{"id":"W4416014008","doi":"10.1609/aiide.v21i1.36855","title":"Game-Based Platforms for Studying Virtual Agent Believability","year":2025,"lang":"","type":"article","venue":"Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment","topic":"Social Robot Interaction and HRI","field":"Psychology","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Simon Fraser University","funders":"","keywords":"Virtual agent; Perception; Relevance (law); Presentation (obstetrics); Virtual machine; Reliability (semiconductor); Virtual reality; Adaptation (eye)","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":[0.0005352843,0.0006608224,0.0007814584,0.0003465655,0.0004345116,0.0008404643,0.0009025933,0.0002677966,0.0004905565],"category_scores_gemma":[0.0008573302,0.0005254875,0.0006344151,0.0004291824,0.0008007081,0.0006345993,0.0004146674,0.0007153685,0.00008212554],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0004839906,"about_ca_system_score_gemma":0.0002013132,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00005401971,"about_ca_topic_score_gemma":0.00001617654,"domain_scores_codex":[0.9961083,0.00003075527,0.001526363,0.001111818,0.0005285647,0.0006942109],"domain_scores_gemma":[0.996649,0.0008412402,0.0009407749,0.0003395853,0.001049492,0.0001798973],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.01091424,0.00517077,0.004147321,0.0003891253,0.000940101,9.603016e-7,0.01352191,0.00008160727,0.003647204,0.5720831,0.0008679517,0.3882357],"study_design_scores_gemma":[0.002300897,0.01327942,0.005006642,0.00980946,0.0008311582,0.000009064303,0.2641006,0.04411632,0.4285168,0.2129639,0.01664728,0.002418435],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9098036,0.00009907543,0.01393189,0.005966421,0.005640055,0.004404083,0.0003160799,0.00007230819,0.05976652],"genre_scores_gemma":[0.9944569,0.00005412316,0.00002941313,0.001420087,0.0001250792,0.0003489572,0.00001260541,0.00003456932,0.003518237],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.4248696,"threshold_uncertainty_score":0.9997197,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.09585965332053883,"score_gpt":0.3712859447033267,"score_spread":0.2754262913827878,"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."}}