The Girvan-Newman algorithm detects communities by iteratively removing edges with the highest betweenness centrality — the number of shortest paths passing through an edge. Bridge edges connecting communities carry the most traffic, so they score highest and are removed first. The process repeats until the graph splits into natural clusters. Modularity Q measures partition quality: Q = (observed edges in community) − (expected by random). The algorithm runs in O(m²n) time, practical for moderate networks.