{"id":"W4403947637","doi":"10.1017/dmp.2024.194","title":"Repeatability, Reproducibility, and Diagnostic Accuracy of a Commercial Large Language Model (ChatGPT) to Perform Disaster Triage Using the Simple Triage and Rapid Treatment (START) Protocol","year":2024,"lang":"en","type":"article","venue":"Disaster Medicine and Public Health Preparedness","topic":"COVID-19 diagnosis using AI","field":"Medicine","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Alberta","funders":"","keywords":"Triage; Repeatability; Reproducibility; Protocol (science); Simple (philosophy); Computer science; Medical emergency; Medicine; Statistics; Pathology; Mathematics","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.1370146,0.0008408886,0.0008105074,0.001223464,0.0007623102,0.002223037,0.0015258,0.001189681,0.002175855],"category_scores_gemma":[0.3624922,0.000637911,0.001320139,0.0006617003,0.001557699,0.001861462,0.002506982,0.001417495,0.001662047],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001611581,"about_ca_system_score_gemma":0.002076964,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002020875,"about_ca_topic_score_gemma":0.002009084,"domain_scores_codex":[0.856838,0.1055358,0.01452625,0.009421884,0.01194622,0.001731824],"domain_scores_gemma":[0.6960599,0.2160862,0.01704729,0.03495394,0.03378638,0.00206626],"domain_codex":null,"domain_gemma":"reproducibility","domain_candidate":"reproducibility","domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.01446221,0.003195584,0.5578787,0.002410642,0.001542202,0.0009020072,0.03008587,0.0215239,0.02282387,0.002794273,0.03161681,0.3107639],"study_design_scores_gemma":[0.00221425,0.01767809,0.5835519,0.001818806,0.00158899,0.00272696,0.01173133,0.2602815,0.06521256,0.006592586,0.04549783,0.001105177],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8981349,0.0004251003,0.08595272,0.001246605,0.0007532167,0.004776497,0.001910304,0.00289914,0.0039016],"genre_scores_gemma":[0.9510774,0.0001065794,0.04191434,0.0007858875,0.0001047715,0.003386872,0.001441911,0.0004515732,0.0007307122],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.8629854,"threshold_uncertainty_score":0.724611,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1553389319643564,"score_gpt":0.4616742235302203,"score_spread":0.3063352915658639,"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."}}