{"id":"W2962059918","doi":"10.1145/3306618.3314244","title":"Actionable Auditing","year":2019,"lang":"en","type":"article","venue":"","topic":"Ethics and Social Impacts of AI","field":"Social Sciences","cited_by":586,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Audit; Computer science; Transparency (behavior); Accounting; Benchmark (surveying); Kairos; IBM; Business; Computer security","routes":{"ca_aff":true,"ca_fund":false,"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":["insufficient_payload"],"consensus_categories":["insufficient_payload"],"category_scores_codex":[0.0005045254,0.00002187852,0.00003792809,0.00001284587,0.0003117678,0.0000916616,0.00006794491,0.00005731017,0.00438707],"category_scores_gemma":[0.0002730209,0.00002064587,0.00002295689,0.0000772861,0.00004369442,0.000300324,0.00001097454,0.00007932461,0.0009368742],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00003680064,"about_ca_system_score_gemma":0.0001234576,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003922982,"about_ca_topic_score_gemma":0.0009798596,"domain_scores_codex":[0.999547,0.00003051438,0.00004052714,0.00005680604,0.0001782424,0.0001469318],"domain_scores_gemma":[0.9996886,0.0001096084,0.00001897094,0.00004467853,0.00008470876,0.00005348442],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"not_applicable","study_design_scores_codex":[6.681412e-7,0.000007222002,0.003284609,0.000001667479,0.000003558851,2.472467e-7,0.004617983,0.000002784072,0.00019464,0.9838145,0.006484396,0.001587713],"study_design_scores_gemma":[0.00009603646,0.00001548404,0.002839229,0.000006257936,0.00000191161,8.130905e-8,0.01935218,0.0000171915,0.0001549066,0.07100801,0.9064087,0.0001000041],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"other","genre_gemma":"empirical","genre_scores_codex":[0.1460095,0.000005389658,0.00002011923,0.01155744,0.0002789455,0.00004873779,2.18326e-7,0.000052952,0.8420267],"genre_scores_gemma":[0.8839878,0.00002221925,0.0001913771,0.001118148,0.0001958173,6.007153e-7,3.146013e-7,0.000002387344,0.1144813],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9128065,"threshold_uncertainty_score":0.999841,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03074059557481843,"score_gpt":0.3728602563927866,"score_spread":0.3421196608179681,"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."}}