{"id":"W4410881775","doi":"10.2196/69521","title":"Chatbots’ Role in Generating Single Best Answer Questions for Undergraduate Medical Student Assessment: Comparative Analysis","year":2025,"lang":"en","type":"article","venue":"JMIR Medical Education","topic":"AI in Service Interactions","field":"Computer Science","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Chatbot; Consistency (knowledge bases); Computer science; Test (biology); Quality (philosophy); Data science; Scale (ratio); Medical education; Psychology; Artificial intelligence; Medicine","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":[],"consensus_categories":[],"category_scores_codex":[0.08008385,0.0004729868,0.0008186584,0.005630756,0.001229205,0.002337751,0.001500846,0.0009931476,0.002410135],"category_scores_gemma":[0.2754615,0.0004816685,0.001158912,0.002819347,0.001598765,0.002704041,0.004004675,0.001014495,0.0007972005],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00227951,"about_ca_system_score_gemma":0.002502911,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001595855,"about_ca_topic_score_gemma":0.003086543,"domain_scores_codex":[0.9260542,0.05420369,0.006773857,0.003290634,0.008201482,0.001476104],"domain_scores_gemma":[0.5912045,0.3382192,0.01334225,0.01010367,0.04255834,0.004572],"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":[0.007998962,0.003492356,0.4315792,0.002786623,0.0007385752,0.0003835222,0.105855,0.001731037,0.005483981,0.002575913,0.002555178,0.4348196],"study_design_scores_gemma":[0.0007421669,0.01980495,0.8707122,0.001909731,0.001086191,0.001110711,0.05949295,0.01748844,0.01207485,0.002067004,0.01323863,0.0002722017],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9864049,0.0008899177,0.007006892,0.0001301997,0.00005054588,0.001162459,0.0001966036,0.0001449451,0.004013481],"genre_scores_gemma":[0.988727,0.0002734746,0.00921136,0.00005846537,0.00002091485,0.001004455,0.0002261602,0.00004105182,0.0004371046],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.08008385,"threshold_uncertainty_score":0.4235289,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02190070761803816,"score_gpt":0.4410937310131128,"score_spread":0.4191930233950747,"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."}}