{"id":"W4410528874","doi":"10.2196/71434","title":"Evaluating the Usability, Technical Performance, and Accuracy of Artificial Intelligence Scribes for Primary Care: Competitive Analysis","year":2025,"lang":"en","type":"article","venue":"JMIR Human Factors","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":23,"is_retracted":false,"has_abstract":true,"ca_institutions":"Trillium Health Centre; Public Health Ontario; Women's College Hospital; University of Toronto","funders":"","keywords":"Preprint; Usability; Computer science; Artificial intelligence; Machine learning; Human–computer interaction; World Wide Web","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":true,"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":[0.0005395472,0.0001236534,0.0003286053,0.0002588936,0.0003507708,0.00002942493,0.0001327963,0.00009527611,0.0000424496],"category_scores_gemma":[0.000567698,0.00008654933,0.0001457541,0.0006807982,0.0003583896,0.00008463281,0.00006383473,0.0001977975,0.00000138627],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001587827,"about_ca_system_score_gemma":0.0002466349,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002007195,"about_ca_topic_score_gemma":0.0002554648,"domain_scores_codex":[0.998648,0.00006315159,0.0005924459,0.0002713611,0.0002333982,0.0001916849],"domain_scores_gemma":[0.9979258,0.001069856,0.0001513591,0.0003150017,0.0004839294,0.00005400671],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0003734271,0.0003491399,0.5974587,0.001698151,0.0002774877,2.17732e-7,0.01087437,0.0001052876,0.009541866,0.02229445,0.00008067976,0.3569463],"study_design_scores_gemma":[0.00004779727,0.001404998,0.9120637,0.0003360138,0.0009743918,0.000001015825,0.02321098,0.001711965,0.05518042,0.004448703,0.000424016,0.0001960125],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.996299,0.0001239878,0.000779741,0.0004082373,0.0001217155,0.001226655,0.00001284183,0.00002822889,0.0009996245],"genre_scores_gemma":[0.9990261,0.0000225686,0.0004404228,0.0001291758,0.00007789519,0.0001425585,0.00009208001,0.000006618396,0.00006263244],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.3567502,"threshold_uncertainty_score":0.3529379,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2285946247328254,"score_gpt":0.5077540762863881,"score_spread":0.2791594515535627,"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."}}