{"id":"W4393082517","doi":"10.2196/55595","title":"Exploring the Performance of ChatGPT-4 in the Taiwan Audiologist Qualification Examination: Preliminary Observational Study Highlighting the Potential of AI Chatbots in Hearing Care","year":2024,"lang":"en","type":"article","venue":"JMIR Medical Education","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":19,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Audiologist; Audiology; Psychology; Rehabilitation; Health care; Medicine; Hearing loss; Neuroscience","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":true,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.006802095,0.0003549596,0.0003890015,0.001134801,0.001495095,0.001155786,0.0007064796,0.001260396,0.001813136],"category_scores_gemma":[0.03424026,0.00040411,0.0004714601,0.0004731053,0.0009540861,0.0009714553,0.002337051,0.00147325,0.0008114796],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001248528,"about_ca_system_score_gemma":0.002142763,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.006044167,"about_ca_topic_score_gemma":0.01135619,"domain_scores_codex":[0.9958106,0.002072405,0.0004403072,0.0003602815,0.0006438609,0.0006724975],"domain_scores_gemma":[0.9792238,0.007233224,0.005211481,0.001174369,0.003021655,0.00413545],"domain_codex":null,"domain_gemma":"evaluation","domain_candidate":"evaluation","domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0002792419,0.001796219,0.9190138,0.0001506868,0.00003238118,0.001653021,0.05734635,0.0002452124,0.002575717,0.0001353576,0.001566424,0.0152056],"study_design_scores_gemma":[0.00003860219,0.003332614,0.9175146,0.000110906,0.00002890101,0.001029952,0.06962092,0.002498908,0.001241951,0.0001432071,0.0043624,0.00007695463],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9992107,0.00001796668,0.0001715242,0.00007784665,0.000006892527,0.00008578384,0.00003806315,0.00000756216,0.0003835854],"genre_scores_gemma":[0.9982933,0.00004341036,0.0005693962,0.0001134966,0.00001105825,0.0002075803,0.0001048676,0.000006288508,0.0006505403],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.9931979,"threshold_uncertainty_score":0.03597337,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2756492139671869,"score_gpt":0.4586515506990061,"score_spread":0.1830023367318192,"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."}}