{"id":"W2134023402","doi":"10.1177/0272989x12454579","title":"Model Transparency and Validation","year":2012,"lang":"en","type":"article","venue":"Medical Decision Making","topic":"Health Systems, Economic Evaluations, Quality of Life","field":"Economics, Econometrics and Finance","cited_by":791,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"","keywords":"Transparency (behavior); Computer science; Face validity; External validity; Rigour; Documentation; Management science; Health care; Medicine; Psychology; Engineering; Psychometrics","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":["insufficient_payload"],"consensus_categories":["insufficient_payload"],"category_scores_codex":[0.0164631,0.0001176926,0.0004720138,0.0001922769,0.0001476919,0.00004647834,0.0001797991,0.0001796421,0.001461884],"category_scores_gemma":[0.007485583,0.0001324737,0.00005758966,0.0001183768,0.00004916443,0.0005348016,0.00004880547,0.0001669528,0.0009304354],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001237682,"about_ca_system_score_gemma":0.00005387572,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00001994134,"about_ca_topic_score_gemma":0.000009213353,"domain_scores_codex":[0.9967004,0.0001156558,0.002283828,0.0003127479,0.0002327426,0.0003546267],"domain_scores_gemma":[0.9975901,0.001235346,0.0005491584,0.0002657506,0.00002324879,0.0003364405],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0000429458,0.0002814368,0.2599242,0.0003319355,0.00005461698,0.000002822229,0.006328354,0.001012501,0.000006102747,0.6050491,0.03954737,0.08741859],"study_design_scores_gemma":[0.001496373,0.00006328751,0.07184599,0.0006243041,0.00001416017,0.00003426859,0.0004184522,0.6032331,0.000007254199,0.2783571,0.04322168,0.0006840382],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.3345047,0.002650848,0.6483381,0.008311957,0.0006612546,0.000221618,0.00003163701,0.00004969765,0.005230145],"genre_scores_gemma":[0.9712109,0.000151962,0.02166398,0.006481512,0.000338677,0.00002446358,0.000007210015,0.00001841235,0.0001029082],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.6367062,"threshold_uncertainty_score":0.9998475,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.4916667956298109,"score_gpt":0.4899938086028957,"score_spread":0.001672987026915207,"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."}}