The impact of content and sentiment coherence on information diffusion

Abstract

While online discussion has been extensively studied in the previous literature, the role of information coherence along the course of information diffusion remains underexplored. In this study, I investigate the influence of information coherence in the online discussion context with a focus on the sequential patterns of user posts. I measure the information coherence using text mining techniques by two dimensions: content and sentiment. Using the data from popular online automobile discussion platforms in China, I empirically estimate the effect of information coherence on the duration and breadth of information diffusion. The empirical evidence sheds light on the heterogeneity of the coherence influence. It indicates that relevant content motivates more replies in a shorter duration, whereas consistent sentiment is associated with longer duration and fewer replies. The moderating analysis on the topic further deepens the understanding of the relationship between information coherence and information diffusion. I found that under the topic of information acquisition (vs. experience sharing), relevant content and consistent sentiment can end the discussion about information acquisition faster with fewer replies. Besides, the moderating analyses on user interaction suggest that user interaction does not always increase participation in terms of more replies and longer duration. High content dependency (vs. low content dependency) increase participation more efficiently under low content coherence, whereas high original poster's replies (vs. low original poster's replies) increase participation more efficiently under high sentiment coherence. The findings provide managerial insights for platforms to facilitate online discussion and for companies to facilitate product management.