社交网络的本质，在于能够影响人们的思想，并迅速形成群体行为。而负面群体行为会给社会稳定带来巨大的影响。社交网络中现有群体行为引导方法存在简单低效，容易造成参与者的反向认知、情绪累加和政治冷漠等问题。为了有效自动引导社交网络中的群体行为，本研究首先引入信任智能体，并根据群体行为特征及相关背景设计信任智能体的特征，从而使其被社交网络中的其他参与者信任。然后，根据社交网络中网络结构、参与者个体行为和群体行为互动构成的离散动力系统的非线性动态性质，本研究引入一种动态协调方法，来协调社交网络中参与者的行为，以达到引导参与者尽可能少地参与特定监控的负面群体行为。最后，我们提出的离散动力系统将由一个社交网络群体行为引导原型系统来进行实验验证，它将有效控制社交网络中的负面群体行为，特别是在不同的社会刺激的条件下（例如，新公共策略或群体事件）利用现代计算方法有效自动引导社交网络用户远离负面群体行为。

The essence of social networks is that they can influence people’s public opinions, and then form the group behaviors quickly. However, the negative group behavior influences the stability of society significantly. But the existing group behavior induction approaches are too simple and inefficient, which easily makes participants become reverse recognization, feeling aggregation, political negligence and so on. Hence, in order to automatically induct the group behavior in social networks efficiently, this project introduces trust agent, and then designs the features of trust agents according to group behavior features and related social background features. In addition, with the nonlinear properties of a discrete dynamic system formed by network structure, the personal behaviors and group behaviors of participants in social networks, the dynamics control mechanism can be generated to coordinate the behaviors of participants in social networks to avoid the specific monitored negative group behavior as few as possible. Finally, our proposed discrete dynamic system will be experimentally validated by a social network group behavior induction prototype system, and it effectively controls the negative group behaviors in social networks, more specifically under the different social incentives (e.g., a new public strategy or a group event) the system can effectively automatically induct social network users with modern computational approaches to avoid negative group behaviors.