{"id":"W3132556242","doi":"10.71781/10609","title":"Benchmarking bias mitigation algorithms in representation learning through fairness metrics","year":2022,"lang":"en","type":"article","venue":"Papyrus : Institutional Repository (Université de Montréal)","topic":"Ethics and Social Impacts of AI","field":"Social Sciences","cited_by":10,"is_retracted":false,"has_abstract":true,"ca_institutions":"Université de Montréal","funders":"McGill University; École Polytechnique Fédérale de Lausanne; Microsoft Research","keywords":"Benchmarking; Computer science; Debiasing; Machine learning; Benchmark (surveying); Artificial intelligence; Deep learning; Artificial neural network; Representation (politics); Deep neural networks; Feature learning; Data mining; Algorithm","routes":{"ca_aff":true,"ca_fund":true,"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":["sts"],"consensus_categories":[],"category_scores_codex":[0.001129192,0.0001102888,0.0001521636,0.0002386184,0.01522662,0.0001026015,0.0002534553,0.0001239332,0.00007163364],"category_scores_gemma":[0.0005421659,0.0001460653,0.0001069648,0.001177068,0.000254128,0.0008346016,0.0002016251,0.0005286563,0.000004374253],"about_ca_system_candidate":true,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.005738907,"about_ca_system_score_gemma":0.001075781,"about_ca_topic_candidate":true,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.2301173,"about_ca_topic_score_gemma":0.01371408,"domain_scores_codex":[0.9975452,0.0006571784,0.0002170911,0.000293167,0.0009697628,0.0003175293],"domain_scores_gemma":[0.9990381,0.0003288343,0.0002056535,0.0001144549,0.0001939059,0.0001190753],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"qualitative","study_design_gemma":"qualitative","study_design_scores_codex":[0.0001163781,0.0001917424,0.1671102,0.00001581907,0.00007321808,0.0008412137,0.5509937,0.03846283,0.0009309533,0.2267527,0.0004070475,0.01410423],"study_design_scores_gemma":[0.001425453,0.0001801883,0.06965621,0.00005481896,0.00007613235,0.00007050786,0.738588,0.007360824,0.0007797268,0.01957046,0.1615888,0.0006488874],"study_design_candidate":"qualitative","study_design_consensus":"qualitative","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8711572,0.001794418,0.0009566663,0.003399794,0.0009113238,0.0003001374,0.000009976261,0.0001184305,0.1213521],"genre_scores_gemma":[0.9928266,0.0003154844,0.0007224524,0.0001704883,0.0001750508,0.00001573869,0.00003979185,0.000009078833,0.005725318],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.2164032,"threshold_uncertainty_score":0.9980779,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03166011230794143,"score_gpt":0.2630295987903246,"score_spread":0.2313694864823832,"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."}}