{"id":"W4408725130","doi":"10.1093/jamia/ocaf037","title":"Robust privacy amidst innovation with large language models through a critical assessment of the risks","year":2025,"lang":"en","type":"article","venue":"Journal of the American Medical Informatics Association","topic":"Machine Learning in Healthcare","field":"Computer Science","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Manitoba","funders":"National Center for Advancing Translational Sciences; National Human Genome Research Institute; U.S. National Library of Medicine; National Institute on Aging; Natural Sciences and Engineering Research Council of Canada; University of Texas Health Science Center at Houston; National Cancer Institute; National Science Foundation; National Institutes of Health; Cancer Prevention and Research Institute of Texas","keywords":"Health Insurance Portability and Accountability Act; Computer science; Software portability; Cosine similarity; Usability; Health care; Information retrieval; F1 score; Natural language processing; Artificial intelligence; Data extraction; Health records; Data quality; Similarity (geometry); Machine learning; Confidentiality; Data mining; Metric (unit); Computer security; MEDLINE; Pattern recognition (psychology); Human–computer interaction","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":[],"consensus_categories":[],"category_scores_codex":[0.002800921,0.00009632761,0.0003020011,0.0001106552,0.0001732572,0.00007557539,0.001279327,0.00006784553,0.00000502748],"category_scores_gemma":[0.004525058,0.00005049643,0.00009274744,0.001706588,0.000113041,0.0005659862,0.000381762,0.0009694337,4.304088e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0005285132,"about_ca_system_score_gemma":0.0009873136,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001067261,"about_ca_topic_score_gemma":0.00001225478,"domain_scores_codex":[0.9965706,0.0003755422,0.0009936119,0.00005974976,0.001782104,0.0002183373],"domain_scores_gemma":[0.9954337,0.0007108352,0.002631531,0.0003905609,0.000786505,0.00004683945],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00006427505,0.0006324927,0.3316248,0.0006140007,0.0005375969,0.000007322716,0.03037969,0.05805301,0.00007361338,0.5421027,0.00951068,0.02639985],"study_design_scores_gemma":[0.0005775688,0.0001699202,0.1065865,0.0004798692,0.00004521251,0.00001688497,0.001413015,0.8853242,0.00006952494,0.004742229,0.000485963,0.0000890838],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.265533,0.00001546971,0.7055573,0.02721975,0.0003501207,0.0001247017,0.000003360012,0.00001603438,0.001180295],"genre_scores_gemma":[0.9416922,0.00001791632,0.05316686,0.005014353,0.00005722885,0.000001954894,7.810067e-7,0.000004279475,0.00004447887],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8272712,"threshold_uncertainty_score":0.5417244,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03137265139190625,"score_gpt":0.3852976143721293,"score_spread":0.3539249629802231,"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."}}