On the role of diffusion dynamics on community-aware centrality measures

Theoretical and empirical studies on diffusion models have revealed their versatile applicability across different fields. Community structure in real-world networks strongly affects diffusion. Community-aware centrality measures, which account for intra- and inter-community links, can identify key nodes within and between communities. Despite many diffusion models, few studies jointly assess how seed nodes chosen by community-aware measures perform across different diffusion dynamics and network topologies. This work compares four diffusion models (SI, SIR, IC, LT) on synthetic and real networks using two community detection algorithms to evaluate the effectiveness of community-aware centralities. Results show the diffusive power of selected nodes is shaped by: (i) community structure strength, (ii) model dynamics, and (iii) budget (fraction of seeds). Simple contagions (SI, SIR, IC) display similar patterns under the same community strength and budget, while the complex contagion LT exhibits distinct behavior.