A mixed approach to explore how meaning, interaction, and affective valence construct online communities

In continuously shifting digital environments, the dynamics of online groups invite researchers to rethink what “community” means and how it can be identified. From a social representations perspective, we propose identifying communities through the joint analysis of shared meanings and communicative interactions - the two facets that make representations “social”. We present a strategy to integrate social network analysis and topic detection to explore how meanings and interactions jointly shape online communities. Moreover, we include the affective dimension of communication (valence) via sentiment analysis to identify the connotation of the representations at stake. The different analyses were applied to a corpus of 461 English tweets about COVID-19. Our results show that online communities can be detected by jointly considering shared contents and communicative dynamics. Sentiment analysis further accounts for the valence of these exchanges (i.e., the attitude component). We can identify different communities and dynamics: a) Topic-based groups tend to exhibit neutral or positive tones and higher cohesion; b) Interaction-based groups - centred around visible media actors - display more negative sentiment and polarization; c) Communities with highly central users (i.e., users frequently mentioned by others, especially by already influential users) show more valence-aligned (tone-aligned) rather than dispersed-valence messages. This study illustrates how mixed techniques applied to textual data can deepen the understanding of complex social phenomena, bridging content and relation, valence and meaning. In doing so, it contributes to rethinking how groups and communities can be conceptualised in mediated contexts, where interactions and meanings coconstruct the boundaries of belonging.