{"id":"W4385432609","doi":"10.2196/46800","title":"Assessing ChatGPT’s Capability for Multiple Choice Questions Using RaschOnline: Observational Study","year":2023,"lang":"en","type":"article","venue":"JMIR Formative Research","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Preprint; Mathematics education; Engineering; Psychology; 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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.006832447,0.0003681352,0.0005360076,0.001530024,0.0007556962,0.0009287852,0.0006896188,0.0006615061,0.001395031],"category_scores_gemma":[0.02515869,0.0003765734,0.0008383648,0.000933952,0.0008793385,0.0009667621,0.001403973,0.001371075,0.0005073149],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0005561165,"about_ca_system_score_gemma":0.00104061,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003747807,"about_ca_topic_score_gemma":0.005476093,"domain_scores_codex":[0.996299,0.001628119,0.0004227979,0.0004218971,0.0009235878,0.000304634],"domain_scores_gemma":[0.9835305,0.005875023,0.004169107,0.001828504,0.003072833,0.00152408],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.00008015019,0.0008588096,0.9774752,0.00005415909,0.00004167112,0.0003930335,0.01114021,0.0003164597,0.0007378358,0.00007594628,0.0002934695,0.008533001],"study_design_scores_gemma":[0.00001543405,0.001469674,0.9821188,0.00002799054,0.00002781544,0.000880159,0.01080712,0.002683545,0.0008590632,0.0001183243,0.0009531034,0.00003888162],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9991574,0.00001745686,0.000420874,0.00001854962,0.00000224727,0.0000705852,0.00006582308,0.000007249012,0.0002397681],"genre_scores_gemma":[0.9987104,0.00002738301,0.0007733135,0.00001881605,0.000003253406,0.00009987807,0.0001285323,0.000004067285,0.0002344136],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.006832447,"threshold_uncertainty_score":0.03613389,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.7643370418403466,"score_gpt":0.6629328663880035,"score_spread":0.101404175452343,"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."}}