How Did Ebola Information Spread on Twitter?

Theoretical Background and research questions/hypothesis: Social networks are beneficial for the communications of public health information. In the pre-Internet age, large-scale dissemination of health information relied on mass media broadcasts. The importance of interpersonal communication has been celebrated in the age of social media. The interpersonal transmission of online messages could be analogous to the spread of the infectious diseases, and such type of information diffusion is referred as the viral spreading mechanism. The primary purpose of this project is to examine whether the traditional broadcast mechanism or the viral spreading mechanism dominated the Ebola information diffusion on Twitter. Lessons learned can then contribute towards developing more effective communication strategies.

Methods: Our data was purchased from GNIP, the official Twitter data provider. We obtained all Ebola-related tweets (including retweets and replies) posted from March 23, 2014 to May 31, 2015. We reconstructed the Ebola retweeting paths based on Twitter contents and the following relationships. Social network analysis was performed to investigate the retweeting patterns. 

Results: On average, 91% of the retweets were directly retweeted from the initial tweeters. On average, the maximum number of steps of information transmission between retweeters and the initial tweeters was 3, i.e. Initial user -> Follower -> Follower -> Follower. These observations suggested that large viral networks were uncommon and broadcast spreading was more pervasive than viral spreading for Ebola-related tweets. According to the retweeting and following relationships, four types of initial users were identified: "broadcasters" (38%), "common users" (60%), "influentials" (2%), and "hidden influentials" (<1%). The "broadcasters" were users with many followers, such as celebrities, but received less retweets than expected. "Influentials" were the users who received many retweets as expected, whereas "hidden influentials" were those with much less followers but received more retweets than expected. In addition, messages generally flowed from a few core participants who are most active and well-connected with each other to the other common users (i.e., peripheral participants) and the core participants were more likely to initiate large-scale information flows. 

Conclusions: Broadcast spreading was the dominant mechanism (vs. viral spreading) for infectious disease outbreak-related (e.g., Ebola) information on Twitter, and it could lead to large scale dissemination. 

Implications for research and/or practice: Unlike social networks with viral spreading, the initial tweeters (i.e., sources of information) and their messages are much more important than common users if the broadcast spreading mechanism is dominant. Therefore, as far as Ebola health communication was concerned, "influential" or "hidden influential" sources (e.g., social media accounts operated by traditional mass media) would be important partners to disseminate Ebola-related health information.