{"id":"W4408559650","doi":"10.2196/63353","title":"Assessing ChatGPT’s Capability as a New Age Standardized Patient: Qualitative Study","year":2025,"lang":"en","type":"article","venue":"JMIR Medical Education","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":19,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Preprint; Qualitative research; Medicine; Psychology; Gerontology; Computer science; World Wide Web; Sociology; Social science","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.01406065,0.0003574469,0.0006373464,0.001437798,0.004382926,0.002620123,0.001686081,0.001284615,0.002952916],"category_scores_gemma":[0.0268289,0.0005031832,0.00042172,0.001042098,0.005729032,0.002751778,0.00508028,0.002251565,0.0003421564],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00427205,"about_ca_system_score_gemma":0.005239066,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00424958,"about_ca_topic_score_gemma":0.006707382,"domain_scores_codex":[0.9896797,0.007687866,0.0003869709,0.0004985858,0.0007275491,0.001019371],"domain_scores_gemma":[0.9775924,0.01416215,0.002242682,0.0005330603,0.002127238,0.003342466],"domain_codex":null,"domain_gemma":"methods","domain_candidate":"methods","domain_consensus":null,"study_design_codex":"qualitative","study_design_gemma":"qualitative","study_design_scores_codex":[0.00003807078,0.00009550004,0.008309925,0.0002009257,0.000005422026,0.0009248867,0.9840891,0.00003833163,0.0007985726,0.0005253875,0.0003898699,0.004583952],"study_design_scores_gemma":[0.000004180576,0.0001259043,0.002875927,0.0001369796,0.000003677234,0.0004130295,0.992604,0.0001031982,0.0002709592,0.0001074387,0.003341883,0.00001303173],"study_design_candidate":"qualitative","study_design_consensus":"qualitative","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9953573,0.0002271212,0.001191416,0.001164006,0.00003429766,0.0002378896,0.00009878004,0.00001089256,0.001678245],"genre_scores_gemma":[0.9967833,0.0003598707,0.0008449392,0.0005763359,0.00001520293,0.000299875,0.00004256686,0.00001019825,0.001067791],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.9859393,"threshold_uncertainty_score":0.07436067,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1212212797584189,"score_gpt":0.5691855747991221,"score_spread":0.4479642950407032,"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."}}