Identifying Network Mechanisms Underlying the use of #hongkongpolicebrutality
Rong Wang, Alvin Zhou, François BarRecent Hong Kong protests started in 2019 with the public’s opposition to the government’s proposed amendment to the extradition law but have deeper roots in other contentious matters. The protests were initially peaceful. However, news coverage from both western and Chinese media reported violent clashes between the protesters and the Hong Kong Police, leading to a viral hashtag #hongkongpolicebrutality. The goal of this study is to uncover network mechanisms underlying the use of this hashtag, and to understand what factors may drive users’ decision to retweet a trendy tweet or engage in a more direct conversation.
We conceptualize the retweet ties as information sharing networks while the ties constructed through direct mention or reply to consist conversational networks. We draw from the literature on online activism, collective action, and social movement to identify what factors might lead to Twitter users’ different decisions in participating in the use of #hongkongpolicebrutality. We expect that the virality of the hashtag might be attributed to several mechanisms including the influence of public figures, local and global human rights organizations, and the diversity of twitter accounts that contributed content.
We collected Twitter data using #hongkongpolicebrutality from October 15th 2019 to January 30th 2020. This particular time frame was selected due to the increasing number of media coverage about the tension between the protesters and the police after China’s 70th anniversary parade. 204079 tweets were archived through an open source data collection tool with Twitter Streaming API. Then, we used these tweets’ unique IDs to extract relevant metadata from the official Twitter Search API, which yielded 185959 tweets. We sampled all the English tweets for further analysis (n = 137580).
The following procedures were taken to further construct the dataset. First, all the original tweets and their retweets were selected, generating a total of 46033 unique tweets which had been retweeted. A total of 52497 twitter users were identified as nodes in the weighted and directed retweet network with 126567 ties. Second, the second network was constructed by building a tie between two users that had a direct engagement, which required us to separate actual mentions from mentions that resulted from retweets. Therefore, this network was solely based on conversational engagement, through either direct mention or direct reply. This weighted and directed network contains 15882 Twitter users as nodes and 48724 ties. Third, attribute data were collected for all the sampled Twitter users in the network, including number of followers, number of followees, geolocation, and dates when they joined Twitter.
Exponential Random Graph Modeling (ERGM) will be conducted on two types of networks to identify how user level attributes and structural features may influence users’ decision to either retweet or engage in more direct communication. This study contributes to the network literature by unpacking the multiplexity underlying online collective action and examining network mechanisms that drive different levels of engagement. Implications on how a collective action hashtag goes viral and what types of users are more likely to be more engaged in online activism are provided.