Diffusion in networks and the virtue of burstiness

From Akbarpour and Jackson:

Significance

The contagion of disease and the diffusion of information depend on personal contact. People are not always available to interact with those around them, and the timing of people’s activities determines whether people have opportunities to meet and transmit a germ, idea, etc., and ultimately whether widespread contagion or diffusion occurs. We show that, in a simple model of contagion or diffusion, the greatest levels of spreading occur when there is heterogeneity in activity patterns: Some people are active for long periods of time and then inactive for long periods, changing their availability only infrequently, while other people alternate frequently between being active and inactive. This observation has policy implications for limiting contagious diseases as well as promoting diffusion of information.

Abstract

Whether an idea, information, or infection diffuses throughout a society depends not only on the structure of the network of interactions, but also on the timing of those interactions. People are not always available to interact with others, and people differ in the timing of when they are active. Some people are active for long periods and then inactive for long periods, while others switch more frequently from being active to inactive and back. We show that maximizing diffusion in classic contagion processes requires heterogeneous activity patterns across agents. In particular, maximizing diffusion comes from mixing two extreme types of people: those who are stationary for long periods of time, changing from active to inactive or back only infrequently, and others who alternate frequently between being active and inactive.