Popularity and similarity among friends: An agent-based model for friendship development

Abstract

This work investigates how local preferences, social structural constraints and randomness might affect the development of the friendship network in Facebook. We do this by analyzing a snapshot Facebook dataset of Princeton University’s students, and by building an agent-based simulation for comparison. Several different, but plausible, processes of friendship network development are proposed in which the structural information of the growing network and the student preferences are taken into account and then compared with the data. ‘Network formation based on personal preference and social structure with some randomness’ matches the data best, and is thus the preferred hypothesis for the way that students add “friends” on Facebook.