{"id":"W4398766485","doi":"10.1038/s44277-024-00010-z","title":"Opportunities and risks of large language models in psychiatry","year":2024,"lang":"en","type":"article","venue":"NPP—Digital Psychiatry and Neuroscience","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":68,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"National Institute of Mental Health; Laureate Institute for Brain Research, University of Tulsa; National Institute of General Medical Sciences; National Institute on Drug Abuse; William K. Warren Foundation","keywords":"Psychology; Psychiatry","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":[0.09186461,0.001057759,0.001069177,0.002090239,0.002911511,0.01349642,0.004191648,0.006034369,0.004343329],"category_scores_gemma":[0.1615383,0.001496294,0.001770022,0.001452565,0.01842313,0.02193242,0.01713135,0.01043348,0.001529995],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.005120065,"about_ca_system_score_gemma":0.008947206,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005146672,"about_ca_topic_score_gemma":0.005527715,"domain_scores_codex":[0.9086271,0.07787403,0.00240066,0.002888602,0.007110055,0.001099645],"domain_scores_gemma":[0.7385103,0.2181893,0.005345642,0.02856411,0.007024951,0.002365788],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.0003105068,0.00009225882,0.003654035,0.0004675845,0.0001176972,0.0005057638,0.007277424,0.02049555,0.0008155183,0.8682519,0.01099544,0.08701636],"study_design_scores_gemma":[0.00003757753,0.00004877404,0.0003012096,0.0004463496,0.00003882535,0.0003406843,0.001444831,0.0395265,0.00100774,0.9098888,0.04686021,0.00005852661],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":"theoretical_or_conceptual","genre_codex":"methods","genre_gemma":"commentary","genre_scores_codex":[0.02063523,0.005368059,0.7077913,0.2412192,0.000872794,0.000267176,0.0003997158,0.002110081,0.02133649],"genre_scores_gemma":[0.5438039,0.004542455,0.4168532,0.02397989,0.001529876,0.001005315,0.0005507261,0.00107472,0.006659818],"genre_candidate":"commentary","genre_consensus":null,"teacher_disagreement_score":0.09186461,"threshold_uncertainty_score":0.4858323,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2507422762530556,"score_gpt":0.4461794069120039,"score_spread":0.1954371306589483,"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."}}