{"id":"W7130684127","doi":"10.1109/swc65939.2025.00306","title":"Can LLMs Aid Expert Elicitation for Causal Modeling?","year":2025,"lang":"","type":"article","venue":"","topic":"Bayesian Modeling and Causal Inference","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Calgary","funders":"","keywords":"Expert elicitation; Bayesian network; Robustness (evolution); Causal model; Bayesian probability; Entropy (arrow of time); Bayesian statistics; Causal inference","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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0005715609,0.0004292499,0.000424427,0.0002961624,0.000539424,0.0006799402,0.001147951,0.0003007318,0.00004277418],"category_scores_gemma":[0.0001611825,0.0004291549,0.0002187819,0.0007286235,0.00008333162,0.0004764475,0.0002815118,0.0002729911,0.00003249825],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001608663,"about_ca_system_score_gemma":0.0009771802,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001245452,"about_ca_topic_score_gemma":0.0002604406,"domain_scores_codex":[0.9968091,0.00009795235,0.0007955498,0.00112368,0.0003648941,0.0008088034],"domain_scores_gemma":[0.9978325,0.0002022728,0.000111468,0.0008983883,0.0007321114,0.0002233067],"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":[0.00006448518,0.0002034744,0.00003251045,0.0001083976,0.00009343059,0.000002717413,0.003288274,0.0420187,0.001072533,0.6884014,0.008006705,0.2567074],"study_design_scores_gemma":[0.0007112466,0.0001873036,0.000005457673,0.0001876808,0.0000287012,0.000002217318,0.0002215299,0.9029678,0.001606729,0.09232917,0.001321937,0.0004302997],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.002844977,0.0009929651,0.9715362,0.01533432,0.002300392,0.000602165,0.00001526081,0.0002519903,0.006121702],"genre_scores_gemma":[0.8704304,0.0001918383,0.1170861,0.006074514,0.0002077967,0.0001528682,0.00001233418,0.00002102997,0.005823134],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.8675854,"threshold_uncertainty_score":0.999816,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04432170127200308,"score_gpt":0.3183801185490867,"score_spread":0.2740584172770836,"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."}}