{"id":"W2965865809","doi":"10.1093/jamia/ocz114","title":"The machine giveth and the machine taketh away: a parrot attack on clinical text deidentified with hiding in plain sight","year":2019,"lang":"en","type":"article","venue":"Journal of the American Medical Informatics Association","topic":"Electronic Health Records Systems","field":"Health Professions","cited_by":12,"is_retracted":false,"has_abstract":true,"ca_institutions":"Privacy Analytics (Canada)","funders":"U.S. National Library of Medicine; National Human Genome Research Institute","keywords":"Computer science; Identifier; Sight; Artificial intelligence; Masking (illustration); Natural language processing; Random forest; Computer security; Information retrieval","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":["metaresearch","research_integrity"],"consensus_categories":[],"category_scores_codex":[0.0295206,0.0001759391,0.0007999954,0.00009531599,0.0006646714,0.00004911571,0.0006234198,0.0001860785,0.00004134867],"category_scores_gemma":[0.007715359,0.00006949947,0.0001444136,0.0004730808,0.0002648734,0.0001675194,0.0001376503,0.003850796,0.00004638319],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001050128,"about_ca_system_score_gemma":0.001330162,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0005368346,"about_ca_topic_score_gemma":0.001419402,"domain_scores_codex":[0.9904752,0.004352435,0.002728356,0.00009577435,0.001735498,0.0006127847],"domain_scores_gemma":[0.9769118,0.01547413,0.006770175,0.0003457183,0.0002542416,0.0002439233],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.002336577,0.000139236,0.9338898,0.0002484932,0.0004325719,0.000005690047,0.01284745,0.0001516487,0.000003821083,0.001341459,0.01874575,0.02985748],"study_design_scores_gemma":[0.03709941,0.004488293,0.4953732,0.005833671,0.0003462337,0.0001577741,0.036539,0.2670804,0.00001050603,0.001364096,0.1509375,0.0007699132],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9456627,0.0002168905,0.0003923345,0.04954181,0.001235259,0.001217642,0.00000665379,0.00001251602,0.001714124],"genre_scores_gemma":[0.9879041,0.001182816,0.0001569207,0.009299048,0.0004213346,0.00002695946,0.000002363339,0.00002175598,0.0009846848],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.4385166,"threshold_uncertainty_score":0.9993128,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03774819822446671,"score_gpt":0.4296206950065799,"score_spread":0.3918724967821132,"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."}}