遊戲理論之電腦模擬─對手選擇對於反覆囚犯困局之影響

Abstract

許多合作行為演化的研究領域上，諸如經濟學、社會學、國際關係等等，都可以見到非常多囚犯困局相關的研究。在Muti-Agent Systems的研究領域中，合作是一個主要的追求目標。近年來有越來越多的研究利用囚犯困局探討在Muti-Agent Systems環境中合作行為的浮現。 而在囚犯困局的研究當中，促進合作是一個相當重要的研究領域。在一般的囚犯困局當中，往往整體一開始會急速陷入背叛狀態，使得整體利益降低，在經過冗長的演化之後，合作行為才會逐漸緩慢的浮現，而在背叛到合作這段期間隱含著資源的浪費。越快地促進合作行為浮現一方面可以使得整體利益獲得最佳狀態，另一方面則可以避免無形中資源的消耗、浪費。 在促進合作的研究領域當中，有著許多不同的方法，而偏見的選擇對手機制為其中一種方法。在偏見的對手選擇模型當中又粗略分為「以標籤為基礎的偏見對手選擇」、「以空間為基礎的偏見對手選擇」以及「以親屬關係為基礎的偏見對手選擇」。在各種偏見的對手選擇模型當中有著不同的優缺點，所以本研究試圖提出一個更簡單、易實行並且具備成效的偏見對手選擇方法，來快速的促進反覆囚犯困局的合作行為浮現。

There are many fields, such like economic science, sociology and international relations, in which researchers are using the Prisoner’s Dilemma to study the evolution of cooperation. Cooperation is also one of the major objectives pursued by researchers in the area of multi-agent systems. There are more and more studies aimed at the Prisoner’s Dilemma to investigate the emergence of cooperation in multi-agent systems. Promoting cooperation is an important study in the Prisoner’s Dilemma. In a general Prisoner’s Dilemma game, the population usually falls down to the status of defection rapidly, and thus decreases the benefit of the whole population. Cooperation needs long time to be evolved and the process is slow. The gap between defection and cooperation implies the waste of resources. The more quickly cooperation emerges, the more benefit the population can enjoy. There are many methods to promote cooperation in the Prisoner’s Dilemma, and the biased opponent selection mechanism is one of them. There are three kinds of the biased opponent selection models: tag-based model, space-based model and kinship-based model. There are many advantages and disadvantages in these three models, and this research tries to provide another biased opponent selection method that is simple and effective to trigger the emergence of the cooperation quickly in the Iterated Prisoner’s Dilemma.