{"id":"W1599059832","doi":"10.1177/0272989x15598743","title":"What Is Insufficient about Validation?","year":2015,"lang":"en","type":"letter","venue":"Medical Decision Making","topic":"Health Systems, Economic Evaluations, Quality of Life","field":"Economics, Econometrics and Finance","cited_by":1,"is_retracted":false,"has_abstract":false,"ca_institutions":"McGill University","funders":"","keywords":"Computer science","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":["metaresearch","metaepi_narrow","research_integrity","insufficient_payload"],"consensus_categories":["metaresearch","insufficient_payload"],"category_scores_codex":[0.03411198,0.0005051533,0.002046404,0.0009284636,0.0002652157,0.0007654139,0.001314414,0.001963042,0.01312504],"category_scores_gemma":[0.02136544,0.0005763458,0.000353132,0.0004251575,0.0001585922,0.0008476747,0.0003694493,0.002081422,0.02271523],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001291648,"about_ca_system_score_gemma":0.0007590051,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00007620823,"about_ca_topic_score_gemma":0.00001169849,"domain_scores_codex":[0.9881149,0.0005800957,0.007505831,0.001464176,0.001471603,0.0008634036],"domain_scores_gemma":[0.9895874,0.004509166,0.003732721,0.001443156,0.0002766123,0.0004509433],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"not_applicable","study_design_scores_codex":[0.000006156516,0.00004432467,0.0002910277,0.0002050598,0.00004655623,0.00005288401,0.0008460588,0.00002455744,1.042292e-8,0.001496188,0.9892942,0.007692968],"study_design_scores_gemma":[0.0005362414,0.00005549099,0.0001939478,0.002062632,0.00001099596,0.00002664503,0.0002147344,0.00451283,2.74516e-7,0.04041384,0.951418,0.0005543765],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"commentary","genre_gemma":"commentary","genre_scores_codex":[0.001599206,0.01070153,0.02818695,0.9417273,0.01164962,0.0007746462,0.0002935531,0.0001234059,0.004943748],"genre_scores_gemma":[0.001682042,0.0009879501,0.003225539,0.9829235,0.007538003,0.0001098157,0.0002615913,0.0001226721,0.003148904],"genre_candidate":"commentary","genre_consensus":"commentary","teacher_disagreement_score":0.04119614,"threshold_uncertainty_score":0.9996688,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.3707662617076382,"score_gpt":0.4802594566583135,"score_spread":0.1094931949506753,"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."}}