@article{oai:ipsj.ixsq.nii.ac.jp:00217723,
 author = {西野上, 和真 and 五十嵐, 瞭平 and 岩崎, 敦 and Kazuma, Nishinoue and Ryohei, Igarashi and Atsushi, Iwasaki},
 issue = {4},
 journal = {情報処理学会論文誌},
 month = {Apr},
 note = {本論文は私的観測下の繰り返し囚人のジレンマにおける協力のダイナミクスを分析した．私的観測は，プレイヤが相手の行動についてノイズを含むシグナルを観測し，そのシグナルを他のプレイヤは観測できないという特徴を持つ．ここで，どんな戦略の組が均衡になるかはゲーム理論の有名な未解決問題の1つであり，本論文では戦略空間を状態数2以下の有限状態機械に限定したレプリケータダイナミクスの帰結から，どのような戦略が生き残るかを吟味した．その結果，利得構造に応じて，4つの社会（非協力，不寛容，相互協力，周期協力）が現れることが分かった．とくに周期協力社会で，非専門家の間では有効と信じられてきたしっぺ返し戦略は最大多数になりうるが，他の戦略と共存しなければならないという不安定さを持つ．一方，他の社会では，ある特定の戦略が人口のほぼすべてを占めるようになる．さらにノイズと突然変異率に関する感度分析から十分広いパラメータにおいて同じ傾向を保つことが分かった．, This paper analyzes the dynamics of cooperation in a repeated prisoner's dilemma under private monitoring, where each player privately observes noisy signals about the opponent's actions. What kind of strategy forms an equilibrium is one of the open, but fundamental questions in game theory. We examine what kind of strategies are abundant in the consequences of replicator-mutator dynamics, where the strategy space is restricted to a finite state automata within two states. As a result, we found that four kinds of societies (non-cooperation, intolerant cooperation, mutual cooperation, and cyclical cooperation) emerge according to payoff structures. In particular, in the cyclical cooperation society, the well-known Tit-for-tat strategy, which is believed to be successful for non-experts, can become the most abundant. However, it is not so stable because it must coexist with other strategies. In the other societies, a particular strategy dominates the population. Furthermore, our sensitivity analysis on the parameters of signal distributions and mutation rate shows that the trend persists for a wide range of parameters.},
 pages = {1138--1148},
 title = {私的観測下の繰り返し囚人のジレンマにおける協力のダイナミクス},
 volume = {63},
 year = {2022}
}