{"id":"W4205328242","doi":"10.7554/elife.72737","title":"Meta-Research: investigating disagreement in the scientific literature","year":2021,"lang":"en","type":"article","venue":"DOAJ (DOAJ: Directory of Open Access Journals)","topic":"Misinformation and Its Impacts","field":"Social Sciences","cited_by":35,"is_retracted":false,"has_abstract":true,"ca_institutions":"Université de Montréal","funders":"Air Force Office of Scientific Research; Canada Research Chairs","keywords":"Discipline; Epistemology; Social science; Physical science; Biological sciences; Sociology; Biology; Philosophy","routes":{"ca_aff":true,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[{"model":"gemma","categories":["metaresearch"],"domain":"methods","study_design":"theoretical_or_conceptual","genre":"empirical","about_ca_system":false,"about_ca_topic":false,"confidence":"low","status":"direct model label, unvalidated"},{"model":"gpt","categories":["metaresearch"],"domain":"evaluation","study_design":"observational","genre":"empirical","about_ca_system":false,"about_ca_topic":false,"confidence":"high","status":"direct model label, unvalidated"}],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["sts","scholarly_communication","insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.01808669,0.0001624548,0.0004076098,0.0007168399,0.001533191,0.01188493,0.002962675,0.00009006386,0.01345009],"category_scores_gemma":[0.00462477,0.0001087557,0.0001818844,0.005291317,0.0006056289,0.004585669,0.0006380873,0.0006913483,0.00003279693],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001231124,"about_ca_system_score_gemma":0.00078866,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001093384,"about_ca_topic_score_gemma":0.001560902,"domain_scores_codex":[0.9935699,0.002218755,0.0007028428,0.0002924005,0.002666549,0.000549551],"domain_scores_gemma":[0.9970974,0.0008802174,0.0003948845,0.0004722425,0.0008504475,0.0003048281],"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":[0.00003633051,0.0006181231,0.07926673,0.0001941107,0.0006516741,0.0001297697,0.3798078,0.0002257373,0.0231144,0.02350672,0.4647048,0.02774382],"study_design_scores_gemma":[0.000794822,0.00001280439,0.3882478,0.001256576,0.0002742727,0.00001651361,0.07557075,0.0001333609,0.01466714,0.0759607,0.4423199,0.0007453472],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8853966,0.0249993,0.00002673885,0.006281876,0.0007340948,0.0006662301,0.00003518151,0.00002034015,0.08183964],"genre_scores_gemma":[0.9922897,0.003160148,0.0001685584,0.001261568,0.0001775208,0.00001810065,0.00001815054,0.00001173278,0.002894518],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.3089811,"threshold_uncertainty_score":0.9997666,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.7736723904124272,"score_gpt":0.6687191692555318,"score_spread":0.1049532211568954,"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."}}