{"id":"W4322208105","doi":"10.1101/2023.02.25.23286117","title":"neuroGPT-X: Towards an Accountable Expert Opinion Tool for Vestibular Schwannoma","year":2023,"lang":"en","type":"preprint","venue":"medRxiv","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":12,"is_retracted":false,"has_abstract":true,"ca_institutions":"Dalhousie University; University of British Columbia; University of Calgary","funders":"","keywords":"Context (archaeology); Psychology; Medicine; Family medicine; Medical education; Geography","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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0008018856,0.0003089833,0.0004830129,0.0002229379,0.0001835477,0.0001042578,0.0003050538,0.0004709663,0.000103853],"category_scores_gemma":[0.001042152,0.0002995254,0.0001370982,0.000199844,0.00007435907,0.0001481517,0.0001765017,0.0005725607,0.0002015029],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001840298,"about_ca_system_score_gemma":0.001136979,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002233655,"about_ca_topic_score_gemma":0.00004999265,"domain_scores_codex":[0.9975837,0.00007522158,0.0006680833,0.0007963222,0.0003787928,0.0004979216],"domain_scores_gemma":[0.9979244,0.0001421654,0.0001812997,0.001006445,0.000496179,0.0002494477],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"not_applicable","study_design_scores_codex":[0.002196829,0.002289714,0.2165481,0.008968898,0.0005360791,0.0002037665,0.01957605,0.00205782,0.01815891,0.003429394,0.1293129,0.5967215],"study_design_scores_gemma":[0.0004877931,0.004444767,0.1164979,0.003412811,0.0002796272,0.000131688,0.003481527,0.04082652,0.07285757,0.03999774,0.7152014,0.002380596],"study_design_candidate":"not_applicable","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9703951,0.0009940363,0.00355794,0.01239129,0.009889759,0.002073556,0.00006181501,0.000459776,0.0001767111],"genre_scores_gemma":[0.975217,0.00160368,0.006757509,0.002422018,0.008676945,0.001745957,0.001693663,0.0002273493,0.001655901],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.5943409,"threshold_uncertainty_score":0.9999457,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2848410309325434,"score_gpt":0.4709771573013682,"score_spread":0.1861361263688248,"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."}}