{"id":"W3126829124","doi":"10.1038/s42003-021-01674-5","title":"Systematic auditing is essential to debiasing machine learning in biology","year":2021,"lang":"en","type":"article","venue":"Communications Biology","topic":"Cell Image Analysis Techniques","field":"Biochemistry, Genetics and Molecular Biology","cited_by":40,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"National Institute of Mental Health; Stanley Center for Psychiatric Research, Broad Institute; University of California, San Francisco; Simons Foundation Autism Research Initiative; Broad Institute; McGill University; Harvard University; Simons Foundation; H. Lundbeck A/S; Lundbeckfonden; U.S. Department of Health and Human Services","keywords":"Debiasing; Audit; Computer science; Process (computing); Data science; Machine learning; Artificial intelligence; Code (set theory); Psychology; Accounting; Cognitive science","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.1053681,0.001544998,0.002085967,0.003177166,0.003345397,0.01049506,0.005303515,0.004089185,0.003912234],"category_scores_gemma":[0.3838283,0.00224948,0.00214993,0.002285187,0.01096776,0.01345333,0.01035227,0.01294958,0.004194497],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002392698,"about_ca_system_score_gemma":0.01567863,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003016855,"about_ca_topic_score_gemma":0.00322824,"domain_scores_codex":[0.9034817,0.05340476,0.007818556,0.006753232,0.02628468,0.002257088],"domain_scores_gemma":[0.5858196,0.190649,0.03203177,0.1489974,0.0385417,0.003960487],"domain_codex":null,"domain_gemma":"methods","domain_candidate":"methods","domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.0009689868,0.0005563108,0.02680301,0.00200246,0.0004849485,0.0006951395,0.005136087,0.03824453,0.02048283,0.2375279,0.04230712,0.6247906],"study_design_scores_gemma":[0.0002652151,0.0005081934,0.006265227,0.002740738,0.0001814861,0.001278795,0.0009597963,0.2229842,0.06886337,0.6033212,0.09204603,0.000585833],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01120196,0.0005749289,0.967456,0.007265985,0.0004649992,0.0007823124,0.0003210804,0.008514899,0.00341783],"genre_scores_gemma":[0.1489946,0.0009189789,0.8398343,0.003237048,0.0004487536,0.001277131,0.0006005989,0.002313201,0.002375469],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8946319,"threshold_uncertainty_score":0.5572463,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01569015251213753,"score_gpt":0.3388596447208553,"score_spread":0.3231694922087178,"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."}}