[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-researchers-build-ai-that-asks-smarter-diagnostic-questions":10,"sections":44},{"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":34,"tags":35,"sources":39,"feedback":43,"feedback_at":22,"cost_usd":43,"total_tokens":43},8357,"researchers-build-ai-that-asks-smarter-diagnostic-questions","Researchers Build AI That Asks Smarter Diagnostic Questions","Researchers built ConsultMind, an uncertainty-aware AI framework that picks the next diagnostic question and sharply lifts accuracy in simulated consultations.","A new AI framework decides which question to ask next during a simulated medical workup, and it measurably improves how often the right diagnosis comes out on top.\n\nThe system, called ConsultMind, is built on a pipeline named AutoDisym that automatically constructs a Disorder-Symptom Bayesian Network from clinical knowledge and real diagnosis-labeled patient narratives. As a virtual patient answers questions, ConsultMind updates its probability estimates for each possible disorder and uses the remaining uncertainty to choose the next, most informative question. The researchers tested it across psychiatry, respiratory medicine, and fever-clinic scenarios plus three public datasets. Compared to baseline approaches, ConsultMind improved top-1 diagnostic accuracy by as much as 22.15 percentage points and top-3 accuracy by up to 37.89 points.\n\nMost AI symptom checkers still work off static decision trees or a single large-language-model prompt guessing at what to ask next. Tying the question-selection process to an explicit, updating probability model is closer to how an actual clinician narrows a differential diagnosis, and it gives the system a built-in way to explain its reasoning rather than just outputting a guess. Physician reviewers in the study also rated ConsultMind's differential diagnoses and rationales as clearer, which matters more for adoption than raw accuracy numbers do.\n\nWorth remembering: this is a benchmark study with simulated patients and physician graders, not a clinical deployment, so the jump from a well-scored transcript to something a hospital would actually trust is still an open question.","[\"ai\",\"healthcare\",\"medical-diagnosis\",\"bayesian-networks\"]","2026-09-28T04:00:00.000Z","2026-09-29T02:43:35.952Z","2026-09-29T02:43:43.022Z","published",null,[24,30],{"id":25,"reviewer":26,"round":27,"reason":28,"status":29},"editor-r1","editor",1,"Verify that \"GPT-5.6-Sol\" is an actual confirmed model in the vendor's current lineup before publishing—no such model can be confirmed, so either substantiate it with a source or remove\u002Frephrase the reference rather than repeating the paper's unverified name.","resolved",{"id":31,"reviewer":26,"round":32,"reason":33,"status":29},"editor-r2",2,"Don't name-check the unverifiable 'GPT-5.6-Sol' model and then flag it as unconfirmed in the same sentence — either drop that specific F1-score benchmark and its model name entirely, or independently verify the model exists before citing it; self-flagging an unresolved inconsistency inline isn't a publishable fix.","ai",[34,36,37,38],"healthcare","medical-diagnosis","bayesian-networks",[40],{"name":41,"url":42},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.30796",0,{"sections":45},[46,50,55,60,65,70,75,80,85,90,95,100,105,110],{"name":47,"slug":34,"count":48,"latest_published_at":49},"AI",4918,"2026-09-29T05:13:43.000Z",{"name":51,"slug":52,"count":53,"latest_published_at":54},"Security","security",768,"2026-09-29T01:20:00.000Z",{"name":56,"slug":57,"count":58,"latest_published_at":59},"Policy","policy",406,"2026-09-28T19:04:09.000Z",{"name":61,"slug":62,"count":63,"latest_published_at":64},"Deals","deals",272,"2026-09-28T17:42:46.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":69},"Hardware","hardware",191,"2026-09-28T15:45:00.000Z",{"name":71,"slug":72,"count":73,"latest_published_at":74},"Science","science",154,"2026-09-28T13:19:18.000Z",{"name":76,"slug":77,"count":78,"latest_published_at":79},"Consumer Tech","consumer-tech",139,"2026-09-28T17:09:47.000Z",{"name":81,"slug":82,"count":83,"latest_published_at":84},"Software","software",91,"2026-09-25T20:55:00.000Z",{"name":86,"slug":87,"count":88,"latest_published_at":89},"Dev Tools","dev-tools",87,"2026-09-28T16:11:42.000Z",{"name":91,"slug":92,"count":93,"latest_published_at":94},"Startups","startups",80,"2026-09-28T17:50:28.000Z",{"name":96,"slug":97,"count":98,"latest_published_at":99},"General","general",49,"2026-09-28T16:44:57.000Z",{"name":101,"slug":102,"count":103,"latest_published_at":104},"Gaming","gaming",48,"2026-09-25T18:35:21.000Z",{"name":106,"slug":107,"count":108,"latest_published_at":109},"Reviews","reviews",31,"2026-09-28T14:31:34.000Z",{"name":111,"slug":112,"count":113,"latest_published_at":114},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]