{"id":"W4411272806","doi":"10.1073/pnas.2426153122","title":"Human–AI collectives most accurately diagnose clinical vignettes","year":2025,"lang":"en","type":"article","venue":"Proceedings of the National Academy of Sciences","topic":"Clinical Reasoning and Diagnostic Skills","field":"Medicine","cited_by":30,"is_retracted":false,"has_abstract":true,"ca_institutions":"Innovation Cluster (Canada)","funders":"Max-Planck-Institut für Bildungsforschung; HORIZON EUROPE Framework Programme; Universität zu Köln; European Commission; Max-Planck-Gesellschaft; Deutsche Forschungsgemeinschaft; HORIZON EUROPE European Research Council; Kaiser Permanente; University of Oxford; Harvey Mudd College","keywords":"Artificial intelligence; Psychology; Computer science; History; Natural language processing","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":[],"consensus_categories":[],"category_scores_codex":[0.008536821,0.001030286,0.0008072293,0.004655598,0.001470311,0.003045232,0.001048332,0.002123883,0.003014611],"category_scores_gemma":[0.04964059,0.0003692583,0.0009816338,0.001733596,0.0008811192,0.002437689,0.002819461,0.001387254,0.00129526],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001200477,"about_ca_system_score_gemma":0.001279139,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0036957,"about_ca_topic_score_gemma":0.008485151,"domain_scores_codex":[0.9936587,0.003006223,0.000474891,0.00171179,0.000852045,0.0002963383],"domain_scores_gemma":[0.9808029,0.01268355,0.001896028,0.002111192,0.00160349,0.0009029815],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"observational","study_design_scores_codex":[0.001358437,0.0009776071,0.30721,0.0008582701,0.00102318,0.001317255,0.0147595,0.04672363,0.01686459,0.004821827,0.01982821,0.5842575],"study_design_scores_gemma":[0.0002125704,0.0007207663,0.1487562,0.0006360085,0.0009860246,0.002202954,0.01512262,0.7304406,0.0216086,0.04656355,0.03241595,0.0003341807],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.84482,0.001411302,0.1281744,0.005712661,0.0002383229,0.0005981509,0.00133903,0.002709105,0.01499707],"genre_scores_gemma":[0.9218639,0.0001879825,0.07493941,0.0005105477,0.00009756098,0.00008432087,0.001121609,0.00007648336,0.001118139],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.008536821,"threshold_uncertainty_score":0.04514754,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1042687726226228,"score_gpt":0.4661704352003379,"score_spread":0.3619016625777151,"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."}}