A trust network group consensus decision-making method considering the interference effect of multi-source reference path information is proposed to address the complex issue of expert opinion correction and moderator compensation costs in the trust network group decision-making environment. Firstly, in response to the unwillingness of non-consensus experts to adjust their opinions during the process of opinion revision, a non-consensus expert adjustment willingness recognition model is established, considering the relative out-degree centrality and confidence level. Secondly, considering the willingness of non-consensus experts to accept the opinions of trusted experts and the opinions of moderators, a feedback mechanism considering quantum interference in a probabilistic linguistic environment is established by combining multi-source reference path information in the form of a wave function to investigate the interference effect in the brain of non-consensus experts who are willing to adjust. Based on this, a minimum opinion adjustment cost consensus model is established by further considering the constraints of minimizing moderator compensation costs and improving individual consensus. Then, using a trust penalty mechanism to reduce the impact of non-consensus experts who are unwilling to adjust their opinions on group consensus reaching, while considering the minimization of trust adjustment costs and group consensus reaching constraints, a minimum trust adjustment cost consensus model is developed. Finally, the proposed method is applied to emergency decision-making for unexpected events, verifying its feasibility.