{"id":"W4415645090","doi":"10.1080/00016489.2025.2577834","title":"Evaluating the accuracy and reproducibility of ChatGPT responses in the context of cochlear implantation","year":2025,"lang":"en","type":"article","venue":"Acta Oto-Laryngologica","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"Toronto Western Hospital; Western University","funders":"","keywords":"Cochlear implantation; Context (archaeology); Reproducibility; Cochlear implant; Hearing loss","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":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.004830503,0.00007086407,0.0002066765,0.00007062607,0.00009725559,0.000007757821,0.0001350521,0.00007423024,0.00003618253],"category_scores_gemma":[0.02067952,0.00003801797,0.00003634487,0.0003054366,0.0001713785,0.00005617763,0.00003922389,0.0001868056,0.000001817544],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00002384593,"about_ca_system_score_gemma":0.0001912911,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001424861,"about_ca_topic_score_gemma":0.0001007745,"domain_scores_codex":[0.9983504,0.0005465987,0.0004980665,0.0003243891,0.0001571143,0.0001233758],"domain_scores_gemma":[0.9944859,0.004276264,0.0002044048,0.0008199873,0.0001966212,0.00001687309],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.003537137,0.0004566507,0.5808133,0.0005116679,0.0000487731,0.000004920537,0.0325066,0.000003960196,0.1411493,0.002754711,0.001020216,0.2371927],"study_design_scores_gemma":[0.0001062762,0.0008191313,0.9185534,0.0001783198,0.00005443712,0.00002187499,0.01865549,0.0003060274,0.05783109,0.002885483,0.0005388162,0.0000496482],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9836246,0.0005011908,0.000005784704,0.0143322,0.0001182181,0.000929188,0.000005604896,0.000008459129,0.0004747325],"genre_scores_gemma":[0.9979164,0.0002672024,0.00007395281,0.001559682,0.0000357847,0.00005554367,0.00001119465,0.000002653432,0.00007758274],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.3377401,"threshold_uncertainty_score":0.9875697,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2512198779434433,"score_gpt":0.501128908778339,"score_spread":0.2499090308348957,"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."}}