[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-128248-en":3,"doc-seo-128248-105":31,"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":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":28,"seo_description":14,"update_tm":29,"read_time":30},128248,2336475104736,"Quinn","https://ap-avatar.wpscdn.com/avatar/22000c4c5e0e5b17e70?x-image-process=image/resize,m_fixed,w_180,h_180&k=1786591360781797222",8,"Research & Report","Interpretability and Privacy for Trustworthy Machine Learning - Bridging Theoretical and User-Centric Perspectives - Dissertation","Data-driven machine learning is increasingly used in high-stakes scenarios, raising the need for trustworthy and socially responsible machine learning (TSRML). Regulatory requirements such as GDPR and the AI Act emphasize explainability and privacy, including human oversight and protection of end users. This thesis addresses a disconnect between theoretical work and user-centered practice by reconciling interpretability and privacy viewpoints, proposing methods with theoretical and user-centric alignment and analyzing practical side effects.","Interpretability and Privacy for Trustworthy Machine Learning: Bridging Theoretical and User-Centric Perspectives  \nDissertation  \nder Mathematisch-Naturwissenschaftlichen Fakultät der Eberhard Karls Universität Tübingen zur Erlangung des Grades eines  \nDoktors der Naturwissenschaften  \n(Dr. rer. nat. )  \nvorgelegt von  \nTobias Leemann  \naus Ingolstadt  \nTübingen  \nGedruckt mit Genehmigung der Mathematisch-Naturwissenschaftlichen Fakultät der Eberhard Karls Universität Tübingen.  \nTag der mündlichen Qualifikation: 27.03.2025  \nDekan: Prof. Dr. Thilo Stehle  \n1. Berichterstatter/-in: Prof. Dr. Gjergji Kasneci  \n2. Berichterstatter/-in: Prof. Dr. Seong Joon Oh  \nAbstract  \nWith data-driven applications increasingly deployed in high-stakes scenarios, the importance of trustworthy and socially responsible machine learning (TSRML) has been recognized and gained substantial research attention. Moreover, TSRML requirements are successively mandated by recent regulations such as the European Union’s General Data Protection Regulation and the Artificial Intelligence (AI) Act. Most prominently, these regulations require careful consideration of explainability and privacy aspects of AI systems. They also make clear that explanations should allow human oversight and protect end users, taking a human-centric perspective. For accountable deployment in user-facing applications, TSRML systems therefore need theoretical guarantees regarding interpretability and privacy, but should be designed around human users. Despite this necessity, we find that the fields of interpretable and private machine learning suffer from a disconnect between foundational theoretical work and user-centered research in practice. To tackle this gap, we explore interpretable and private machine learning from both a theoretical and user-centric viewpoint with the goal of reconciling common misalignments between the two. In the first part of this thesis, we study interpretability through conceptual explanations, counterfactual explanations, and feature attribution methods. We outline how user-friendly conceptual explanations suffer from theoretical identifiability issues and Counterfactual explanations often neglect the requirement for human oversight and the user perspective as they incentivize adversarial actions over real improvements. Finally, popular feature attributions can be theoretically misaligned with underlying predictive models such as transformers. We propose rigorous techniques to reconcile both perspectives for these three explainability techniques. In the second part, we study computational strategies to protect privacy in machine learning and their side effects on users. We show how handing back control over specific features to the users results in a discrimination risk and  \nhow implementing privacy notions such as differential privacy (DP) may result in users being subjected to excessively noisy decisions. We propose strategies to mitigate the two issues, showing that privacy guarantees are often possible without significant impact on predictive performance. Our findings suggest that it is necessary to consider theoretical and user-centric perspectives in conjunction and that tractable tools to reconcile both perspectives often exist. To conclude, we discuss how interpretability and privacy can be combined, explore connections between the two, and outline remaining steps towards practical implementations of TSRML.  \nKurzfassung  \nDurch die zunehmende Nutzung datengesttzter Modelle zum Treffen risikoreicher Entscheidungen erfhrt das Gebiet des vertrauenswrdigen und sozial verantwortlichen maschinellen Lernens (TSRML) erhhte Aufmerksamkeit in der Wissenschaft. Darber hinaus werden die Anforderungen an TSRML nachund nach in Vorschriften der Europischen Union wie der Datenschutzgrundverordnung (GDPR) und dem gerade in Kraft getretenen AI Act festgeschrieben. DieseVerordnungen verlangen eine sorgfltige Prfung von aufknstlicher Intelligenz (KI) basierenden","cbCaiia1K1uXBcWC","https://ap.wps.com/l/cbCaiia1K1uXBcWC","pdf",13750542,3,1,226,"English","en",105,"# Abstract\n## Interpretability: conceptual, counterfactual, and feature attribution perspectives\n## Privacy: computational protections and user side effects\n## Reconciling theory and user-centric requirements for TSRML","[{\"question\":\"Why does TSRML require both explainability and privacy in high-stakes settings?