[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-127664-105":59,"doc-detail-127664-en":135},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":128,"head_meta":130,"extra_data":132,"updated_unix":134},105,"en","from-mistakes-to-insights-counterfactual-explanations-for-incorrect-machine-learning-predictions-completed-research-paper","From Mistakes to Insights - Counterfactual Explanations for Incorrect Machine Learning Predictions - Completed Research Paper","","Machine learning models often act as black boxes, creating interpretability gaps in high-stakes settings such as finance, healthcare, and criminal justice. Interpretable machine learning and explainable AI seek explanations that support trust, transparency, and accountability. Counterfactual explanations offer actionable alternatives by showing how input or feature changes could lead to different outcomes, yet existing evaluation emphasizes fidelity without addressing errors. This study rectifies and explains incorrect predictions in binary classification (beneficial vs adverse) by generating counterfactuals via an explanation tree derived from black-box outputs.",{"@graph":69,"@context":127},[70,84,106],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/from-mistakes-to-insights-counterfactual-explanations-for-incorrect-machine-learning-predictions-completed-research-paper/127664/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/from-mistakes-to-insights-counterfactual-explanations-for-incorrect-machine-learning-predictions-completed-research-paper/127664.png","ImageObject",300,407,{"name":92,"@type":93},"Olivia Brown","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-17","2026-08-05",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",13,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119,123],{"name":110,"@type":111,"acceptedAnswer":112},"What problem does the study address in explainable AI methods?","Question",{"text":113,"@type":114},"The study addresses how existing XAI/IML approaches rely on fidelity and can mistakenly treat misclassified outputs as correct, leading to explanations that do not correspond to the model’s error.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"How does the proposed approach generate explanations for incorrect predictions?",{"text":118,"@type":114},"It uses a black-box classifier and then fits a decision tree—an explanation tree—based on the black-box classification results to identify suitable counterfactual examples for explanations.",{"name":120,"@type":111,"acceptedAnswer":121},"What types of errors and decision settings are considered?",{"text":122,"@type":114},"The focus is on binary classification between beneficial and adverse outcomes, covering two misclassification types: beneficial-to-adverse (b2a) and adverse-to-beneficial (a2b).",{"name":124,"@type":111,"acceptedAnswer":125},"Who receives the explanations and how are research questions framed?",{"text":126,"@type":114},"Explanations are tailored to individuals for cases where a beneficial class is predicted as adverse, and to organizations for cases where an adverse class is predicted as beneficial.","https://schema.org",{"og:url":83,"og:type":129,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":131,"canonical":83},"index,follow",{"doc_id":133,"site_id":62},127664,1785940613,{"code":4,"msg":5,"data":136},{"doc_id":133,"user_id":137,"nickname":92,"user_avatar":138,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":139,"file_id":140,"file_url":141,"file_type":142,"file_size":143,"view_count":105,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":14,"language":144,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":145,"faqs":146,"seo_title":147,"seo_description":67,"update_tm":134,"read_time":24},962084925636,"https://ap-avatar.wpscdn.com/davatar_994ba38a5ba835b3df7d355c54d3ed8d","Association for Information Systems  \nAIS Electronic Library (AISeL)  \n\n| NEAIS 2023 Proceedings | New England Chapter of Association for Information Systems |\n| --- | --- |\n| 10-28-2023\u003Cbr>From Mistakes to Insights: Counterfactual Explanations for Incorrect Machine Learning Predictions\u003Cbr>Amir Asrzad\u003Cbr>University of Massachusetts Lowell, [Amir_Asrzad@uml.edu](Amir_Asrzad@uml.edu)\u003Cbr>Xiao-Bai Li\u003Cbr>University of Massachusetts Lowell, [xiaobai_li@uml.edu](xiaobai_li@uml.edu)\u003Cbr>Follow this and additional works at: [https://aisel.aisnet.org/neais2023](https://aisel.aisnet.org/neais2023) |  |\n\nRecommended Citation  \nAsrzad, Amir and Li, Xiao-Bai, \"From Mistakes to Insights: Counterfactual Explanations for Incorrect Machine Learning Predictions\" (2023) . NEAIS 2023 Proceedings. 10.  \n[https://aisel.aisnet.org/neais2023/10](https://aisel.aisnet.org/neais2023/10)  \nThis material is brought to you by the New England Chapter of Association for Information Systems at AIS Electronic Library (AISeL) . It has been accepted for inclusion in NEAIS 2023 Proceedings by an authorized administrator of AIS Electronic Library (AISeL) . For more information, please [contact](contact elibrary@aisnet.org)[ elibrary@aisnet.org](contact elibrary@aisnet.org).  \nFrom Mistakes to Insights: Counterfactual Explanations for Incorrect Machine Learning Predictions  \nCompleted Research Paper  \nAmir Asrzad University of Massachusetts Lowell  \n[Amir_Asrzad@uml.edu](Amir_Asrzad@uml.edu)  \nXiao-Bai Li University of Massachusetts Lowell  \n[Xiaobai_Li@uml.edu](Xiaobai_Li@uml.edu)  \nABSTRACT  \nMachine learning (ML) has revolutionized various industries with powerful predictive capabilities. However, the lack of interpretability in these black box models poses challenges in high-stakes domains like finance, healthcare, and criminal justice. Interpretable machine learning (IML) or explainable AI (XAI) aims to address these challenges by developing methods that provide meaningful explanations for human understanding (Molnar, 2023) . By enhancing interpretability, we can establish trust, transparency, and accountability in AI systems, ensuring fairness and reliability in their outputs. Counterfactual explanations have gained popularity as an XAI/IML method (Verma, Boonsanong, Hoang, Hines, Dickerson and Shah, 2022) . Unlike traditional methods, counterfactual explanations don't directly explain the \"why\" behind a decision. Instead, they present alternative scenarios, or counterfactuals, illustrating how changes in inputs or features could lead to different outcomes. For instance, if an ML model predicts a loan default, a counterfactual explanation can advise the applicant on the factors that could secure loan approval. Counterfactual explanations are easy to understand, persuasive, and provide actionable insights (Fernández-Loría, Provost, and Han, 2022), leading to increased research attention (Guidotti, 2022) .  \nFidelity is a crucial criterion for evaluating XAI/IML methods, measuring their ability to approximate black box model predictions accurately. However, existing approaches solely prioritize fidelity and overlook errors. When a black box model misclassifies an instance, interpretable methods, based on fidelity, mistakenly treat the misclassified result as correct and attempt to explain the incorrect outcome. This misinterpretation has significant implications for subsequent actions, and no existing studies have addressed this issue.  \nIn this study, we address the problem of rectifying and explaining incorrect predictions made by AI and ML models. Our focus is on classification problems with two categorical outcomes: beneficial and adverse. Two types of errors exist: misclassifying beneficial as adverse (b2a) and misclassifying adverse as beneficial (a2b) . We distinguish between two types of errors, and our research questions are (1) how to explain misclassifications for individuals when a beneficial class is classified as adverse and (2) for organizations wh","cbCainoDovcdt3PL","https://ap.wps.com/l/cbCainoDovcdt3PL","pdf",65900,"English","# Abstract\n## Interpretable ML and counterfactual explanations\n## Fidelity and the overlooked role of errors\n## Problem setting and research questions\n## Proposed method: explanation tree and counterfactual selection\n## Contributions and empirical validation","[{\"question\":\"What problem does the study address in explainable AI methods?\",\"answer\":\"The study addresses how existing XAI/IML approaches rely on fidelity and can mistakenly treat misclassified outputs as correct, leading to explanations that do not correspond to the model’s error.\"},{\"question\":\"How does the proposed approach generate explanations for incorrect predictions?\",\"answer\":\"It uses a black-box classifier and then fits a decision tree—an explanation tree—based on the black-box classification results to identify suitable counterfactual examples for explanations.\"},{\"question\":\"What types of errors and decision settings are considered?\",\"answer\":\"The focus is on binary classification between beneficial and adverse outcomes, covering two misclassification types: beneficial-to-adverse (b2a) and adverse-to-beneficial (a2b).\"},{\"question\":\"Who receives the explanations and how are research questions framed?\",\"answer\":\"Explanations are tailored to individuals for cases where a beneficial class is predicted as adverse, and to organizations for cases where an adverse class is predicted as beneficial.\"}]","From Mistakes to Insights - Counterfactual Explanations for Incorrect Machine Learning Predictions - Completed Research Paper | PDF"]