Identifying the structure patterns to govern the performance of localization in regulating innovation diffusion

Abstract The macro social influence is recognized as a non-negligible ingredient in innovation propagation: more adopters in the network lead to a higher adoption tendency for the rest individuals. A recent study to incorporate such a crucial mechanism shows that sufficient intensity of macro-level social influence can cause a change from a continuous to discontinuous transition, further indicating the existence of a tricritical point. Although network localization strength determines the tricritical point, it remains unclear what network quantities govern the performance of localization in regulating innovation diffusion. To address this issue, we herein consider the model incorporating both the micro- and macro-levels social influence. We present a dynamic message-passing method to analytically treat both the outbreak threshold and recovered population, and validate the predictions through agent-based simulations. Extensive analysis on the classical synthetic networks shows that sparsely available connections, and relatively heterogeneous degree distribution, either assortative or extremely disassortative configurations are favorable for continuous transition. In such cases, the employed network can yield a strong localization effect so that the innovation is trapped in the configurations composed of the hubs with high non-backtracking centrality. We further explore the dependence of both tricritical point and localization strength on three structural quantities: network density, heterogeneity, and assortativity, which gives a clear physical picture of the joint effects of the three structure quantities on the localization strength. Finally, we conclude that the core-periphery structure, being sensitive to the change of the three structure quantities, essentially determines localization strength, and further regulates the phase transition.