Standard approaches to the study of information diffusion draw on analogies to the transmission of diseases or computer viruses, and find that adding more random ties to a network increases the speed of information propagation through it. However, a person sharing information in a social network differs from a computer transmitting a virus in two important respects: a person may not have the opportunity to pass the information to every tie, and may be unwilling to pass the information to certain ties even when presented with the opportunity. Accounting for these two features reveals that, while additional random ties allow information to jump to distant regions of a network, they also change the composition of network neighborhoods. When the latter increases the proportion of neighbors to whom people are less willing to pass information, the result can be a net decrease in diffusion. I show that this is the case in heterogeneous, homophilous networks: the addition of random ties strictly impedes information dissemination, and the impediment is increasing in both original homophily and the number of new ties.
ASJC Scopus subject areas
- Artificial Intelligence