{"id":"W4413940306","doi":"10.1002/bcp.70223","title":"Assessing ChatGPT's capability in understanding and reporting antiretroviral therapy drug–drug interaction effects: Quantitative and qualitative results from the ACCURATE‐DDI study","year":2025,"lang":"en","type":"article","venue":"British Journal of Clinical Pharmacology","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto; Toronto General Hospital; St. Michael's Hospital; St. Paul's Hospital; University of British Columbia","funders":"","keywords":"Drug; Drug-drug interaction; Medicine; Antiretroviral therapy; Drug interaction; Pharmacology; Pharmacotherapy; Antiretroviral drug; Human immunodeficiency virus (HIV); Intensive care medicine; Internal medicine; Family medicine; Viral load","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"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":[0.05668836,0.0004536873,0.0006332684,0.002725147,0.0008906654,0.001812284,0.001128377,0.0007883564,0.001837623],"category_scores_gemma":[0.2048368,0.0003891731,0.0008425862,0.001610954,0.00292223,0.002070557,0.005821905,0.001053848,0.0003797486],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002390636,"about_ca_system_score_gemma":0.001834781,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004392762,"about_ca_topic_score_gemma":0.003847828,"domain_scores_codex":[0.9521013,0.03638731,0.003032486,0.001642299,0.00581765,0.001018941],"domain_scores_gemma":[0.5688995,0.3723991,0.01744492,0.01034107,0.02706114,0.003854361],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.001450507,0.0006008483,0.636299,0.001551577,0.0002911958,0.0009237607,0.2365942,0.002110438,0.003883476,0.00112407,0.005614639,0.1095564],"study_design_scores_gemma":[0.0002153873,0.002593224,0.7387536,0.001027886,0.0002865567,0.001761479,0.2152766,0.0131553,0.005924875,0.002411852,0.01829641,0.0002968339],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9896263,0.000159922,0.003647637,0.0004973806,0.0000258251,0.0006746344,0.0007410999,0.00009562767,0.004531547],"genre_scores_gemma":[0.990506,0.000133104,0.007026117,0.0002764962,0.00001785286,0.0008701149,0.0003826317,0.00003741127,0.0007502136],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.05668836,"threshold_uncertainty_score":0.2998003,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.5475769401988588,"score_gpt":0.6490025932472533,"score_spread":0.1014256530483945,"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."}}