{"id":"W4392350413","doi":"10.52609/jmlph.v4i1.113","title":"Evaluating the Precision of ChatGPT Artificial Intelligence in Emergency Differential Diagnosis","year":2024,"lang":"en","type":"article","venue":"The Journal of Medicine Law & Public Health","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Medical diagnosis; Triage; Artificial intelligence; Differential diagnosis; Computer science; Differential (mechanical device); Human intelligence; Machine learning; Medical emergency; Medicine; Engineering; Pathology","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":"codex-gemma-dda1882f352a","candidate_categories":["insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.01378807,0.0001587425,0.0005230703,0.00030683,0.0001996433,0.00001996026,0.0003627205,0.0000862254,0.0009760454],"category_scores_gemma":[0.002749726,0.00007549326,0.000120022,0.001030522,0.000343119,0.0001844334,0.00004542634,0.0009027309,0.00001277459],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002447394,"about_ca_system_score_gemma":0.001337993,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003313483,"about_ca_topic_score_gemma":0.000752899,"domain_scores_codex":[0.9948973,0.0009160311,0.002475558,0.0001678132,0.001113197,0.0004300798],"domain_scores_gemma":[0.9963,0.001812032,0.000666621,0.0003707325,0.0005703997,0.000280181],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.0002252761,0.0002531048,0.001030913,0.000409236,0.00004933473,0.000006187556,0.02657096,0.0000509488,0.0004896565,0.02043693,0.002344204,0.9481332],"study_design_scores_gemma":[0.00104081,0.06162416,0.04131907,0.03177305,0.001478654,0.002711183,0.1689608,0.1244924,0.01597096,0.4863448,0.06308062,0.001203414],"study_design_candidate":"design_other","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6035377,0.01047518,0.005697589,0.3760306,0.003415244,0.0005875287,0.000001944044,0.00001461752,0.0002395795],"genre_scores_gemma":[0.9862383,0.009767518,0.000126926,0.001387683,0.002397341,0.00001612105,0.000003917401,0.00001943046,0.00004276347],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.9469298,"threshold_uncertainty_score":0.9999372,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.4049069743792399,"score_gpt":0.5374027853009822,"score_spread":0.1324958109217422,"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."}}