Research on the influence mechanism of emotional communication on Twitter (X) and the effect of spreading public anger
Year: 2025
Type: article
Abstract: BACKGROUND: Anger-driven content has a significant influence on engagement and discourse on social media platforms, particularly Twitter. This study investigates the mechanisms through which anger spreads on Twitter, analyzing the role of user interactions, algorithmic amplification, and network dynamics in shaping emotional discourse. METHODS: A dataset of 5000 tweets was collected from Twitter using API access, spanning political discourse, social movements, and crisis-related discussions. Sentiment and toxicity analysis were conducted using VADER and the Perspective API, while network analysis applied the Louvain method to identify clusters of anger-driven interactions. Regression models, propensity score matching, and time-series analysis were employed to examine the relationship between toxicity, engagement, and anger amplification. The predictive performance of other models was compared with the meta-model. RESULTS: Among 5000 tweets analyzed, 70.0 % originated from ordinary users and 30.0 % from public figures. Toxic content (20.0 %) was strongly correlated with engagement (retweets: r = 0.52, p < 0.001; likes: r = 0.48, p < 0.001; replies: r = 0.43, p = 0.002). Regression confirmed that toxicity significantly predicted anger spread (β = 0.320, p < 0.001), with visibility score also exerting a strong effect (β = 0.045, p < 0.001). Anger-driven discussions formed large clusters (avg. 215.4 users), and toxic tweets received 27.1 % higher visibility and 85.7 % more retweets. The proposed meta-model (RF + SVM + ARIMA) achieved the highest accuracy (0.89), outperforming deep learning approaches such as BERT (0.87) and LSTM (0.86). These amplification results are based on observational modeling of engagement patterns and should be interpreted as indicative of algorithmic intervention. CONCLUSION: Our findings underscore the need for targeted interventions, including algorithmic adjustments to mitigate the disproportionate amplification of anger, as well as enhanced content moderation strategies to foster healthier digital interactions.
Source: Acta Psychologica
Authors: Chunqiong Wu, Shan Jiang, Jianhong Sun, Yingqi Liu
Institutions: Yango University, Ningbo University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,670
OpenAlex ID: W4414334608