Answers are grounded in this paper and cite their sources.
The rapid growth of social media has transformed misinformation dissemination into a multimodal and network-driven phenomenon. This study proposes a multimodal analytical framework to investigate how textual claims, visual signals, emotional expressions, and user cognitive biases jointly influence misinformation diffusion on social media platforms. This paper establishes a data set with 4,000 public posts obtained from January to March 2026 on Sina Weibo platform as the study context. 2,000 posts are selected as misinformation posts based on their refutation record while 2,000 posts are chosen as factual posts based on their verification record. A multimodal detection model based on Chinese RoBERTa and ResNet-50 and repost-comment networks are built in the experiments, the results show that the multimodal model has AUC of 0.934 ± 0.012 and F1 of 0.887 ± 0.016 and is 0.092 more than that of the text-based methods. Misinformation posts have a greater diffusion intensity, since there is a 3.84 ± 1.26-level rise in the average depth of the reposts and an increase of 0.027 ± 0.009 in network density. Negativity bias, confirmation bias, and authority cues are significant predictors in the regression analysis.
No takes yet. Share an insight, caveat, or question.
王淑一 (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: