{"id":"W4385572053","doi":"10.18653/v1/2023.findings-acl.369","title":"Shielded Representations: Protecting Sensitive Attributes Through Iterative Gradient-Based Projection","year":2023,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":7,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Azrieli Foundation; Israel Science Foundation; Open Philanthropy Project","keywords":"Computer science; Projection (relational algebra); ENCODE; Task (project management); Representation (politics); Artificial intelligence; Machine learning; Iterative method; Pattern recognition (psychology); Algorithm","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":[],"consensus_categories":[],"category_scores_codex":[0.003022176,0.001966996,0.0014566,0.0008262218,0.0007126348,0.001589225,0.002085815,0.001721015,0.001710048],"category_scores_gemma":[0.01289079,0.0006676246,0.001145796,0.0008396679,0.001793514,0.003514254,0.00347072,0.003581368,0.001057599],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000793317,"about_ca_system_score_gemma":0.001723839,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002904068,"about_ca_topic_score_gemma":0.003061656,"domain_scores_codex":[0.9984451,0.0006351864,0.00006618655,0.0002672841,0.000432299,0.0001539802],"domain_scores_gemma":[0.9962888,0.001637407,0.0003364508,0.0009487465,0.0006134366,0.0001751954],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0004269434,0.0004004867,0.003096016,0.0002040819,0.0002097989,0.000202281,0.0008406566,0.3669367,0.02267436,0.03207349,0.008285213,0.56465],"study_design_scores_gemma":[0.00002213729,0.00008669506,0.0002211981,0.00001375202,0.00001981812,0.00005029732,0.0000436242,0.9687006,0.005970068,0.02412345,0.0007313333,0.0000170476],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.03038566,0.0001941402,0.966742,0.0003105237,0.00003757107,0.00007068438,0.00006481788,0.001187814,0.001006895],"genre_scores_gemma":[0.5888548,0.0003445837,0.4047509,0.000550898,0.000116608,0.0003361623,0.0005995988,0.0004831941,0.003963179],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.003022176,"threshold_uncertainty_score":0.01598299,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.07561534063835071,"score_gpt":0.3180728255231723,"score_spread":0.2424574848848216,"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."}}