Relationships under the Microscope with Interaction-Backed Social Networks

Binary friendship declarations typical of online social networks have been shown inadequate to properly capture dynamic and meaningful social relationships between users. Interaction networks, on the other hand, rely on statistical inference and assumptions on the nature of what “friendship” means. This paper analyzes an interaction-backed social network, where an interaction network and a declared social network coexist without constraining each other. We show quantitatively how many interactions take place within a declared relationship, but also that there are interactions between users without a declared relationship. By measuring interactions between declared and non-declared pairs of users, we can discover how levels of interaction wax and wane, and how attention is diverted from existing relationships to forming new ones. Our quantitative analysis can also serve scientists to create interaction workloads from declared social networks or infer social networks from interaction traces.