Affective Expression and Agenda‐Setting in Indian Political Discourse on Social Media

ABSTRACT

Social media platforms have become key spaces where political discourse unfolds and public engagement takes shape. Yet users bring different emotional styles and expressive habits into these environments, producing visible variation in how political conversations evolve online. This study analyses 50,000 tweets from Indian political discussions on X (formerly Twitter) to examine how emotional expression, engagement behaviours and influence patterns differ across users. Applying sentiment analysis (TextBlob, validated against a transformer‐based BERT model), Latent Dirichlet Allocation (LDA) for topic modelling, Poisson generalized linear models for engagement analysis and mention‐based network mapping, the study addresses three research questions: how sentiment relates to engagement behaviour, which political topics dominate discourse and how they are framed emotionally, and how influence is distributed among users. The findings show that tweets with a more positive emotional tone receive higher levels of affective endorsement through likes. More specifically, positively framed tweets are expected to receive approximately 23% more likes than neutral tweets, whereas negatively framed tweets receive fewer likes, suggesting that affective validation and content amplification operate through distinct behavioural pathways. Election‐related content displays markedly stronger emotional polarization, suggesting that politically contentious themes evoke sharper affective responses. The network analysis further indicates that agenda‐setting influence is concentrated among a small group of elite actors, around whom much of the expressive and behavioural activity clusters. By integrating sentiment orientation, topic salience and network centrality into a unified analytical framework and tracing these differences at scale, the study contributes to work on emotional contagion, digital polarization and political communication by providing measurable digital indicators of the emotional climate of political discourse in digitally networked public spheres.