{"id":"W4396573797","doi":"10.1093/ced/llae158","title":"Assessing GPT-4’s diagnostic accuracy with darker skin tones: underperformance and implications","year":2024,"lang":"en","type":"letter","venue":"Clinical and Experimental Dermatology","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Medicine; Dermatology; Audiology","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":[0.01599082,0.000489838,0.0003537479,0.001044633,0.000435391,0.001248117,0.0007788248,0.0009043205,0.002390536],"category_scores_gemma":[0.09810396,0.0001627716,0.0004541447,0.000602719,0.001135319,0.001385801,0.00108637,0.00103856,0.001241101],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0008208288,"about_ca_system_score_gemma":0.0006387131,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002988943,"about_ca_topic_score_gemma":0.002458572,"domain_scores_codex":[0.9857696,0.007251839,0.001719189,0.0008527036,0.003451957,0.0009546714],"domain_scores_gemma":[0.9222111,0.05148475,0.00748063,0.004372113,0.012763,0.001688411],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0007630341,0.0001991509,0.9015266,0.0001341345,0.00007044905,0.001187167,0.001635253,0.001184526,0.004194051,0.0003510199,0.003422539,0.08533213],"study_design_scores_gemma":[0.00005626044,0.002089767,0.9307504,0.0002858659,0.0001358007,0.01101349,0.005842793,0.02309245,0.01447704,0.001994908,0.01018668,0.00007436981],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"commentary","genre_scores_codex":[0.9825714,0.0004990513,0.002540289,0.004749198,0.000260887,0.00006726099,0.0001557281,0.00007975627,0.009076416],"genre_scores_gemma":[0.9961588,0.0001444902,0.002185423,0.0005597926,0.00007870259,0.0000217617,0.00007035914,0.00001780439,0.0007629099],"genre_candidate":"commentary","genre_consensus":null,"teacher_disagreement_score":0.01599082,"threshold_uncertainty_score":0.08456856,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1912669036070457,"score_gpt":0.5078072894596163,"score_spread":0.3165403858525706,"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."}}