Unpacking the network dynamics of online political discussions: Stochastic actor-oriented modeling with political/sociopsychological/linguistic factors

The underlying mechanisms of online political discussions which may involve power dynamics have seldom been explored through a dynamic network approach, even though discussions themselves are inherently relational and dynamic processes. It remains unclear how discussions are shaped over time between egos/alters with different political/sociopsychological/linguistic attributes as well as by the existing discussion ties surrounding them. Based on a stochastic actor-oriented modeling constructed from pre-survey/virtual discussion/post-survey data, we found some evidence that, in the politically heterogeneous context, shared political ideology, alter’s political tolerance level, and the extent of negative/strong/assertive language used by alter are associated with the likelihood of discussion tie creation/maintenance. In the homogeneous context, the extent to which ego perceives alter as politically knowledgeable and personally likable was clearly associated with ego’s tendency to create or maintain discussion ties with alter over time. This study uncovers alter/dyad-centered variables that are critical, but have been overlooked, for communication research.