{"id":"W4387394671","doi":"10.1609/aiide.v19i1.27499","title":"Entropy as a Measure of Puzzle Difficulty","year":2023,"lang":"en","type":"article","venue":"Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment","topic":"Artificial Intelligence in Games","field":"Computer Science","cited_by":6,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Alberta","funders":"Canadian Institute for Advanced Research","keywords":"Computer science; Exploit; Ranking (information retrieval); Witness; Entropy (arrow of time); Variety (cybernetics); Set (abstract data type); Artificial intelligence; Machine learning; Theoretical computer science","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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0002818915,0.0003063159,0.0003749812,0.0002357223,0.0001570151,0.0004074185,0.0014711,0.00008304218,0.00004952318],"category_scores_gemma":[0.0007274515,0.0002202357,0.0002159895,0.000722289,0.0004233809,0.001148151,0.000793953,0.0002691887,0.0001928779],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00007350332,"about_ca_system_score_gemma":0.00005546034,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00004005831,"about_ca_topic_score_gemma":0.00000517636,"domain_scores_codex":[0.9974844,0.00001996984,0.0007569676,0.0005872491,0.0007372943,0.0004141209],"domain_scores_gemma":[0.9981575,0.0002090875,0.0005115475,0.0003187179,0.0006769605,0.0001261859],"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":[0.0002378148,0.0004456631,0.0007985721,0.00004922392,0.00008607773,0.000002432019,0.005934571,0.00004624202,0.06139208,0.7724273,0.0002595438,0.1583205],"study_design_scores_gemma":[0.00003988749,0.0006507728,0.0003735259,0.0004823021,0.00001413151,0.000008916765,0.006422929,0.02244488,0.859367,0.1095273,0.0003906627,0.0002776982],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9660059,0.00002304033,0.007376653,0.004303342,0.0006296573,0.0007978357,0.00003241958,0.0001574083,0.02067371],"genre_scores_gemma":[0.9990717,0.00006036253,0.0001080164,0.0001244553,0.0000369335,0.00004270102,0.000001988014,0.00001492454,0.0005389463],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.7979749,"threshold_uncertainty_score":0.8980953,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04936860510762716,"score_gpt":0.2936480159845815,"score_spread":0.2442794108769543,"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."}}