<p>Public affairs decision-making with public participation is one of the key scenarios for multi-attribute group decision-making (MAGDM). However, existing decision frameworks face challenges, including inaccurate attribute weighting, incomplete social trust networks, and limited dimensions for evaluating expert influence. To address these issues, this paper proposes a novel public participation-oriented MAGDM framework. First, a loss aversion sentiment analysis (SA) model is developed to derive attribute weights. This model integrates loss aversion theory into traditional SA frameworks, incorporating irrational psychological traits of experts. It enables more accurate extraction of public sentiment and reduces computational bias in weight calculation. Second, a two-stage trust propagation mechanism is designed to dynamically adjust initial trust values. In the first stage, a trust amendment model based on intuitionistic fuzzy opinion distance is introduced. This model quantifies changes in trust among experts before trust propagation, which are caused by differences in opinion distances. In the second stage, these corrected trust values are incorporated into the trust propagation model to construct a complete social trust network. Third, drawing on the structural characteristics of social trust networks, this paper integrates the Von Neumann entropy (VNE) and the gravity model to propose an entropy gravity model for expert weighting, thereby meeting the weight calibration requirements of dynamic decision-making scenarios. Finally, a numerical case study is conducted to verify the effectiveness and rationality of the proposed decision-making framework.</p>