{"id":"W3204942516","doi":"10.1609/aiide.v17i1.18894","title":"Explaining Deep Reinforcement Learning Agents in the Atari Domain through a Surrogate Model","year":2021,"lang":"en","type":"article","venue":"Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment","topic":"Explainable Artificial Intelligence (XAI)","field":"Computer Science","cited_by":9,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Alberta","funders":"Natural Sciences and Engineering Research Council of Canada; Alberta Machine Intelligence Institute","keywords":"Reinforcement learning; Artificial intelligence; Computer science; Suite; Domain (mathematical analysis); Transformation (genetics); Replicate; Surrogate model; Representation (politics); Deep learning; Machine learning; Mathematics; Law","routes":{"ca_aff":true,"ca_fund":true,"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":[0.001148564,0.0006567251,0.0005251347,0.0002510331,0.0002920261,0.0008289441,0.001193265,0.00125328,0.002565252],"category_scores_gemma":[0.00573745,0.0003185266,0.0005328318,0.0001710813,0.001241227,0.001357879,0.001191396,0.002295447,0.0003065022],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0008917047,"about_ca_system_score_gemma":0.0008324276,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002570648,"about_ca_topic_score_gemma":0.00383435,"domain_scores_codex":[0.9995139,0.0002510794,0.00001889618,0.00008539401,0.00008390414,0.00004680379],"domain_scores_gemma":[0.9981694,0.001191142,0.0001864025,0.0002309826,0.0001382405,0.00008374814],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0000623612,0.00004501752,0.0008400641,0.00004542251,0.00002263949,0.0001055255,0.0001738292,0.9429148,0.00148061,0.03928208,0.0006313379,0.01439628],"study_design_scores_gemma":[0.00000660973,0.00001589151,0.0000498517,0.000004155701,0.00000182878,0.000006727137,0.000008227826,0.9879293,0.0002415986,0.01152398,0.0002092342,0.000002585945],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.09195109,0.0001355787,0.901287,0.0009023796,0.00004361856,0.00008680115,0.0001552953,0.0005826411,0.004855697],"genre_scores_gemma":[0.8779274,0.00008454482,0.1183904,0.0001607607,0.00001687664,0.0001413048,0.0001506698,0.0000713871,0.003056869],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.002570648,"threshold_uncertainty_score":0.008581638,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.07958754341074467,"score_gpt":0.3097414533509233,"score_spread":0.2301539099401787,"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."}}