Abstract: Suppose that n persons with their own behavioral characteristics live in a limited area. Some persons here are willing to be kind to others, while others choose to be selfish. Some are penny-pinching, while others may be weathercocks. When two persons meet, they will determine their behaviors according to their own behavior characteristics, and accordingly, obtain their own profits. What kinds of person will succeed in a long run?
To discuss the above problem, the agent-based simulation method is used to describe the interaction between persons in grass-roots organizations by using the repeated prisoner' dilemma model, where the different behavior characteristics of persons are modeled as strategies, and four simulation models with no manager, strongly supervised managers, weakly supervised managers, and no supervised managers are constructed, respectively.
In the no manager model, we design 12 strategies to depict the behavior of different persons and find that the “cunning men” get the highest profits. The reason may be that the persons of this strategy can not only ensure that they will not be retaliated by selfish persons but also benefit from friendly ones. It implies that no manager mechanism is not a suitable form for grass-roots organizations.
In the models with managers, managers are elected. Three sub-models are discussed below. In the model with strongly supervised managers, managers must always cooperate in the interaction with others. In this case, the “good men” with the always-cooperation-strategy, who are always cooperative with others, get the highest profits. It is may be that the election mechanism provides them with an important opportunity to be elected as managers to improve their profits. In the model with weakly supervised managers, some elected managers may choose to betray, while others still cooperate. Therefore, each strategy is changed into two strategies, for example, the always-cooperation-strategy is changed into two strategies: the really-always-cooperation-strategy, who is still cooperative after being elected as a manager, and the like-always-cooperation-strategy, who will betray after becoming a manager. The reason may be that although choosing betrayal can obtain high returns in a short period of time, the election mechanism makes it difficult for such agents in the future election, and as a result, the “good men” with the really-always-cooperation-strategy get the highest profits in this case. In the model with no supervised managers, some elected managers may not only choose to betray but also even encroach on the profits of others during their tenure. Although it is likely to lose the future election, elected managers still may have accumulated a lot of wealth by way of encroachment, which may be why the “hypocrites” with the like-always-cooperation-strategy get the highest profits in this case.
A good mechanism should be able to protect the interests of “good persons”. Therefore, from the perspective of individuals, the best mechanism is with strongly supervised managers, the second is with weekly supervised managers, the third is with no managers and the last is with no supervised managers. Furthermore, from the perspective of organizations, we compare the average profits of all persons under different mechanisms, and also find that the best mechanism is with strongly supervised managers, the second is with weekly supervised managers, the third is with no managers and the last is with no supervised managers. Whether from the perspective of individuals or organizations, the mechanism with supervised managers, especially with strongly supervised managers, is the most efficient organizational form since it can protect the interests of “good persons” and thus lead people to being “good”.

Key words: interpersonal interactions, grassroots organization governance, iterated prisoner's dilemma, agent-based modeling and simulation

摘要： 运用基于Agent仿真的方法,用重复囚徒困境模型来刻画基层组织中个体间的交互行为,并把个体的不同行为特点抽象为博弈策略,从而构建出人际交互仿真模型。然后,对比分析无管理者、管理者受强监督、管理者受弱监督、以及管理者不受监督机制下各种博弈策略的平均收益和全部个体的总平均收益。结果表明,四种机制的优劣顺序为:管理者受强监督机制>管理者受弱监督机制>无管理者机制>管理者不受监督机制。管理者受监督的机制才是最有效率、最能保障“好人”的利益从而能够“引人向善”的组织形式。

关键词: 人际交互, 基层组织治理, 重复囚徒困境, Agent建模与仿真