Abstract: This study investigates the intrinsic mechanisms of social bots in disseminating misinformation from the dual perspectives of emotional fields and content topics. Drawing on a large-scale dataset of misinformation on Weibo, and employing methods including machine learning, zero-shot sentiment classification, BERTopic, and negative binomial regression, this research explores how social bots participate in misinformation propagation through dual pathways: group emotional fields and topic engagement heat. Key findings include: 1) social bots tend to engage more actively with public issues that have high social influence or are prone to controversy; 2) the involvement of bots significantly increases emotional entropy, thereby enhancing the diversity of collective emotional expression; 3) emotional polarization is more likely to emerge as a natural outcome of human user interaction rather than being directly triggered by bots; 4) both emotional entropy and emotional polarization exhibit dual effects and topic heterogeneity in influencing the virality of misinformation. This study provides a dual-dimensional analytical framework for understanding misinformation dissemination mechanisms in the era of human-bot symbiosis.

Key words: Social bots, Misinformation, Emotion, Topic modeling, Machine learning