{"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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0006203149,0.0005425228,0.0007813571,0.0002247481,0.0001331397,0.0001840605,0.0006172,0.0005546014,0.00001360325],"category_scores_gemma":[0.000791997,0.0005642684,0.0001961083,0.0003433529,0.00003516364,0.00001059149,0.001219861,0.0007046359,0.0000496524],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000094085,"about_ca_system_score_gemma":0.0002760899,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00007991782,"about_ca_topic_score_gemma":0.000006339573,"domain_scores_codex":[0.9972976,0.000195523,0.0009255389,0.0008443326,0.0001745811,0.0005623673],"domain_scores_gemma":[0.9982354,0.00002913103,0.0005374862,0.0008088462,0.0001850124,0.0002041196],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"bench_or_experimental","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.0000404435,0.00004729809,0.004037067,0.006122673,0.0002172548,0.00001003016,0.00005867646,0.0004307786,0.9888308,0.00009448741,0.000108534,0.000001946664],"study_design_scores_gemma":[0.001389117,0.0003790268,0.008581898,0.00613046,0.0002607291,2.192702e-7,0.00006528373,0.02049483,0.9579191,0.000005711417,0.002567863,0.00220576],"study_design_candidate":"bench_or_experimental","study_design_consensus":"bench_or_experimental","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9864796,0.002774282,0.008363856,0.0005166144,0.0005246022,0.001107573,0.0001025258,0.0001128824,0.00001801625],"genre_scores_gemma":[0.9934473,0.0002250678,0.00504798,0.000581886,0.000399287,0.0001731556,0.000005166123,0.0001130539,0.000007099898],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.03091171,"threshold_uncertainty_score":0.9996809,"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."}}