[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-82445-en":3,"doc-seo-82445-105":29,"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":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":13,"seo_description":14,"update_tm":27,"read_time":28},82445,7971461741311,"Ophelia","https://ap-avatar.wpscdn.com/avatar/74000253aff267980c6?x-image-process=image/resize,m_fixed,w_180,h_180&k=1779345379180704826",8,"Research & Report","ConceptSMILE Auditing the Trustworthiness of Concept-Based Explainable AI","ConceptSMILE audits the trustworthiness of concept-based explainable AI, where concept-level outputs improve human interpretability but do not inherently guarantee reliability. The model-agnostic perturbation framework extends SMILE-style perturbation logic to human-understandable concepts by perturbing input regions, measuring concept-response shifts, applying locality weighting, and fitting an XGBoost surrogate to approximate local concept behavior. Reliability is evaluated via attribution accuracy, surrogate fidelity, faithfulness, stability, and consistency. Experiments on retinal fundus images compare MedSAM visual concepts with VLM semantic concepts.","CONCEPTSMILE: AUDITING THE TRUSTWORTHINESS OF  \nCONCEPT-BASED EXPLAINABLE AI  \narXiv :2607 .09649v 1 [ cs .AI] 10 Jul 2026  \nMohadeseh Mollapour  \nSchool of Computer Science University of Hull  \nHull, United Kingdom [m.mollapour-papkyadeh-2023@hull.ac.uk](m.mollapour-papkyadeh-2023@hull.ac.uk)  \nZeinab Dehghani  \nPhD Researcher (Engineering), WMG University of Warwick  \nCoventry, United Kingdom [sara.dehghani@warwick.ac.uk](sara.dehghani@warwick.ac.uk)  \nKoorosh Aslansefat  \nSchool of Computer Science University of Hull Hull, United Kingdom [k.aslansefat@hull.ac.uk](k.aslansefat@hull.ac.uk)  \nBhupesh Kumar Mishra  \nSchool of Computer Science University of Hull Hull, United Kingdom [bhupesh.mishra@hull.ac.uk](bhupesh.mishra@hull.ac.uk)  \nTejal Shah  \nSchool of Computing Newcastle University Newcastle upon Tyne, United Kingdom [tejal.shah@ncl.ac.uk](tejal.shah@ncl.ac.uk)  \nZhibao Mian  \nSchool of Computer Science University of Hull Hull, United Kingdom [z.mian2@hull.ac.uk](z.mian2@hull.ac.uk)  \nJuly 13, 2026  \nABSTRACT  \nConcept-based explainable artificial intelligence (AI) can make model reasoning more humanunderstandable, but concept-level outputs are not automatically trustworthy. We introduce ConceptSMILE, a model-agnostic perturbation-based auditing framework for evaluating the reliability of concept-based explanations. Rather than replacing SMILE, ConceptSMILE extends its perturbationbased logic from feature or region level attribution to the auditing of human-understandable concept explanations. The framework perturbs input regions, measures concept-response shifts, applies locality weighting, and fits an XGBoost surrogate to approximate local concept behaviour. Reliability is assessed through attribution accuracy, surrogate fidelity, faithfulness, stability, and consistency. We evaluate ConceptSMILE on retinal fundus images by comparing MedSAM-derived visual concepts with VLM-based semantic concepts. Results show that reliability varies across concepts and pathways: MedSAM achieves stronger spatial attribution and the highest surrogate fidelity (R2 = 0 .8503, R2w = 0 .8465), while the VLM pathway shows stronger vessel faithfulness and stronger stability under selected artefact conditions. ConceptSMILE provides an independent audit layer for evaluating the trustworthiness of concept-based XAI.  \n1 Introduction  \nArtificial intelligence systems are increasingly used in high-stakes domains where model decisions must be not only accurate but also transparent, reliable, and accountable. Deep learning, vision transformers, multimodal models, and foundation models have achieved strong performance in areas such as medical imaging, autonomous systems, environmental monitoring, and decision support. However, these models often remain difficult to interpret because their internal reasoning is usually represented through high-dimensional latent features rather than human-understandable  \nA PREPRINT-JULY 13, 2026  \nevidence [1] . This lack of transparency becomes particularly problematic when AI systems are used in settings where incorrect or poorly justified decisions may affect human safety, clinical judgement, or public trust [2] .  \nExplainable Artificial Intelligence (XAI) has emerged to address this problem by making model behaviour more understandable to users and domain experts. Widely used post-hoc methods such as LIME and SHAP provide important tools for identifying influential input features [3, 4] . In computer vision, Grad-CAM has also been widely used to highlight image regions that contribute to a model prediction [5] . These methods have been applied across many domains, including medical image analysis, where heatmaps and saliency maps can show which parts of an image contribute to a prediction. Nevertheless, region-based attribution is often insufficient for high-stakes decision-making. A heatmap may show where a model attends, but it does not necessarily explain what clinically, semantically, or operationally mean","cbCaioSQwUjqtvyc","https://ap.wps.com/l/cbCaioSQwUjqtvyc","pdf",6120868,1,31,"English","en",105,"# Abstract\n# Introduction\n## Motivation for trustworthy explanations\n## Existing XAI methods and limitations\n## Concept-based explainable AI approaches\n## Why concept explanations may be unreliable\n# Proposed framework: ConceptSMILE","[{\"question\":\"What problem does ConceptSMILE address in concept-based XAI?\",\"answer\":\"ConceptSMILE addresses the gap between interpretability and trustworthiness: concept-level explanations can look plausible while being unreliable due to spurious correlations, leakage, instability, or incomplete representations.\"},{\"question\":\"How does ConceptSMILE evaluate the reliability of concept explanations?\",\"answer\":\"It perturbs input regions, measures concept-response shifts, uses locality weighting, and trains an XGBoost surrogate to approximate local concept behavior, then scores attribution accuracy, surrogate fidelity, faithfulness, stability, and consistency.\"},{\"question\":\"What do the retinal fundus image results suggest about different concept pathways?\",\"answer\":\"Reliability varies across concepts and pathways: MedSAM shows stronger spatial attribution and the highest surrogate fidelity, while the VLM pathway shows stronger vessel faithfulness and improved stability under selected artifact conditions.\"}]",1784180414,78,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"conceptsmile-auditing-the-trustworthiness-of-concept-based-explainable-ai","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/conceptsmile-auditing-the-trustworthiness-of-concept-based-explainable-ai/82445/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does ConceptSMILE address in concept-based XAI?","Question",{"text":75,"@type":76},"ConceptSMILE addresses the gap between interpretability and trustworthiness: concept-level explanations can look plausible while being unreliable due to spurious correlations, leakage, instability, or incomplete representations.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does ConceptSMILE evaluate the reliability of concept explanations?",{"text":80,"@type":76},"It perturbs input regions, measures concept-response shifts, uses locality weighting, and trains an XGBoost surrogate to approximate local concept behavior, then scores attribution accuracy, surrogate fidelity, faithfulness, stability, and consistency.",{"name":82,"@type":73,"acceptedAnswer":83},"What do the retinal fundus image results suggest about different concept pathways?",{"text":84,"@type":76},"Reliability varies across concepts and pathways: MedSAM shows stronger spatial attribution and the highest surrogate fidelity, while the VLM pathway shows stronger vessel faithfulness and improved stability under selected artifact conditions.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"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":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":45,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":45,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]