Sunday 09 March 2025
In a recent study, researchers delved into the world of social media and information operations (IOs). IOs are carefully crafted campaigns designed to shape public opinion, often by spreading misinformation or influencing online conversations. The researchers aimed to uncover the differences in these campaigns between English- and Spanish-speaking communities.
The team analyzed over a million tweets about the 2024 US presidential election, focusing on the behavior of users who drove these information operations. They constructed similarity graphs from behavioral patterns, allowing them to identify IO drivers in both languages.
One key finding was that the topics and political indicators discussed by IO drivers differed significantly between English and Spanish speakers. English-speaking IO drivers focused on issues like politics, economy, and healthcare, while their Spanish-speaking counterparts emphasized immigration, education, and social justice.
The researchers also explored the role of bilingual users who post in both languages. They discovered that these individuals exhibit distinct behaviors compared to monolingual users, often serving as bridges between different online communities.
To detect IO drivers, the team employed various network-based methods, including edge filtering and node pruning. Edge filtering involves removing edges with low weights, while node pruning involves deleting nodes with low centrality scores. The researchers found that a combination of both techniques worked best, but noted that each language responded differently to these methods.
The study also revealed differences in the most popular web domains shared between English and Spanish tweets. While English-speaking users favored conservative media outlets like Breitbart and Fox News, their Spanish-speaking counterparts focused on domains such as Voz de América and Clarín.
The researchers visualized the sentiment shift over time for IO drivers from both languages, finding a similar trend for positive and negative sentiments. However, they noted that the peak of positive sentiment in English-speaking users coincided with a significant increase in August 2024, while Spanish-speaking users exhibited a more sustained positive sentiment throughout the period.
The team’s analysis also highlighted the importance of caution when using centrality scores to detect social bots. They found that nodes with high eigenvector centrality often corresponded to organic users rather than bots.
Overall, this study sheds light on the complex landscape of information operations and their differences across linguistic boundaries. By understanding these patterns, researchers can develop more effective strategies for detecting and mitigating the impact of IOs on online discourse.
Cite this article: “Language Barriers in Information Operations: A Comparative Study of English- and Spanish-Speaking Communities”, The Science Archive, 2025.
Information Operations, Social Media, Language, Politics, Election, Misinformation, Online Conversations, Network Analysis, Sentiment Analysis, Centrality Scores.
