[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-ai-panels-replace-expert-delphi-surveys-for-risk-estimates":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},9808,"ai-panels-replace-expert-delphi-surveys-for-risk-estimates","AI Panels Replace Expert Delphi Surveys for Risk Estimates","Researchers swapped months-long expert panels for LLM personas that debate and revise risk estimates, and the numbers hold up against real experts.","A new technique teaches large language models to run Delphi-style risk assessments, the slow, expert-heavy process normally reserved for things nobody can directly measure.\n\nThe classical Delphi method gathers anonymous experts, has them estimate an unknown quantity, share reasoning, and revise their numbers over several rounds. It produces calibrated, auditable estimates but takes months of specialist time, which puts it out of reach for most teams. Researchers built \"Scalable Delphi,\" which runs the same protocol with multiple LLM personas instead of human panelists, iterating and sharing rationales automatically. Tested on AI-augmented cybersecurity risk and ice-sheet contributions to sea-level rise, across three benchmarks and three reproduced expert studies, the system's estimates improved as evidence was added, matched expert panels on most quantities, and correlated with ground truth at Pearson r=0.91-0.98.\n\nThat last part matters more than the benchmark numbers suggest. Delphi estimates exist precisely because the underlying quantities (novel cyberattack probabilities, ice-sheet melt decades out) can't be checked against a ground truth in real time. Making the process cheap and rerunnable means more organizations could get calibrated risk numbers on demand instead of commissioning a panel and waiting months, and testing which evidence moves the needle becomes a software operation instead of a committee exercise.\n\nStrong correlation with human experts is reassuring, not proof. Experts can be wrong together, especially about the unobservable. The real test for Scalable Delphi will come when it's used to make a call nobody can check for years.","[\"ai\",\"risk-assessment\",\"forecasting\",\"research\"]","2026-10-02T04:00:00.000Z","2026-10-03T11:27:49.439Z","2026-10-03T11:27:54.379Z","published",null,[],"ai",[24,26,27,28],"risk-assessment","forecasting","research",[30],{"name":31,"url":32},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.08889",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",6058,{"name":40,"slug":41,"count":42,"latest_published_at":18},"Security","security",848,{"name":44,"slug":45,"count":46,"latest_published_at":18},"Policy","policy",439,{"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",176,{"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"]