{"id":"W4411272807","doi":"10.1073/pnas.2501660122","title":"Take caution in using LLMs as human surrogates","year":2025,"lang":"en","type":"article","venue":"Proceedings of the National Academy of Sciences","topic":"Biomedical Text Mining and Ontologies","field":"Biochemistry, Genetics and Molecular Biology","cited_by":29,"is_retracted":false,"has_abstract":true,"ca_institutions":"Quest University Canada","funders":"","keywords":"Psychology; Political science","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":[],"consensus_categories":[],"category_scores_codex":[0.000614454,0.00004513139,0.00007103015,0.00008742673,0.00008980188,0.000007548667,0.0003235121,0.00009701953,0.000002521256],"category_scores_gemma":[0.0004803892,0.0000310604,0.00003309107,0.0003603931,0.0006480744,0.000006426168,0.0001030281,0.00006449431,1.523183e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00001125708,"about_ca_system_score_gemma":0.00003514219,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00002007718,"about_ca_topic_score_gemma":4.887308e-7,"domain_scores_codex":[0.9993585,0.000004475297,0.0001610878,0.0001424844,0.0002484807,0.00008492415],"domain_scores_gemma":[0.9997754,0.00001678446,0.0001134839,0.000004829386,0.00007953399,0.00001003896],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"bench_or_experimental","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.000005275558,0.00002279033,0.01562732,0.00002024215,0.000005465783,1.622972e-9,0.00002504856,0.00002461223,0.9737945,0.009843667,0.0002200034,0.0004110308],"study_design_scores_gemma":[0.0001358442,0.00004046008,0.08400393,0.0000873953,0.000004179337,0.000002040157,0.0001923214,0.0002349232,0.8861831,0.02852741,0.0005451454,0.00004323002],"study_design_candidate":"bench_or_experimental","study_design_consensus":"bench_or_experimental","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9940125,0.0001957208,0.000001665972,0.001258006,0.0000144965,0.00004720626,0.000002263036,0.000002775749,0.004465417],"genre_scores_gemma":[0.9989296,0.00001408079,0.0006889364,0.0001433489,0.00002354495,0.000002126787,2.016277e-7,9.939881e-7,0.0001971251],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.08761141,"threshold_uncertainty_score":0.2387857,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05139893177855222,"score_gpt":0.3707658822394894,"score_spread":0.3193669504609372,"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."}}