Betweenness centrality
Betweenness centrality is a measure of the extent to which a node in a network lies on the shortest paths between other nodes. It was introduced by sociologist Linton Freeman in 1977 as a way to quantify the "brokerage" role of individuals in social networks. A node with high betweenness centrality is not necessarily the one with the most connections (that is degree centrality); it is the one that most frequently mediates the flow of information, resources, or influence between parts of the network that would otherwise be disconnected.
The mathematical formulation is straightforward. For a node v, its betweenness centrality is the sum, over all pairs of nodes (s,t) not including v, of the fraction of shortest paths between s and t that pass through v. In a network of n nodes, this can be normalized by dividing by the maximum possible value, (n−1)(n−2)/2, yielding a score between 0 and 1.
Interpretation: Broker vs. Bridge
Betweenness centrality captures a specific kind of network importance: the capacity to control or monitor flows. In social network analysis, a person with high betweenness centrality occupies a "broker" position — they can gatekeep information, introduce ideas from one cluster to another, or selectively amplify or suppress signals. This is the individual-level analogue of the structural holes concept developed by Ronald Burt.
But betweenness centrality is not identical to brokerage. An actor can be structurally peripheral yet span a critical hole between two large clusters. Conversely, an actor can be highly central within a closed cluster without spanning any structural holes at all. The relationship between the two concepts is explored in Structural holes.
The "bridge" interpretation differs from the "broker" interpretation in emphasis. A broker is an agent who exploits a structural position for advantage. A bridge is a topological feature that happens to connect clusters. The same node can be a bridge without being a broker (if it lacks awareness of its position or capacity to exploit it) and a broker without being the only bridge (if there are alternative paths that happen to be longer).
Ecological Applications
In ecological networks, betweenness centrality identifies species whose loss would most fragment the food web. These are not always the most abundant species or the apex predators. They are often "keystone species" that occupy critical intermediary positions — species that feed on multiple trophic levels and are preyed upon by multiple predators. As noted in Bioaccumulation, species with high betweenness centrality can become disproportionate reservoirs of toxins because they channel flows from multiple contaminated pathways.
The concept also applies to metapopulation dynamics. Habitat patches that serve as dispersal corridors between otherwise isolated populations have high betweenness centrality in the landscape network. Their conservation is often more critical for species persistence than protecting the largest or most species-rich patches, because their loss severs the network into disconnected components.
Computational and Algorithmic Aspects
Computing exact betweenness centrality for all nodes in a network with n nodes and m edges requires O(nm) time for unweighted graphs using Brandes' algorithm and O(nm + n² log n) for weighted graphs. For large networks — social media platforms, the internet, brain connectomes — exact computation is prohibitive, and approximation algorithms are used. The Girvan-Newman algorithm uses betweenness centrality iteratively to detect community structure: it removes the edge with highest betweenness centrality, recomputes, and repeats, producing a dendrogram of nested communities.
Betweenness centrality is not without limitations. It assumes that flow follows shortest paths, which is realistic for some processes (packet routing, gossip) but not for others (random walks, diffusion). Variants such as random walk betweenness centrality relax this assumption. It also treats all pairs equally, which may not match real-world importance: some pairs may matter more than others.
The Systems View
From a systems perspective, betweenness centrality is a diagnostic of network vulnerability. High-betweenness nodes are single points of failure: their removal disproportionately increases the average path length and can fragment the network. This is why infrastructure networks — power grids, transportation systems, the internet — are designed with redundancy to reduce betweenness centrality at any single node. The presence of a node with very high betweenness centrality is not a sign of strength; it is a sign of structural fragility.
The measure also reveals the gap between local and global network properties. A node can have low degree centrality (few direct connections) but high betweenness centrality (mediating between otherwise distant clusters). This is the network-science version of the "weak tie" phenomenon: the most important connections are often not the strongest ones.
The persistent confusion of betweenness centrality with general importance — the assumption that the most central node is the most powerful, influential, or valuable — is a failure mode that runs through the entire network-science literature. Betweenness centrality measures one thing: mediation. It says nothing about power, nothing about value, nothing about legitimacy. To treat a broker as a leader, or a bridge as a destination, is to misread the network entirely.