{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,1,20]],"date-time":"2026-01-20T04:05:21Z","timestamp":1768881921836,"version":"3.49.0"},"reference-count":25,"publisher":"MDPI AG","issue":"1","license":[{"start":{"date-parts":[[2026,1,19]],"date-time":"2026-01-19T00:00:00Z","timestamp":1768780800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["MAKE"],"abstract":"<jats:p>Understanding the decision-making behavior of machine learning models is essential in domains where individual predictions matter, such as medical diagnosis or sports analytics. While explainable artificial intelligence (XAI) methods such as SHAP provide instance-level feature attributions, they mainly summarize typical decision behavior and offer limited support for systematically exploring atypical yet correctly classified cases. In this work, we introduce the Classification Outlier Variability Score (COVAS), a framework designed to support hypothesis generation through the analysis of explanation variability. COVAS operates in the explanation space and builds directly on SHAP value representations. It quantifies how strongly an individual instance\u2019s SHAP-based explanation deviates from class-specific attribution patterns by aggregating standardized SHAP deviations into a single score. Consequently, the applicability of COVAS inherits the model- and data-agnostic properties of SHAP, provided that explanations can be computed for the underlying model and data. We evaluate COVAS on publicly available datasets from the medical and sports domains. The results show that COVAS reveals explanation-space outliers not captured by feature-space outlier detection or prediction uncertainty measures. Robustness analyses demonstrate stability across parameter choices, class imbalance, model initialization, and model classes. Overall, COVAS complements existing XAI techniques by enabling targeted instance-level inspection and facilitating XAI-guided hypothesis formulation.<\/jats:p>","DOI":"10.3390\/make8010024","type":"journal-article","created":{"date-parts":[[2026,1,19]],"date-time":"2026-01-19T14:58:54Z","timestamp":1768834734000},"page":"24","update-policy":"https:\/\/doi.org\/10.3390\/mdpi_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["COVAS: Highlighting the Importance of Outliers in Classification Through Explainable AI"],"prefix":"10.3390","volume":"8","author":[{"ORCID":"https:\/\/orcid.org\/0009-0008-3717-6559","authenticated-orcid":false,"given":"Sebastian","family":"Roth","sequence":"first","affiliation":[{"name":"Faculty of Mathematics, Informatics, Technology, University of Applied Sciences Koblenz, 53424 Remagen, Germany"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-6119-533X","authenticated-orcid":false,"given":"Adrien","family":"Cerrito","sequence":"additional","affiliation":[{"name":"Haute-Ecole Arc Sant\u00e9, HES-SO University of Applied Sciences and Arts Western Switzerland, 2800 Del\u00e9mont, Switzerland"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0009-0003-3828-8214","authenticated-orcid":false,"given":"Samuel","family":"Orth","sequence":"additional","affiliation":[{"name":"Faculty of Mathematics, Informatics, Technology, University of Applied Sciences Koblenz, 53424 Remagen, Germany"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-9625-7075","authenticated-orcid":false,"given":"Ulrich","family":"Hartmann","sequence":"additional","affiliation":[{"name":"Faculty of Mathematics, Informatics, Technology, University of Applied Sciences Koblenz, 53424 Remagen, Germany"},{"name":"Institut f\u00fcr Medizintechnik und Informationsverarbeitung, 56070 Koblenz, Germany"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-5035-6473","authenticated-orcid":false,"given":"Daniel","family":"Friemert","sequence":"additional","affiliation":[{"name":"Faculty