Twitter Conversation Dynamics

Rasha Abdulla, Alessandro Dondero, Navid Hassanpour, David Montez, Augusto Valeriani

Question and Objectives: Mapping Twitter conversation dynamics, creating a GUI for network mapping purposes, assessing the effect of external interventions in social processes on the structure of conversation network.

H1: Network structure of a (viral) conversation is denser than a normal conversation, static conversation networks are star-shaped, viral ones have smaller average betweenness and a larger number of cliques.

Methodology: Comparing conversation network statistics (Replies and Retweets in the dataset) before and after a certain intervention during the event

Findings:

What we expect to see: If we see a growing density in the overall network, will should see diminished centrality among previously influential nodes.

During political campaigns we normally see a situation where opinion leaders (politicians, journalists and celebrities) tweet – i.e. they start the conversations--and citizens (ordinary users) reply or retweet. (Star structure)

When something unexpected happens, citizens talk (retweet and replies) much more among themselves (Different typologies of tree structure)

An unexpected event disrupts the structure of conversations and thus reduces the centrality of opinion leaders within the political discussion.