{"id":"W3080527613","doi":"10.2196/19918","title":"Is Artificial Intelligence Better Than Human Clinicians in Predicting Patient Outcomes?","year":2020,"lang":"en","type":"article","venue":"Journal of Medical Internet Research","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":38,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Calgary","funders":"University of Calgary","keywords":"Psychology; Computer science; Artificial intelligence; Medicine","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":["research_integrity","insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.00486636,0.0001398566,0.0005137719,0.0004241728,0.0000703937,0.00006253996,0.0005288701,0.0003200737,0.002459712],"category_scores_gemma":[0.0075907,0.0001050226,0.0002083244,0.0005847526,0.000275434,0.000136323,0.000174883,0.003212637,0.0001863961],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000225126,"about_ca_system_score_gemma":0.0006898423,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0009319423,"about_ca_topic_score_gemma":0.000293952,"domain_scores_codex":[0.9937201,0.0004114151,0.002006342,0.0002884988,0.003031265,0.0005424071],"domain_scores_gemma":[0.9968484,0.0009045767,0.0002980968,0.0002088813,0.0007843448,0.0009557176],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.0005457352,0.0008021906,0.5892813,0.0001608197,0.00009005619,0.000895677,0.02699248,0.000004645972,0.0002964104,0.0003966589,0.016468,0.3640661],"study_design_scores_gemma":[0.001545842,0.05698868,0.2345157,0.01774042,0.0003223481,0.001727635,0.2148923,0.1044301,0.23536,0.0566428,0.07395563,0.001878592],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8014085,0.000130505,0.0008034526,0.1966086,0.0004592116,0.0002256692,0.000001266589,0.000009849166,0.0003529998],"genre_scores_gemma":[0.9891122,0.0001151056,0.0001907143,0.00902706,0.001458163,0.000007995738,0.000002684574,0.00002067653,0.0000653797],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.3621875,"threshold_uncertainty_score":0.999087,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.5084866340275597,"score_gpt":0.5799768971393031,"score_spread":0.07149026311174345,"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."}}