formulations of the classi􀀌ers as SSAT/SMT formulas or by sampling.  \nIn this paper, we propose an e􀀎cient fairness veri􀀌er, called FVGM, that encodes the correlations among features as a Bayesian network. In contrast to existing veri􀀌ers, FVGM proposes a stochastic subset-sum based approach for verifying linear classi􀀌ers. Experimentally, we show that FVGM leads to an accurate and scalable assessment for more diverse families of fairness-enhancing algorithms, fairness attacks, and group/causal fairness metrics than the state-of-the-art fairness veri􀀌ers. We also demonstrate that FVGM facilitates the computation of fairness in􀀍uence functions as a stepping stone to detect the source of bias induced by subsets of features.  \n1. Introduction  \nThe signi􀀌cant improvement of machine learning (ML) over the decades has led to a host of applications of ML in high-stake decision-making such as college admission (Martinez Neda et al., 2021), hiring of employees (Ajunwa et al., 2016), and recidivism prediction (Tollenaarand Van der Heijden, 2013; Dressel and Farid, 2018) . ML algorithms often have an accuracycentric learning objective, which may cause them to be biased towards certain part of the dataset belonging to a certain economically or socially sensitive groups (Landy et al., 1978; Zliobaite, 2015; Berk, 2019) . The following example illustrates a plausible case of unfairness induced by ML algorithms.  \n∗. Source code: [https://github.com/meelgroup/justicia](https://github.com/meelgroup/justicia)  \nExample 1 Following (Ghosh et al., 2021, Example 1.), let us consider an ML problem where the classi􀀌er decides the eligibility of an individual for health insurance given their income and 􀀌tness (Figure 1) . Here, the sensitive feature `age' (A) follows a Bernoulli distribution, and income (I) and 􀀌tness (F ) follow Gaussian distributions. We generate 1000 samples from these distributions and use them to train a Support Vector Machine (SVM) classi􀀌er. The decision boundary of this classi􀀌er is 9:37I + 9:75F 􀀀 0:34A 􀀕 9:4, where A = 1 denotes the sensitive group `age 􀀕 40'. This classi􀀌er selects an individual above and below 40 years of age with probabilities 0:24 and 0:86, respectively. This illustratesa disparate treatment of individuals of two age groups by the SVM classi􀀌er.  \n~~ ~~ age \u003C 40 ~~ ~~ age  40  \n0.0 0.5 income  \n1.0 0.0 0.5 1.0  \nfitness  \nFigure 1: Age-dependent distributions of income and 􀀌tness in Example 1 .  \nIn order to identify and mitigate the bias of ML classi􀀌ers, di􀀋erent fairness de􀀌nitions and fairness algorithms have been proposed (Hardt et al., 2016; Kusner et al., 2017; Mehrabi et al., 2019) . In this paper, we focus on two families of fairness de􀀌nitions: group and causal fairness. Group fairness metrics, such as disparate impact and equalized","cbCaioeyKnn0P6RT","https://ap.wps.com/l/cbCaioeyKnn0P6RT","pdf",1393886,33,"English","# Introduction\n## Fairness verification problem and motivation\n## Fairness definitions: group and causal fairness","[{\"question\":\"What is FVGM and how does it improve fairness verification?\",\"answer\":\"FVGM is an efficient fairness verifier that encodes feature correlations using a Bayesian network. It verifies linear classifiers using a stochastic subset-sum approach, enabling accurate and scalable assessment across fairness metrics and fairness-enhancing algorithms or attacks.\"}]","Algorithmic Fairness Verification with Graphical Models - Efficient Fairness Verifier FVGM | PDF",83]