[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85987-en":3,"doc-seo-85987-105":30,"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":13,"seo_description":14,"update_tm":28,"read_time":29},85987,1374391975076,"Riley","https://ap-avatar.wpscdn.com/avatar/14000253ca4ec9f6853?x-image-process=image/resize,m_fixed,w_180,h_180&k=1783305029341752051",8,"Research & Report","Auditing Construct Overlap in Explainable Machine Learning: Evidence from Burnout-Depression Prediction Across Student Cohorts","Explainable machine learning (XAI) pipelines applied to composite mental health outcomes can produce seemingly robust, cross-population-stable risk hierarchies that reflect artifacts of how outcomes are constructed. Using an ElasticNet pipeline on 886 medical students, validated across 2,580 longitudinal observations and 701 non-medical students with identical instruments, the hierarchy is dominated by trait anxiety and health satisfaction (Kendall τ = 1.0) with transfer R2 = 0.41–0.49. Residualization experiments show the apparent signal collapses when shared psychometric overlap is removed, with model R2 dropping to 0.016 and deployment ruled out by wide conformal prediction intervals.","arXiv :2607 . 10633v 1 [ cs .LG] 12 Jul 2026  \nAuditing Construct Overlap in Explainable Machine Learning: Evidence from Burnout-Depression Prediction  \nAcross Student Cohorts  \nAlireza Dehghan 1 Negin Ashrafi2,*  \n1 Sharif University of Technology, Tehran, Iran  \n2 University of Southern California, California, USA  \n*  \n[Corresponding author: ashrafin@usc.edu](Corresponding author: ashrafin@usc.edu)  \nAbstract. Explainable machine learning (XML) pipelines applied to composite mental health outcomes can produce apparently-robust, cross-population-stable risk hierarchies that are largely artefacts of how the outcome was constructed. We demonstrate this using an ElasticNet pipeline applied to 886 medical students at the University of Lausanne (primary cohort, 2022), validated across 2,580 longitudinal observations at three time points and 701 non-medical students from eight faculties; all three datasets share identical instruments.  \nThe pipeline produces a hierarchy in which trait anxiety and health satisfaction dominate wherever the outcome is measured, with Kendall τ = 1 .0 for the top-two positions across all five evaluation sets and consistent transfer performance (R2 : 0 .41–0.49) . Two residualization experiments, which isolate shared variance between correlated variables via regression, reveal the mechanism: when trait anxiety (STAI-T) is residualized against the co-included depression subscale (CES-D, r = 0 .72), model R2 drops from 0 .41 to 0 . 16 and STAI-T falls from rank 1 to rank 6; when burnout subscales are residualized against CES-D, R2 collapses to 0 .016. Prediction intervals average 35 .4 units on a 0–100 scale (2.4 outcome standard deviations), independently ruling out individual-level deployment. The residualization protocol is the paper’s transferable contribution: any XAI study combining correlated predictor and outcome constructs should apply this check before interpreting apparent stability as a finding.  \nKeywords: construct overlap; explainability artefact; burnout; depression; residualization; XAI; SHAP; conformal  \nprediction; medical students; multi-cohort validation  \n1 Introduction  \nBurnout and depression among medical students are prevalent and consequential: reported burnout rates consistently exceed those of the general population, and untreated distress impairs clinical empathy and patient care quality [1–3] . The application of machine learning to self-report survey data offers a route to automated risk stratification; paired with SHAP explanations, such pipelines appear to identify stable, crosspopulation risk hierarchies, a consistent rank-ordering of predictors by importance, in which trait anxiety and health satisfaction dominate [4, 5] .  \nThe objective of this study is not to identify definitive risk factors, but to test whether the stability produced by such a pipeline is real. Specifically: trait anxiety (STAI-T) and the CES-D depression scale are highly correlated (r = 0 .72 in this sample) . The CES-D is embedded inside the composite burnoutdepression outcome. When a predictor and an outcome component share a latent construct, the predictor’s explanatory dominance may reflect psychometric overlap rather than a genuine risk relationship. We call this construct overlap, and we introduce a residualization protocol to quantify how much of the apparent hierarchy survives once the overlap is removed.  \nThe principal contributions are threefold. First, we validate a residualization protocol for detecting construct overlap artefacts: removing the anxiety– depression psychometric channel collapses R2 from 0.41 to 0.016 (burnout-specific outcome) and STAI-T  \nfrom rank 1 to rank 6 . Second, we demonstrate that apparent cross-cohort stability is not evidence of a portable risk structure when the signal is anchored toa fixed psychometric correlation: the model transfers with R2 = 0 .41–0.49 across five evaluation sets because the STAI-T/CES-D bivariate relationship is constant, not because t","cbCaieq88HIkLQ1m","https://ap.wps.com/l/cbCaieq88HIkLQ1m","pdf",590042,5,1,9,"English","en",105,"# Introduction\n# Related Work\n# Methodology\n# Results\n# Discussion\n# Conclusion","[{\"question\":\"What problem does the paper address in explainable machine learning for mental health prediction?\",\"answer\":\"It examines how XAI pipelines can yield stable, cross-cohort risk hierarchies that are artifacts caused by construct overlap in composite outcomes rather than genuine risk relationships.\"},{\"question\":\"What is the key mechanism revealed by the residualization experiments?\",\"answer\":\"When trait anxiety is residualized against the co-included depression subscale, model performance collapses (R2 from 0.41 to 0.16) and rankings change; residualizing burnout subscales against CES-D further collapses R2 to 0.016.\"},{\"question\":\"Why does the paper argue that current models are not suitable for individual-level deployment?\",\"answer\":\"Conformal prediction intervals are wide on a 0–100 scale, averaging about 35.4 units (≈2.4 outcome standard deviations), making individual-level clinical prediction not actionable with the current feature sets.\"}]",1784207595,23,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":87,"head_meta":89,"extra_data":91,"updated_unix":28},"auditing-construct-overlap-in-explainable-machine-learning-evidence-from-burnout-depression-prediction-across-student-cohorts","",{"@graph":36,"@context":86},[37,54,69],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/auditing-construct-overlap-in-explainable-machine-learning-evidence-from-burnout-depression-prediction-across-student-cohorts/85987/",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":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-07-24","2026-07-16",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},"What problem does the paper address in explainable machine learning for mental health prediction?","Question",{"text":76,"@type":77},"It examines how XAI pipelines can yield stable, cross-cohort risk hierarchies that are artifacts caused by construct overlap in composite outcomes rather than genuine risk relationships.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"What is the key mechanism revealed by the residualization experiments?",{"text":81,"@type":77},"When trait anxiety is residualized against the co-included depression subscale, model performance collapses (R2 from 0.41 to 0.16) and rankings change; residualizing burnout subscales against CES-D further collapses R2 to 0.016.",{"name":83,"@type":74,"acceptedAnswer":84},"Why does the paper argue that current models are not suitable for individual-level deployment?",{"text":85,"@type":77},"Conformal prediction intervals are wide on a 0–100 scale, averaging about 35.4 units (≈2.4 outcome standard deviations), making individual-level clinical prediction not actionable with the current feature sets.","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},[94,98,102,106,110,115,120,123,127,130,134],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":103,"show_sort_weight":104,"slug":105},"Exam",70,"exam",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":22,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":20,"slug":137},19,"General","general"]