The containment of misinformation diffusion on social media is a critical challenge in computational social science. However, prevailing intervention strategies predominantly rely on static topological metrics or time-agnostic learning models, thereby overlooking the profound impact of temporal–demographic heterogeneity. This oversight frequently results in a “spatio-temporal mismatch”, where limited intervention resources are misallocated to structurally central but temporarily inactive nodes, particularly during non-stationary propagation bursts driven by exogenous triggers. To bridge this gap, we propose a Spatio-Temporal Deep Reinforcement Learning (ST-DRL) framework for proactive misinformation defense. By seamlessly integrating continuous trigonometric time encoding with demographic-aware Graph Attention Networks, our model explicitly captures the coupling dynamics between group-specific circadian rhythms and event-driven transmission surges. Extensive simulations on heterogeneous networks demonstrate that ST-DRL achieves a Peak Prevalence Reduction of 93.2%, significantly outperforming static heuristics and approaching the theoretical upper bound of oracle-assisted baselines. Crucially, interpretability analysis reveals that the agent autonomously evolves a “Preemptive Strike” strategy—prioritizing the sanitization of high-risk bridge nodes, such as bots, prior to event onsets—thus establishing a new paradigm for predictive rather than reactive network governance.

Social media has become an increasingly important platform for public opinion formation and mobilization. Through online communication, people can easily express their views on hot topics and form a consensus, which can lead to offline action and amplify public opinion. To help design effective communication strategies to guide public opinion, we propose a dynamic model of online and offline coupled susceptibility-latency-forwarding-immunity (C-SLFI). The model considers the dual identities of human netizens and real people in the information dissemination environment to understand the environment with online and offline communication behaviour. Based on the numerical analysis of the model, the fitting in our C-SLFI model matches the actual accident situation. In addition, sensitivity analysis of parameters shows that the average exposure rate and forwarding probability in both online and offline dimensions play an important role in controlling information dissemination. Therefore, we propose some effective strategies for guiding public opinion based on these findings.