本文基于群体最后通牒博弈模型，将信息反馈作为实验变量设计实验，其中，从第二轮起，对照组返回被试上一轮次的策略及收益；返回对手信息组在对照组的基础上返回所有与被试角色不同的群体的策略；返回同组信息组在对照组的基础上返回所有与被试角色相同的群体的策略及对应收益。 结论是返回同组信息对提高分配的公平度和效率有较强促进作用，返回对手信息组次之，对照组公平度和效率最差。返回信息对“搭便车”行为具有明显抑制作用，且两种信息返回方式抑制效果相近。本文对产生这种现象的原因从被试类型及学习模型角度进行分析，并给出猜想，使用实验数据进行了初步验证。对照组的被试行为可用强化学习模型进行解释；返回对手信息组被试行为可同时用信念学习模型与建立声誉同时解释；返回同组信息的被试行为可用模仿学习解释。

We designed experiments based on group ultimatum game model and used information feedback on every round as the experiments variable. The control groups return their own strategies from the second rounds and payoff on last round to every agent; the groups which return rival information return the strategies of the group that is different from agents' role and their own payoff to every agent; the experiments which returns the same groups' information return the strategies and payoff of the their own group with them and their own payoff to every agent. Our results show that returning the same group's information plays a strong role in improving the fairness and efficiency of distribution, returning the opponents' information comes the second, and control groups is the worst. We also find return group information could restrain the free-rider behavior. We analyzed the causes of this phenomenon and gives some guess and explanation viewed from the type of agents and the learning models, and use the experiment data to support our explanation. The behaviors of the control groups can be explained reinforcement learning, while agents' strategies of the groups with rival information could be explained by the belief learning and reputation-building. And the strategies of agents who have the same groups' information can be explained by imitation learning.