{"id":"W4406945781","doi":"10.2196/66207","title":"Medical Misinformation in AI-Assisted Self-Diagnosis: Development of a Method (EvalPrompt) for Analyzing Large Language Models","year":2025,"lang":"en","type":"article","venue":"JMIR Formative Research","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":25,"is_retracted":false,"has_abstract":true,"ca_institutions":"St. Michael's Hospital; University of Toronto; University of Waterloo","funders":"","keywords":"Preprint; Misinformation; Computer science; Psychology; Artificial intelligence; Natural language processing; World Wide Web; Computer security","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":true,"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":[0.01236632,0.001377644,0.0007699523,0.002062063,0.0005503664,0.002484016,0.001909641,0.001781168,0.003729126],"category_scores_gemma":[0.06774455,0.0005971896,0.001218541,0.0008025423,0.001030965,0.003498024,0.003482008,0.002738131,0.001115368],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001161898,"about_ca_system_score_gemma":0.001510026,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001993995,"about_ca_topic_score_gemma":0.002667656,"domain_scores_codex":[0.9905304,0.006104835,0.0006076095,0.001312807,0.001295098,0.0001492006],"domain_scores_gemma":[0.9246541,0.06747018,0.002121083,0.003040747,0.002316578,0.000397378],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.0008230475,0.0006783546,0.01533006,0.0009219809,0.0003400439,0.001061531,0.004939689,0.08792328,0.01367271,0.01899397,0.015152,0.8401633],"study_design_scores_gemma":[0.00005571809,0.0001190092,0.001564548,0.00007285077,0.00003109491,0.0003556164,0.0005632259,0.9671533,0.008016926,0.01694897,0.00507918,0.00003955882],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.01955841,0.0001106019,0.9687812,0.0003633832,0.00003770809,0.0003452989,0.0004030841,0.009553546,0.0008467609],"genre_scores_gemma":[0.1665168,0.00006605004,0.8305568,0.0002023824,0.00003076446,0.0005524803,0.0008785407,0.0004101314,0.0007861411],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.01236632,"threshold_uncertainty_score":0.06540018,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.219269846472011,"score_gpt":0.5887760165265201,"score_spread":0.3695061700545091,"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."}}