In view of the possible non-cooperative behavior in large group emergency decision-making, considering the social network relationship of decision-making experts, a consensus-based method for managing non-cooperative behavior in large group emergency decision-making is proposed. Firstly, the Louvain clustering method based on modularity in social network is introduced to cluster decision experts to reduce the complexity of large group emergency decision-making. Secondly, two kinds of non-cooperative behaviors are defined, the identification and detection model of non-cooperative behaviors is established. Then, the concepts of trust risk coefficient, preference risk coefficient and comprehensive risk coefficient are defined. These risk coefficients are used to manage the aggregation of non-cooperative behaviors in different degrees, so that a decision-making scheme with high consensus level can be obtained. Finally, the feasibility and effectiveness of the method proposed in this paper are verified by the case of the 4.25 Tibet earthquake.