\",\"answer\":\"Because regulations like GDPR and the AI Act mandate careful consideration of explainability and privacy, while also requiring human oversight and protection of end users in user-facing deployments.\"},{\"question\":\"What gaps does the thesis identify between theoretical and user-centered interpretability research?\",\"answer\":\"It finds misalignments where conceptual explanations face theoretical identifiability issues, counterfactual explanations can neglect human oversight and user perspective, and popular feature attributions can be theoretically misaligned with underlying predictive models such as transformers.\"},{\"question\":\"How does the thesis address privacy protections and their effects on users?\",\"answer\":\"It studies computational privacy strategies and shows that returning control of specific features can create discrimination risk, while differential privacy may lead to excessively noisy decisions; mitigation strategies are proposed with limited impact on predictive performance.\"}]","Interpretability and Privacy for Trustworthy Machine Learning - Bridging Theoretical and User-Centric Perspectives - Dissertation | PDF",1785946183,570,{"code":4,"msg":32,"data":33},"ok",{"site_id":25,"language":24,"slug":34,"title":13,"keywords":35,"description":14,"schema_data":36,"social_meta":87,"head_meta":89,"extra_data":91,"updated_unix":29},"interpretability-and-privacy-for-trustworthy-machine-learning-bridging-theoretical-and-user-centric-perspectives-dissertation","",{"@graph":37,"@context":86},[38,54,69],{"@type":39,"itemListElement":40},"BreadcrumbList",[41,45,49,51],{"item":42,"name":43,"@type":44,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":46,"name":47,"@type":44,"position":48},"https://docshare.wps.com/document/","Document",2,{"item":50,"name":12,"@type":44,"position":20},"https://docshare.wps.com/document/research-report/",{"item":52,"name":13,"@type":44,"position":53},"https://docshare.wps.com/document/interpretability-and-privacy-for-trustworthy-machine-learning-bridging-theoretical-and-user-centric-perspectives-dissertation/128248/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":24,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":42,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-25","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},"Why does TSRML require both explainability and privacy in high-stakes settings?","Question",{"text":76,"@type":77},"Because regulations like GDPR and the AI Act mandate careful consideration of explainability and privacy, while also requiring human oversight and protection of end users in user-facing deployments.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"What gaps does the thesis identify between theoretical and user-centered interpretability research?",{"text":81,"@type":77},"It finds misalignments where conceptual explanations face theoretical identifiability issues, counterfactual explanations can neglect human oversight and user perspective, and popular feature attributions can be theoretically misaligned with underlying predictive models such as transformers.",{"name":83,"@type":74,"acceptedAnswer":84},"How does the thesis address privacy protections and their effects on users?",{"text":85,"@type":77},"It studies computational privacy strategies and shows that returning control of specific features can create discrimination risk, while differential privacy may lead to excessively noisy decisions; mitigation strategies are proposed with limited impact on predictive performance.","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":25},{"code":4,"msg":5,"data":93},[94,98,102,106,111,116,121,124,129,132,136],{"id":21,"doc_module":4,"doc_module_name":47,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":48,"doc_module":4,"doc_module_name":47,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":47,"category_name":103,"show_sort_weight":104,"slug":105},"Exam",70,"exam",{"id":107,"doc_module":4,"doc_module_name":47,"category_name":108,"show_sort_weight":109,"slug":110},5,"Comic",60,"comic",{"id":112,"doc_module":4,"doc_module_name":47,"category_name":113,"show_sort_weight":114,"slug":115},6,"Technology",50,"technology",{"id":117,"doc_module":4,"doc_module_name":47,"category_name":118,"show_sort_weight":119,"slug":120},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":47,"category_name":12,"show_sort_weight":122,"slug":123},30,"research-report",{"id":125,"doc_module":4,"doc_module_name":47,"category_name":126,"show_sort_weight":127,"slug":128},9,"Religion & Spirituality",20,"religion-spirituality",{"id":127,"doc_module":4,"doc_module_name":47,"category_name":130,"show_sort_weight":127,"slug":131},"World Cup","world-cup",{"id":133,"doc_module":4,"doc_module_name":47,"category_name":134,"show_sort_weight":133,"slug":135},10,"Lifestyle","lifestyle",{"id":137,"doc_module":4,"doc_module_name":47,"category_name":138,"show_sort_weight":107,"slug":139},19,"General","general"]