Influencers, Multipliers, and the Structure of Polarization: How Political Narratives Circulate on Twitter/X
Mathematical Approach
To uncover these patterns, the study combined advanced computational methods. A machine-learning based topic modeling approach inductively classified millions classified millions of tweets into thematic categories across thousands of trending topics, creating a high-dimensional map of public discourse. Retweet activity was then modeled as directed network, with links representing retweets between users. Applying stochastic block modeling — a statistical method for detecting community structure in networks — the researchers identified, for each trend, whether the retweet network is polarized or not, and extracted the clusters that correspond to opposing ideological camps. Finally, an alignment metric was developed to measure how consistently individual users stayed within the same ideological block across topics. Multipliers - whose authenticity is also supported by the study - consistently showed higher alignment scores than influencers, indicating their role in binding issues together.
Issue Alignment Across Topics
Polarization is not limited to a few “hot-button” issues but spans multiple political fields simultaneously. The analysis of the top 1,000 influencers and multipliers reveals that multipliers are more active and maintain stronger ideological alignment across topics than influencers. While most political issues are highly aligned, non-political topics such as gaming and music attract different user groups and remain weakly polarized. A similar realignment effect appeared for Ukraine-related discussions, where some right-leaning influencers broke with their cluster’s dominant pro-Russian stance — a pattern again less visible among multipliers. Across all topics, multipliers consistently show higher issue alignment than influencers. The topic alignment matrix shows a high issue alignment across topics, except for Music and Gaming, with a gradual difference between strongly aligned issues like Covid, Journalism, and German politics, and less aligned topics like Social politics.
User activity patterns show that multipliers are typically active in a larger number of trends than influencers. A comparison of the size of political clusters highlights further differences. Among the 1000 most retweeted influencers, we observe a majority of accounts belonging to the left-liberal cluster. This is reversed for multipliers: among the 1000 most active multipliers, we observe a majority of users from the right-conservative cluster.
Future Research Directions
The authors highlight the importance of considering the role of social media in shaping public opinion and the need for further research into the mechanisms driving polarization. The study has limitations that call for further research, such as the need to shed light on the discursive reasons of issue alignment, which the authors plan on investigating through the lens of conflicting narratives. Further research is needed to explore the alignment of regular users and to determine if similar patterns of polarization exist on other social media platforms. Preliminary findings suggest that consistent polarization and issue alignment may hold for the majority of users, and that influential hyper-active users on other platforms, such as Facebook, may have similar effects as multipliers on Twitter.