of Mathematics, Informatics, Technology, University of Applied Sciences Koblenz, 53424 Remagen, Germany"},{"name":"Institut f\u00fcr Medizintechnik und Informationsverarbeitung, 56070 Koblenz, Germany"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"1968","published-online":{"date-parts":[[2026,1,19]]},"reference":[{"key":"ref_1","doi-asserted-by":"crossref","first-page":"78","DOI":"10.3390\/make5010006","article-title":"XAIR: A Systematic Metareview of Explainable AI (XAI) Aligned to the Software Development Process","volume":"5","author":"Clement","year":"2023","journal-title":"Mach. Learn. Knowl. Extr."},{"key":"ref_2","doi-asserted-by":"crossref","first-page":"169","DOI":"10.3390\/make5010010","article-title":"Explainable Machine Learning","volume":"5","author":"Garcke","year":"2023","journal-title":"Mach. Learn. Knowl. Extr."},{"key":"ref_3","doi-asserted-by":"crossref","first-page":"41111","DOI":"10.1109\/ACCESS.2025.3546681","article-title":"A Literature Review on Applications of Explainable Artificial Intelligence (XAI)","volume":"13","author":"Kalasampath","year":"2025","journal-title":"IEEE Access"},{"key":"ref_4","doi-asserted-by":"crossref","first-page":"128111","DOI":"10.1016\/j.neucom.2024.128111","article-title":"Explainable artificial intelligence: A survey of needs, techniques, applications, and future direction","volume":"599","author":"Mersha","year":"2024","journal-title":"Neurocomputing"},{"key":"ref_5","doi-asserted-by":"crossref","first-page":"109370","DOI":"10.1016\/j.compeleceng.2024.109370","article-title":"A review of Explainable Artificial Intelligence in healthcare","volume":"118","author":"Sadeghi","year":"2024","journal-title":"Comput. Electr. Eng."},{"key":"ref_6","unstructured":"Lundberg, S.M., and Lee, S.I. (2017). A Unified Approach to Interpreting Model Predictions. Adv. Neural Inf. Process. Syst., 30."},{"key":"ref_7","unstructured":"Molnar, C. (2023). Interpreting Machine Learning Models with SHAP: A Guide with Python Examples and Theory on Shapley Values, Christoph Molnar. [1st ed.]."},{"key":"ref_8","doi-asserted-by":"crossref","first-page":"966","DOI":"10.3390\/make3040048","article-title":"Analysis of Explainers of Black Box Deep Neural Networks for Computer Vision: A Survey","volume":"3","author":"Buhrmester","year":"2021","journal-title":"Mach. Learn. Knowl. Extr."},{"key":"ref_9","doi-asserted-by":"crossref","unstructured":"Janoudi, G., Uzun (Rada), M., Fell, D.B., Ray, J.G., Foster, A.M., Giffen, R., Clifford, T., and Walker, M.C. (2024). Outlier analysis for accelerating clinical discovery: An augmented intelligence framework and a systematic review. PLoS Digit. Health, 3.","DOI":"10.1371\/journal.pdig.0000515"},{"key":"ref_10","doi-asserted-by":"crossref","unstructured":"Cai, J., Hu, W., Yang, Y., Yan, H., and Chen, F. (2024). Outlier detection in spatial error models using modified thresholding-based iterative procedure for outlier detection approach. Bmc Med. Res. Methodol., 24.","DOI":"10.1186\/s12874-024-02208-3"},{"key":"ref_11","doi-asserted-by":"crossref","first-page":"625","DOI":"10.1002\/mds.27655","article-title":"Genetic mimics of cerebral palsy","volume":"34","author":"Pearson","year":"2019","journal-title":"Mov. Disord."},{"key":"ref_12","doi-asserted-by":"crossref","first-page":"1451","DOI":"10.1002\/mds.30225","article-title":"Neurophysiology of Atypical Parkinsonian Syndromes: A Study Group Position Paper","volume":"40","author":"Suppa","year":"2025","journal-title":"Mov. Disord."},{"key":"ref_13","doi-asserted-by":"crossref","unstructured":"Datta, A., Sen, S., and Zick, Y. (2017). Algorithmic Transparency via Quantitative Input Influence. Transparent Data Mining for Big and Small Data, Springer.","DOI":"10.1007\/978-3-319-54024-5_4"},{"key":"ref_14","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1016\/S0169-7439(99)00047-7","article-title":"The Mahalanobis distance","volume":"50","author":"Massart","year":"2000","journal-title":"Chemom. Intell. Lab. Syst."