[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-study-questions-whether-smart-joint-picks-improve-rehab-ai":10,"sections":34},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":24,"tags":25,"sources":29,"feedback":33,"feedback_at":22,"cost_usd":33,"total_tokens":33},9510,"study-questions-whether-smart-joint-picks-improve-rehab-ai","Study Questions Whether Smart Joint Picks Improve Rehab AI","An audit of rehab-exercise AI finds that picking specific joints barely helps accuracy once evaluation methods are controlled properly.","A new audit finds that a popular trick for teaching AI to judge exercise form barely survives scrutiny of its own evaluation choices.\n\nResearchers re-examined claims that hand-picking specific joints improves skeleton-based classifiers that judge whether a rehab exercise is performed correctly. Using 1,057 repetitions from ten subjects in the REHAB24-6 dataset, they found the reported advantage shrank from 0.055 to 0.020 AUROC depending on how results were averaged, and neither figure clears zero with confidence. When they generated 1,000 random joint subsets matched only on size, 14 matched or beat the original hand-picked selection; once those random subsets also matched structural traits like using both body sides and including the trunk, 145 did. Only one classifier, an RBF-SVM, kept a positive edge under the stricter per-subject comparison; logistic regression and a random-convolution baseline did not.\n\nThe study is a case study in how one evaluation choice, averaging results across all data versus averaging per person, can manufacture or erase an apparent breakthrough. That matters because rehab-tracking AI is increasingly pitched to clinics and apps as a judge of correct movement, and an overstated joint-selection claim could get built into a product that is really just reading noise as signal.\n\nCall it a smaller-scale echo of the reproducibility reckoning that has already hit wearable fitness and medical-imaging AI: impressive numbers, a modest dataset, and a result that mostly evaporates once someone checks the math.","[\"ai\",\"machine learning\",\"rehabilitation\",\"research methods\"]","2026-10-02T04:00:00.000Z","2026-10-02T22:31:57.967Z","2026-10-02T22:32:02.885Z","published",null,[],"ai",[24,26,27,28],"machine learning","rehabilitation","research methods",[30],{"name":31,"url":32},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.01188",0,{"sections":35},[36,39,43,47,52,56,60,65,70,75,80,85,90,95],{"name":37,"slug":24,"count":38,"latest_published_at":18},"AI",5886,{"name":40,"slug":41,"count":42,"latest_published_at":18},"Security","security",835,{"name":44,"slug":45,"count":46,"latest_published_at":18},"Policy","policy",438,{"name":48,"slug":49,"count":50,"latest_published_at":51},"Deals","deals",317,"2026-10-01T22:00:00.000Z",{"name":53,"slug":54,"count":55,"latest_published_at":18},"Hardware","hardware",199,{"name":57,"slug":58,"count":59,"latest_published_at":18},"Science","science",171,{"name":61,"slug":62,"count":63,"latest_published_at":64},"Consumer Tech","consumer-tech",155,"2026-10-01T19:54:10.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":69},"Dev Tools","dev-tools",96,"2026-10-01T16:57:03.000Z",{"name":71,"slug":72,"count":73,"latest_published_at":74},"Software","software",93,"2026-09-30T21:41:11.000Z",{"name":76,"slug":77,"count":78,"latest_published_at":79},"Startups","startups",90,"2026-10-01T21:55:22.000Z",{"name":81,"slug":82,"count":83,"latest_published_at":84},"Gaming","gaming",53,"2026-10-02T02:50:39.000Z",{"name":86,"slug":87,"count":88,"latest_published_at":89},"General","general",50,"2026-09-30T21:37:54.000Z",{"name":91,"slug":92,"count":93,"latest_published_at":94},"Reviews","reviews",31,"2026-09-28T14:31:34.000Z",{"name":96,"slug":97,"count":98,"latest_published_at":99},"How-To","how-to",7,"2026-10-01T09:00:00.000Z"]