{"id":"W4400079353","doi":"10.2196/59213","title":"A Language Model–Powered Simulated Patient With Automated Feedback for History Taking: Prospective Study","year":2024,"lang":"en","type":"article","venue":"JMIR Medical Education","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":114,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Preprint; Computer science; World Wide Web","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":true,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0002404321,0.0001511813,0.0002065151,0.0001750808,0.00006873003,0.00002384855,0.00006823505,0.0001499918,0.0005039118],"category_scores_gemma":[0.0006437923,0.0001152326,0.00005245694,0.000292116,0.00006846559,0.0001078071,0.00001162093,0.0002738964,0.00005203618],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001042941,"about_ca_system_score_gemma":0.005492847,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0004146411,"about_ca_topic_score_gemma":0.0000730703,"domain_scores_codex":[0.998439,0.00004176775,0.0003859408,0.0003941761,0.0004995809,0.0002395674],"domain_scores_gemma":[0.9988862,0.0001193028,0.00009331457,0.0002409469,0.0003582806,0.0003019774],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"qualitative","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.002058985,0.01243879,0.009570227,0.001661622,0.0003407432,0.00009540101,0.4443128,0.0003599001,0.0006161158,0.001078021,0.1163179,0.4111495],"study_design_scores_gemma":[0.0006221636,0.005104125,0.008153191,0.001940866,0.0003117405,0.0001198412,0.06813852,0.8998412,0.0009330871,0.0005982848,0.01372992,0.0005070179],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9887462,0.000996251,0.0004677292,0.003149746,0.001340991,0.003598179,0.000003103302,0.0006379496,0.001059907],"genre_scores_gemma":[0.9959389,0.00000687143,0.0004032873,0.001015314,0.0004058989,0.0009892893,0.0001135324,0.0000399758,0.00108693],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.8994814,"threshold_uncertainty_score":0.9744075,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.06459609217325434,"score_gpt":0.4453597336472424,"score_spread":0.380763641473988,"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."}}