Alignment and Differentiation: How Language and Network Proximity Drive Opinion-Based Group Formation Online

This study examines the interplay between language and social connectedness in forming opinion-based groups on social media. Drawing on small-world theory and social identity theory, we propose a dual-layer approach that combines semantic and network analysis to investigate the dynamics of group formation on X/Twitter during the 2021 COVID-19 vaccination campaign in South Africa. Our findings reveal a nuanced process of social sorting, where users actively align with like-minded individuals and differentiate from opposing viewpoints, shaping the formation of distinct online communities. Although a weak correlation exists between network distance and semantic similarity, our parallel examination highlights a clear trend of clustering around shared opinions. Notably, we find evidence that distance and proximity are not opposites but qualitatively different social psychological processes. Language becomes a more prominent marker of group membership over time, especially among pro-vaccination users. This suggests a dynamic process where opinion alignment is solidified through linguistic differentiation. Such sorting mechanism leads to the creation of social barriers and group boundaries. Our findings underscore the necessity of considering both structural and symbolic aspects in understanding the complex dynamics of opinion-based groups on social media. We propose a multilayer design to understand how individuals sort themselves and others within the complexities of online information flows. This approach unveils how people align with and differentiate from others, shaping the formation of opinion-based groups on social media.