The four types of centrality used in social network analysis, degree, betweenness, closeness, and Eigenvector centrality, have a variety of different applications. Briefly, degree centrality is the total number of direct connects one node has with another, closeness considers the average distance from one node to the other nodes in the network, betweenness considers a node’s position within the totality of all shortest paths between any two given nodes within a network, and Eigenvector centrality rates the importance of the other nodes connected to the specific node in question. The application of these modes of centrality depends on the research question. For example, if one is interested in specifying individuals within a network that act as gatekeepers, betweenness might be the most appropriate metric, as it considers a node’s position in important paths within the network. Alternatively, if one is interested in the spread of a pathogen, Eigenvector centrality might help identify node that could compromise the entire network. If one is interested in determining who’s on the “inside” and who is relatively isolated within a network, closeness centrality would be a good choice. Finally, there’s degree centrality, which is somewhat descriptively weak as it doesn’t consider the broader network, but if a node exhibits a high level of degree centrality, it’s probably fairly important within its network.
Measures of centrality don’t feature prominently within every analysis section that utilize social network analysis, but researchers frequently use these metrics to describe a network. For example, in Grandjean’s (2016) article, “A social network analysis of Twitter: Mapping the digital humanities community,” Grandjean constructs a directional network using Twitter following/er relationships, and ranks them according to degree centrality, betweenness centrality, and Eigenvector centrality – note, Grandjean uses in-degree and out-degree centrality measures, which indicate the direction of the relationships which the degree centrality ranks, because Grandjean’s data is directional in nature. Simiarly, Murthy & Lewis (2014) also construct social networks based off of social media data. They utilize a survey instrument which also informs much of their analysis, but one of the metrics which they utilize to find “hubs” within their network is degree centrality – they also analyze this population of individuals with high degree centrality and find that they are more likely to be women. Personally, I find this methodology surprising – given the authors’ interest in finding hubs, ranked centrality metrics might bolster the conclusions made from their survey analysis, and I think that it is a very practical form of analysis for social media projects.
Grandjean, M. (2016). A social network analysis of Twitter: Mapping the digital humanities community. Cogent Arts & Humanities, 3(1). https://doi.org/10.1080/23311983.2016.1171458
Murthy, D., & Lewis, J. P. (2014). Social Media, Collaboration, and Scientific Organizations. American Behavioral Scientist, 59(1), 149–171. https://doi.org/10.1177/0002764214540504