Why is it called information centrality?

The original current-flow work connects the measure to the earlier information-centrality formulation.

Can adding an edge matter even when no shortest path becomes shorter?

Yes. Additional routes can reduce effective resistance.

Is its value always between zero and one?

Not under every scaling and conductance choice. The convention and units affect the numerical range.

AI Glossary / Graphs and symbolic AI / Advanced

Current Flow Closeness

Definition

Current-flow closeness is a centrality measure based on effective resistance between vertices in an electrical-network model of a graph. A vertex scores higher when its total effective resistance to the other vertices is smaller. The measure is also known as information centrality.

What it means

Ordinary closeness uses shortest-path distances. Current-flow closeness replaces those distances with a quantity that reflects the combined effect of alternative connections. Several routes can make two vertices electrically close even when their shortest route does not change.

Effective resistance is the potential difference required to send one unit of current between two vertices when the rest of the graph remains available. It incorporates the whole network, not just one selected path.

The electrical interpretation does not require the graph to represent wires. It can serve as a structural model for information or interaction networks, provided the analogy suits the question. A high score does not independently establish influence, traffic volume or service quality.

Numerical conventions matter. The NetworkX implementation considered here returns the reciprocal of the sum of effective resistances to the other vertices. A convention using reciprocal average resistance adds a factor of n - 1, so raw values from different sources may differ even when rankings agree.

How it works

Represent a connected undirected graph with positive edge conductances. A larger conductance represents a stronger or easier connection; resistance is its reciprocal. Construct the weighted graph Laplacian and solve the relevant linear systems with one potential reference fixed, or use an equivalent matrix formulation.

For each pair of vertices, determine effective resistance. Sum those values from the vertex being scored to all others, then invert the sum under the stated NetworkX convention.

Alternative routes lower effective resistance relative to having only one route. This makes the measure sensitive to redundancy that ordinary shortest-path distance can ignore. However, it does not directly model independent failure probabilities or operational capacity constraints.

Changing every conductance by the same factor changes the score's scale. Comparisons therefore need compatible units and preprocessing. Missing edges or different choices about which entities belong in the graph can also alter the result.

The cited implementation requires connectedness for nontrivial graphs and does not accept directed graphs. Analyzing components separately changes the population in each sum, so component scores should not be combined without a justified comparison convention.

Three practical examples

A graph analyst compares nodes that have similar shortest-path distances but different numbers of alternative connections.
A communications study examines how a proposed extra link changes effective connectivity under a conductance model.
A teaching dashboard displays ordinary and current-flow closeness side by side to show that different notions of distance produce different centrality measures.

Think of it this way

Think of reaching a destination through several available corridors rather than evaluating only the shortest corridor. Multiple alternatives can improve overall connectivity even when the shortest route remains unchanged.

The analogy is limited because effective resistance follows electrical laws, not walking time. A second corridor's effect on people depends on congestion, access and behavior that this graph model may not represent.

Worked example

A path becomes a triangle

Start with an undirected path A-B-C whose two edges each have resistance 1. With no alternative routes, effective resistance from A to B is 1, from B to C is 1 and from A to C is 2.

Under reciprocal-sum scaling, B has score 1/(1 + 1) = 1/2. A and C each have score 1/(1 + 2) = 1/3. B is electrically closer to the rest of the path.

Now add edge A-C with resistance 1, making a triangle. Between any pair, the direct resistance 1 is in parallel with the two-edge resistance 2. Their equivalent resistance is 1/(1 + 1/2) = 2/3.

Every vertex now has total effective resistance 2/3 + 2/3 = 4/3, giving current-flow closeness 3/4. The vertices tie by symmetry.

B's ordinary shortest-path distances to A and C were already 1 and remain 1, so its ordinary closeness does not improve in this comparison. Its current-flow closeness rises from 1/2 to 3/4 because the added connection supplies another route. The calculation isolates a structural effect; it is not a measured performance improvement.

Illustrative case study

Fictional case: Cedar Networks compares two proposed office-link designs. A shortest-path report shows no change in distance from the central hub to either neighboring hub, so the first review concludes that the added link has no connectivity benefit.

A current-flow analysis reveals reduced effective resistance. The team uses that finding to motivate a closer operational review, checking actual bandwidth, shared physical routes and expected demand before making a decision.

The lesson is that a metric can reveal a structural property without settling the business case.

Common misconceptions and mistakes

Current-flow closeness uses only the shortest route. Effective resistance reflects alternative connections.
It is the same as current-flow betweenness. Closeness concerns connection to others; betweenness concerns intermediary flow.
A score is meaningful without its scaling convention. Reciprocal-sum and reciprocal-average forms differ.
Edge length can be used directly as conductance without thought. That can invert the intended relationship.
A higher score proves better reliability. Failure mechanisms are not automatically represented.

Business use

Current-flow closeness is useful when the analysis concerns distributed connectivity rather than only shortest travel. It can complement other graph measures in infrastructure, information-network or relationship studies.

State the graph scope, conductance interpretation and scaling convention alongside the results. Check sensitivity to plausible weight changes and missing links. If comparisons span different network sizes, consider the effect of the resistance sum and any normalization. Use the score to support a defined question rather than as a generic ranking of organisational value or importance.

Check your understanding

In the unit-resistance path A-B-C, what is B's current-flow closeness under reciprocal-sum scaling?

Answer: 1/2, because its effective resistances to A and C are both 1, and 1/(1 + 1) = 1/2.

Related terms

Current Flow Betweenness
Normalization
Effective resistance
information centrality
conductance
graph Laplacian
closeness centrality
parallel paths

Learn next

Repeat the triangle calculation after doubling every conductance. Determine how effective resistances and the reciprocal-sum scores change. Explain why this scaling effect makes unit conventions essential when comparing two weighted networks.

Sources

NetworkX current-flow closeness documentation: https://networkx.org/documentation/stable/reference/algorithms/generated/networkx.algorithms.centrality.current_flow_closeness_centrality.html

NetworkX source confirms reciprocal-sum output and graph requirements: https://networkx.org/documentation/stable/_modules/networkx/algorithms/centrality/current_flow_closeness.html

Brandes and Fleischer, original research: https://www.uni-konstanz.de/algo/publications/bf-cmbcf-05.pdf

The path-to-triangle calculation and fictional review are original illustrations.

Previous: Current Flow Betweenness Next: Curriculum Learning