<p>The rapid development of social media platforms has fundamentally reshaped large-scale group decision-making (LSGDM), introducing complex information flows that extend beyond traditional trust-based relationships. Existing consensus reaching processes (CRPs) primarily rely on trust networks, which emphasize interpersonal ties but often fail to capture the broader, real-world diffusion of information. To address this challenge, we propose a novel consensus model for LSGDM based on information diffusion theory in social networks. First, experts are clustered into subgroups using the Infomap algorithm, which reflects the underlying structure of information flow. Second, the independent cascade (IC) model is applied to simulate the information diffusion process, thereby quantifying expert influence and determining subgroup and individual weights. Third, a hybrid feedback mechanism is developed to guide consensus formation, where experts are categorized into leaders and followers with distinct interactive and automatic adjustment strategies. Finally, a hierarchical framework is introduced to manage noncooperative behavior among leaders. This framework balances the preservation of minority opinions with group consensus by protecting the weights of highly influential leaders while penalizing less influential ones. The feasibility and effectiveness of the proposed model are demonstrated through an illustrative example and extensive simulation experiments. Furthermore, comprehensive comparative analyses with classical opinion dynamics models and traditional trust-based CRPs demonstrate that our information diffusion mechanism significantly accelerates consensus convergence and robustly protects valuable minority insights.</p>