},{"key":"ref_15","doi-asserted-by":"crossref","unstructured":"Breunig, M., Kriegel, H., Ng, R., and Sander, J. (2000, January 15\u201318). LOF: Identifying Density-Based Local Outliers. Proceedings of the 2000 ACM SIGMOD International Conference on Management of Data, New York, NY, USA.","DOI":"10.1145\/342009.335388"},{"key":"ref_16","doi-asserted-by":"crossref","first-page":"630","DOI":"10.1016\/j.ejor.2025.05.039","article-title":"Evaluating the Stability of Model Explanations in Instance-Dependent Cost-Sensitive Credit Scoring","volume":"326","author":"Ballegeer","year":"2025","journal-title":"Eur. J. Oper. Res."},{"key":"ref_17","first-page":"346","article-title":"Interpretable Machine Learning for Imbalanced Credit Scoring","volume":"310","author":"Chen","year":"2023","journal-title":"Eur. J. Oper. Res."},{"key":"ref_18","doi-asserted-by":"crossref","unstructured":"Saarela, M. (2024). Recent Applications of Explainable Artificial Intelligence: A Systematic Review. Appl. Sci., 14.","DOI":"10.3390\/app14198884"},{"key":"ref_19","unstructured":"Wolberg, W., Mangasarian, O., and Street, N. (1993). Breast Cancer Wisconsin (Diagnostic), UCI Machine Learning Repository."},{"key":"ref_20","unstructured":"Mathan, J., and Community, K. (2025, June 20). FIFA 2018 Match Statistics: Man of the Match Dataset. Available online: https:\/\/www.kaggle.com\/datasets\/mathan\/fifa-2018-match-statistics."},{"key":"ref_21","doi-asserted-by":"crossref","unstructured":"Friemert, D., Schnur, D., Runkel, S., Borsch, J., Karamanidis, K., Dellen, B., Thieme, L., Fiedler, A., Jaeckel, U., and Hartmann, U. (2025). Limitations of Public Biomechanical Movement Datasets for Deep Learning: Issues of Metadata, Standardization, and Variety in Motion Types. medRxiv.","DOI":"10.1101\/2025.05.29.25328474"},{"key":"ref_22","doi-asserted-by":"crossref","first-page":"167","DOI":"10.1007\/s40279-012-0011-z","article-title":"Key Properties of Expert Movement Systems in Sport","volume":"43","author":"Seifert","year":"2013","journal-title":"Sport. Med."},{"key":"ref_23","doi-asserted-by":"crossref","first-page":"117","DOI":"10.1080\/17408989.2018.1552673","article-title":"Principles of nonlinear pedagogy in sport practice","volume":"24","author":"Correia","year":"2018","journal-title":"Phys. Educ. Sport Pedagog."},{"key":"ref_24","doi-asserted-by":"crossref","unstructured":"Bento, J., Saleiro, P., Cruz, A.F., Figueiredo, M.A.T., and Bizarro, P. (2021, January 14\u201318). TimeSHAP: Explaining Recurrent Models through Sequence Perturbations. Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining, Virtual.","DOI":"10.1145\/3447548.3467166"},{"key":"ref_25","first-page":"104435","article-title":"WindowSHAP: An Efficient Framework for Explaining Time Series Classifiers","volume":"145","author":"Nayebi","year":"2023","journal-title":"J. Biomed. Inform."}],"container-title":["Machine Learning and Knowledge Extraction"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.mdpi.com\/2504-4990\/8\/1\/24\/pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,1,19]],"date-time":"2026-01-19T15:19:44Z","timestamp":1768835984000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.mdpi.com\/2504-4990\/8\/1\/24"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,1,19]]},"references-count":25,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2026,1]]}},"alternative-id":["make8010024"],"URL":"https:\/\/doi.org\/10.3390\/make8010024","relation":{},"ISSN":["2504-4990"],"issn-type":[{"value":"2504-4990","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,1,19]]}}}