Attentive Betweenness Centrality (ABC): Considering Options and Bandwidth When Measuring Criticality

Betweenness centrality measures how critical a node is to information flow in a network. A node is critical (and hence should have high betweeness) if it is on many shortest paths. Two shortcomings of such a measure are: (i) It ignores nodes on “almost shortest” paths; (ii) It assumes that a node can provide the same attention to information flow through each of those shortest paths, no matter how many shortest paths the node controls. There have been attempts to address these concerns in the literature, with partial success. We provide a new measure, attentive betweenness centrality (ABC), that measures criticality by the amount of attention a node devotes to the information flow between other nodes. Our measure addresses both the aforementioned concerns and can be computed efficiently. It performs as well or better than betweenness centrality on both stylized networks and large scale real data networks, and hence provides a useful tool for measuring node criticality.