{"id":"W4415198176","doi":"10.48550/arxiv.2505.09662","title":"When Large Language Models are More PersuasiveThan Incentivized Humans, and Why","year":2025,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Misinformation and Its Impacts","field":"Social Sciences","cited_by":6,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Templeton World Charity Foundation; Ministerio de Ciencia, Innovación y Universidades; Good Ventures Foundation; Social Sciences and Humanities Research Council of Canada; Deutsche Forschungsgemeinschaft; Open Philanthropy Project; National Science Foundation","keywords":"Persuasion; Conviction; Incentive; Human language; Converse","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.008353909,0.0007478095,0.0004483994,0.0005398214,0.0004469531,0.003530967,0.0004585231,0.00167053,0.005567361],"category_scores_gemma":[0.05947968,0.0003656285,0.000331283,0.0002031579,0.001233783,0.003873955,0.001368251,0.001560788,0.001259446],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0004811998,"about_ca_system_score_gemma":0.000477104,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0005386059,"about_ca_topic_score_gemma":0.0008755386,"domain_scores_codex":[0.9916942,0.005549634,0.0002686088,0.0009999286,0.001160093,0.0003275711],"domain_scores_gemma":[0.9600451,0.02932048,0.004229424,0.003650689,0.001615637,0.001138688],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.006971194,0.00219964,0.1440461,0.004586966,0.001876137,0.001206588,0.01824114,0.03148482,0.197783,0.1075412,0.02466395,0.4593993],"study_design_scores_gemma":[0.002283694,0.008061804,0.1444633,0.001123546,0.002069517,0.002733993,0.01063999,0.2359146,0.1317364,0.3115276,0.1486727,0.0007728556],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9105027,0.00178421,0.0456465,0.008732542,0.0004038707,0.0001733007,0.0002506055,0.001292859,0.03121355],"genre_scores_gemma":[0.9845204,0.0002002534,0.01142005,0.0009930972,0.00006882662,0.00006252741,0.0001104817,0.0001234458,0.002500957],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.008353909,"threshold_uncertainty_score":0.04418021,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.07936080907955605,"score_gpt":0.2482118106493216,"score_spread":0.1688510015697655,"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."}}