{"id":"W4400047099","doi":"10.1093/jla/laae003","title":"Large Legal Fictions: Profiling Legal Hallucinations in Large Language Models","year":2024,"lang":"en","type":"article","venue":"The Journal of Legal Analysis","topic":"Artificial Intelligence in Law","field":"Social Sciences","cited_by":192,"is_retracted":false,"has_abstract":true,"ca_institutions":"Institute on Governance","funders":"","keywords":"Typology; Hallucinating; Profiling (computer programming); Psychology; Criminology; Law; Political science; Social psychology; Sociology","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.006160736,0.0005099336,0.0004573009,0.003314054,0.0009417268,0.002902313,0.000780736,0.001222437,0.001595095],"category_scores_gemma":[0.07634547,0.0002964157,0.0004772138,0.002170414,0.001208078,0.004495707,0.002711077,0.001705502,0.0006810449],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001110561,"about_ca_system_score_gemma":0.0006987309,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005720588,"about_ca_topic_score_gemma":0.008958582,"domain_scores_codex":[0.9937993,0.004019187,0.0003996601,0.0005041339,0.001065145,0.0002125192],"domain_scores_gemma":[0.9001827,0.08717176,0.004238489,0.004463738,0.002963487,0.0009798174],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.001655423,0.0008934777,0.4490322,0.001154672,0.0005876431,0.003713555,0.02678995,0.1141233,0.008612758,0.01283061,0.03824105,0.3423654],"study_design_scores_gemma":[0.00004701615,0.0002231338,0.05448651,0.0001639974,0.0001204039,0.001092313,0.007161997,0.8920313,0.004331624,0.02660563,0.01361806,0.000118008],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9451967,0.0007804605,0.04349648,0.002355616,0.00006546461,0.0001624535,0.002544133,0.001411044,0.003987571],"genre_scores_gemma":[0.9877161,0.00009418364,0.009300101,0.0001680543,0.00003978254,0.00003908141,0.002149042,0.00007100153,0.0004225408],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.006160736,"threshold_uncertainty_score":0.03258145,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03341118742895829,"score_gpt":0.3685634204669446,"score_spread":0.3351522330379862,"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."}}