The next speaker at AoIR 2015 is Sikana Tanupabrungsun, whose focus is on the use of Twitter by gubernatorial candidates in 36 state elections across the United States in 2014. The focus here is on @mentioning between candidates, and the analysis was conducted using automated content analysis approaches. This found that the most frequent mode of address was to attack other candidates.
Online campaigns have been studied for several years, and a general finding is that incumbents employ more position-taking strategies, while challengers operate more often in attack mode. This may play out on Twitter slightly differently because of the different affordances of the medium, though: activities are public, and any users may get involved in a conversation.
The project identified some 72 personal and campaign accounts, and captured tweets between candidates. Incumbent Democrats tended not to post much, while incumbent Republicans did; interestingly, this pattern was reversed for Democrat and Republican challengers.
Tweets were coded across a number of categories, and it turned out that categories such as Attack, Positioning, Action, and Narrative were the most prominent of these. This was used to develop a predictive model to classify further tweets and thus determine the communicative styles of individual accounts.
In total, challengers sent more calls to actions and association tweets; incumbents sent more positioning tweets and attack tweets. Attack tweets were most prominent overall, but there were fewer attack tweets in close races.
