On Understanding Divergence of Online Social Group Discussion

摘要: We study online social group dynamics based on how members diverge in their discussions. Previous studies mostly focused the link structure to characterize dynamics, whereas behavior of content generation discussions is not well understood. Particularly, we use Jensen-Shannon (JS) divergence measure topics user-generated contents, and it progresses over time. Twitter messages (tweets) multiple real-world events (natural disasters activism) with different times demographics. also model structural user features guidance from two socio-psychological theories, cohesion identity, learn implications discussion divergence. Those show significant correlation By leveraging them are able construct a classifier predict future increase or decrease divergence, which achieves an area under curve (AUC) 0.84 F-1 score (harmonic mean precision recall) 0.8. Our approach allows systematically collective diverging independent formation design. It can help prioritize whom engage communities for specific needs during disaster response coordination, concerns advocacy brand management.