{"id":"W3021228017","doi":"10.1101/2020.05.08.085183","title":"Systematic auditing is essential to debiasing machine learning in biology","year":2020,"lang":"en","type":"preprint","venue":"bioRxiv (Cold Spring Harbor Laboratory)","topic":"vaccines and immunoinformatics approaches","field":"Biochemistry, Genetics and Molecular Biology","cited_by":5,"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; H. Lundbeck A/S; Lundbeckfonden; Simons Foundation Autism Research Initiative; Broad Institute; McGill University; Harvard University; Simons Foundation","keywords":"Debiasing; Generalizability theory; Audit; Computer science; Artificial intelligence; Process (computing); Machine learning; Data science; Psychology; Cognitive science; Accounting","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":[0.06054292,0.001031989,0.001730905,0.002116306,0.002116692,0.007183766,0.003307258,0.00298837,0.001826781],"category_scores_gemma":[0.2821903,0.001317277,0.001043192,0.001844232,0.01095652,0.01114826,0.009491324,0.009734073,0.001029316],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002019638,"about_ca_system_score_gemma":0.009060177,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001692655,"about_ca_topic_score_gemma":0.0014593,"domain_scores_codex":[0.9545107,0.02752679,0.003484678,0.004084192,0.009008325,0.001385211],"domain_scores_gemma":[0.7147004,0.1473854,0.03204814,0.08466624,0.01780572,0.003394059],"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":[0.001044301,0.0005311494,0.07479684,0.001091505,0.0005907494,0.0005625826,0.002702564,0.121441,0.02794527,0.27455,0.01777908,0.4769649],"study_design_scores_gemma":[0.0000694148,0.0001919494,0.00427375,0.0004583335,0.00008055515,0.0002860862,0.0003156628,0.3162473,0.03195371,0.6365477,0.009416512,0.0001590136],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.07902718,0.001260814,0.8958286,0.01564906,0.0005847131,0.0002208485,0.0003190684,0.003827716,0.00328204],"genre_scores_gemma":[0.7042044,0.0008001943,0.2886517,0.003309712,0.0004887498,0.0003267204,0.000342562,0.0006667305,0.001209205],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.9394571,"threshold_uncertainty_score":0.3201854,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01227245685710764,"score_gpt":0.2317291120219246,"score_spread":0.219456655164817,"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."}}