Focusing on the Realistic Challenges of China's Intelligent Social Governance:How Algorithmic Recommendation Shapes Social Conflict and Affective Polarization

In the era of intelligent algorithms, the erosion of institutional trust has emerged as a critical challenge for contemporary social governance. This research systematically investigates how algorithmic recommendation environments catalyze the decline of institutional trust (EIT) within the context of China’s social governance. By integrating Structural Equation Modeling (SEM), Necessary Condition Analysis (NCA), and Multimodal Deep Learning, this study employs a tripartite methodological triangulation to validate the complex dynamics. Study 1 SEM reveal that the erosion of trust follows an affective-dominant pathway rather than a purely cognitive one; Study 2 NCA demonstrates that affective polarization serves as a necessary condition (bottleneck) for trust erosion, suggesting that systemic trust collapse is unattainable unless polarization exceeds a critical threshold. Finally, Study 3 utilizes a multimodal deep learning architecture to analyze high-dimensional data, revealing how the confluence of inflammatory text and polarized visual symbols generates non-linear negative shocks to perceived governance legitimacy (PGL).