{"id":"W4365510493","doi":"10.1038/s41591-023-02294-8","title":"Concerns about the Burden of Proof studies","year":2023,"lang":"en","type":"letter","venue":"Nature Medicine","topic":"Economic, financial, and policy analysis","field":"Economics, Econometrics and Finance","cited_by":13,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto; St. Michael's Hospital","funders":"Canadian Institutes of Health Research; Government of Canada; Foundation for the National Institutes of Health","keywords":"Burden of proof; Proof of concept; Medicine; Computer science; Political science; Law","routes":{"ca_aff":true,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow","research_integrity"],"consensus_categories":["research_integrity"],"category_scores_codex":[0.001075424,0.0004035376,0.001882874,0.000558564,0.0001337474,0.00001624755,0.0008344147,0.001357767,0.0004014024],"category_scores_gemma":[0.001234715,0.0002891901,0.0003646859,0.000587118,0.0006217593,0.00005990838,0.000126129,0.002933282,0.0003787738],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001395884,"about_ca_system_score_gemma":0.00005363076,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001733328,"about_ca_topic_score_gemma":0.0001269851,"domain_scores_codex":[0.9974754,0.00004148927,0.001302749,0.0006153885,0.0001118411,0.0004530805],"domain_scores_gemma":[0.9968597,0.0006899921,0.001372244,0.0008970354,0.0001399224,0.00004107235],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"not_applicable","study_design_scores_codex":[0.000005728777,0.000004431339,0.0002135473,0.0003341684,0.0007594864,0.00002749324,0.002486538,0.00002135608,0.00000126175,0.01214749,0.9837651,0.0002334592],"study_design_scores_gemma":[0.0003936282,0.00008568676,0.0002188047,0.0002782253,0.0001555198,0.00000325403,0.0002737659,0.00007497934,0.00002394531,0.02373026,0.9744841,0.0002777746],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"commentary","genre_gemma":"commentary","genre_scores_codex":[0.001015419,0.09578627,0.00001619457,0.8800818,0.003315721,0.0003804013,0.0006716918,0.00004745338,0.01868509],"genre_scores_gemma":[0.06700642,0.01776901,0.00004271263,0.7195569,0.09222674,0.0001741085,0.0006756469,0.000230906,0.1023175],"genre_candidate":"commentary","genre_consensus":"commentary","teacher_disagreement_score":0.1605248,"threshold_uncertainty_score":0.999956,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08112919910884163,"score_gpt":0.3223733159659963,"score_spread":0.2412441168571546,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}