{"id":"W4388014118","doi":"10.1002/art.42737","title":"Doctor Versus Artificial Intelligence: Patient and Physician Evaluation of Large Language Model Responses to Rheumatology Patient Questions in a <scp>Cross‐Sectional</scp> Study","year":2023,"lang":"en","type":"article","venue":"Arthritis & Rheumatology","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":63,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Alberta","funders":"","keywords":"Medicine; Readability; Rheumatology; Internal medicine; Likert scale; Chatbot; Cross-sectional study; Family medicine; Artificial intelligence; Pathology; Psychology","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":true,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001148846,0.0001923143,0.0004915759,0.000678065,0.0001888677,0.00002303309,0.00008484834,0.0001787697,0.00003418643],"category_scores_gemma":[0.004189312,0.0002081018,0.00006374287,0.0007942381,0.0001507978,0.0001184831,0.00009377082,0.000244221,0.000155396],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001472195,"about_ca_system_score_gemma":0.0004668053,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00165802,"about_ca_topic_score_gemma":0.00522592,"domain_scores_codex":[0.9968827,0.0004788152,0.001133803,0.0005164709,0.0004853849,0.0005027595],"domain_scores_gemma":[0.9976284,0.001096878,0.0002060718,0.0003416229,0.0005603281,0.0001667061],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0009962905,0.00218864,0.614915,0.0001193278,0.00009658356,0.0000478859,0.1230704,0.00389389,0.0008690127,0.009941102,0.0004111254,0.2434507],"study_design_scores_gemma":[0.002301748,0.01145117,0.4042574,0.001363555,0.0001754003,0.0004277918,0.3863298,0.1478983,0.01771832,0.02708015,0.0003499023,0.0006464082],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9961424,0.0006137859,0.0001581528,0.0004471153,0.0009725504,0.001441126,0.00004735512,0.00008684553,0.00009070482],"genre_scores_gemma":[0.9987922,0.0003103304,0.0001638927,0.0001057362,0.00001849784,0.000502346,0.00006846875,0.00002429334,0.00001421085],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.2632594,"threshold_uncertainty_score":0.8486146,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1205783328331111,"score_gpt":0.4459697096145198,"score_spread":0.3253913767814086,"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."}}