{"id":"W4403321301","doi":"10.1186/s12909-024-06115-5","title":"The future of AI clinicians: assessing the modern standard of chatbots and their approach to diagnostic uncertainty","year":2024,"lang":"en","type":"article","venue":"BMC Medical Education","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":11,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto; St. Michael's Hospital","funders":"","keywords":"Medicine; Correctness; MEDLINE; Interpretation (philosophy); Family medicine; Medical physics; Medical education; Artificial intelligence; Computer science","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.1222459,0.0003308528,0.0004484857,0.002505664,0.001957844,0.006235208,0.001501613,0.001542295,0.001573807],"category_scores_gemma":[0.3737449,0.0004350875,0.0004318949,0.0008658923,0.004151835,0.00590153,0.00451069,0.002222135,0.0004658894],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.004933014,"about_ca_system_score_gemma":0.006521396,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003430587,"about_ca_topic_score_gemma":0.003950008,"domain_scores_codex":[0.934199,0.04608456,0.00351922,0.002836215,0.01192044,0.001440525],"domain_scores_gemma":[0.6303095,0.2509722,0.04719742,0.01253401,0.04509009,0.01389673],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0008962708,0.0006381645,0.7993146,0.0004654374,0.0001438048,0.0001768573,0.04817948,0.001013331,0.000616962,0.002220262,0.002468387,0.1438665],"study_design_scores_gemma":[0.0001975598,0.005042612,0.8800967,0.002024868,0.0003149556,0.001433169,0.05876826,0.01916206,0.003351587,0.008015438,0.02130341,0.0002894247],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9739782,0.002185831,0.007556763,0.007423572,0.0001671968,0.0003988906,0.0001139876,0.0001201531,0.00805541],"genre_scores_gemma":[0.9915133,0.0004190944,0.006980259,0.000547639,0.00004648684,0.0001879375,0.00004638512,0.0000136814,0.0002452011],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.1222459,"threshold_uncertainty_score":0.6465059,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08794833893524416,"score_gpt":0.4630393188298464,"score_spread":0.3750909798946022,"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."}}