Centrality Measures (Multiple Flavors)¶

Why It Matters¶

"Important" is multidimensional in networks. The most-connected node (degree) might not control information flow (betweenness) or have fast access (closeness) or connect to other important nodes (eigenvector). Different centrality measures reveal different strategic positions. Betweenness identifies bottlenecks and single points of failure. Closeness identifies efficient broadcasters. Eigenvector identifies elite connections. Confusing these dimensions leads to misidentifying who/what matters for your specific purpose.

The Key Move¶

Before analyzing network importance, ask what kind of importance you care about: connection count, brokerage/control, speed of access, or quality of connections? Calculate the appropriate centrality measure. Compare ranks across measures - nodes with high degree but low betweenness are well-connected but not gatekeepers; nodes with high betweenness but low degree are critical bridges. Map centrality to your actual problem (influence spread = closeness; information control = betweenness; resilience = degree).

Classic Application¶

Epidemic contact tracing. Early strategies targeted high-degree individuals (most contacts). But betweenness centrality better identifies superspreaders who bridge otherwise separate groups. Vaccinating high-betweenness individuals fragments transmission paths more effectively than vaccinating high-degree individuals within already-connected clusters.

Surprising Application¶

Personal productivity systems. Task networks where nodes are projects and edges are dependencies/shared-resources reveal workflow bottlenecks. High betweenness tasks are critical path items - delays cascade widely. High degree tasks are coordination-heavy. High closeness tasks are good leverage points for inspection. Measuring task centrality reveals which work items deserve disproportionate attention versus which are safely parallel.

Failure Modes¶

Treating centrality as one-dimensional - different measures answer different questions; using the wrong measure gives misleading answers. Ignoring dynamics - centrality changes as networks evolve; today's broker may be tomorrow's peripheral node. Assuming correlation - degree and betweenness often don't correlate; high-degree nodes in dense clusters have low betweenness. Computing centrality on wrong network - directed vs. undirected, weighted vs. unweighted fundamentally change rankings.

Go Deeper¶

Newman, "Networks: An Introduction" Chapter 7 (Measures and Metrics); Borgatti, "Centrality and network flow" (Social Networks, 2005)

Leverage Points - Operations Research & Systems Thinking: A Map of the Territory