How do network structures affect behavior diffusion dynamics? One determinant is the social contagion mechanism. Diffusion models assuming contagion from a single contact with infected nodes predict that less clustered ‘small world’ networks can better spread the behavior because the average path length is smaller. While threshold models assuming contagion from multiple contacts to different sources predict that more clustered networks would better facilitate diffusion because of the redundant ties. The other determinant is individuals’ probability of being influenced given the same number of contacts with infected nodes across different network structures they are embedded in. Based on a natural experiment in an online social network website, we find that social contagion requires mult...[ Read more ]

How do network structures affect behavior diffusion dynamics? One determinant is the social contagion mechanism. Diffusion models assuming contagion from a single contact with infected nodes predict that less clustered ‘small world’ networks can better spread the behavior because the average path length is smaller. While threshold models assuming contagion from multiple contacts to different sources predict that more clustered networks would better facilitate diffusion because of the redundant ties. The other determinant is individuals’ probability of being influenced given the same number of contacts with infected nodes across different network structures they are embedded in. Based on a natural experiment in an online social network website, we find that social contagion requires multiple contacts with different infected nodes, and people who are embedded in more clustered local network structures are more likely to be influenced given the same number of contacts with infected nodes. These empirical evidences provide sound support that more clustered network can better facilitate the spread of human behavior.