Who Receives Support, and Who Generates Interaction?: Examining Influencer Types and Post Characteristics Using MRQAP Analysis
In online communities, view counts, upvote counts, and comment counts indicate how users engage with posts. Posts receiving many view counts and upvote counts may be considered supported by other users, whereas posts with many comment counts tend to generate active discussion by prompting additional opinions. Although comment counts can be positively correlated with view or upvote counts, some users may receive high levels of support but limited discussion, while others may generate discussion despite receiving relatively little support. Thus, high support does not necessarily lead to active discussion, and vice versa. Both types of users exhibit high levels of engagement with their posts and can therefore be classified as influencers, but their characteristics differ depending on whether they primarily receive support from other users or encourage additional discussion. Consequently, treating them as a single type of influencer may obscure the distinct influences and roles that each type plays within the community. Accordingly, this study classifies influencers into three types based on whether they meet criteria for the level of support (view counts /upvote counts) and interaction (comment-based conversation networks), and aims to identify how the characteristics of the posts they create differ across these types.
Using data from the Politics/Current Affairs board of FMKorea, one of the largest online communities in South Korea, this study constructs a conversation network based on post–comment and comment–reply relationships. The conversation network constructed from 20,306 posts and 87,162 comments consists of 4,765 nodes and 52,458 edges. Users with high in-eigenvector centrality are classified as “interaction influencers,” while users receiving high levels of support based on view counts and upvote counts are classified as “support influencers.” Users meeting both criteria are classified as “super influencers.” This approach enables a comparison between users who occupy central positions in discussions via posts and comments and those who receive support from the other user. To compare the three influencer groups, we conduct multiple regression quadratic assignment procedure (MRQAP) analyses using two dummy-coded group variables as independent variables. We estimate separate models in which the dependent variable is pairwise similarity in one of four post characteristics: post length, word use, the use of external links, and the number of comments authors leave on their own posts. These analyses examine whether similarities in post characteristics differ by influencer group.
As an expected result of this study, the similarity in word use is expected to decrease significantly among interaction influencers compared to support influencers. The use of common vocabulary, such as popular expressions used within the community, may help influencers receive more support from other users. The similarity in the use of external links is expected to be higher among interaction influencers than among support influencers. Providing information from external sources in posts may encourage users to participate in further discussions through comments. Interaction influencers and super influencers are also expected to have more similar post characteristics to each other than to support influencers.
This study contributes to the literature by distinguishing influencers from two perspectives—levels of support and the formation of interactive discussions—in order to account for their distinct roles their distinct roles, and by identifying how the characteristics of posts created by influencers who receive support differ from those of influencers who lead discussions. From a social perspective, this study provides insight into the different roles that influencers may play in the process of opinion formation in anonymous online communities, such as sharing information from external sources and encouraging interaction.
