Predicting Tie Strength with Ego Network Structures

Abstract

Not all social media “friends” are close friends, but distinguishing them from mere acquaintances is an important task in marketing. The notion of a close friend is reflected in the metric tie strength, but the true tie strength is often unobserved in online social networks. With this research, we propose an approach that predicts real-world tie strength via online data measures of similarity, interaction, and network data. At its core, we assess ego network structures to predict tie strength, i.e., all first-degree connections and the interlinkage among them. Ego networks are easier to obtain than full networks, and researchers can process them more efficiently. We explain why bridging ego network positions could be associated with real-world tie strength and demonstrate the high discriminatory power of related network measures. In combination with measures of similarity and interaction, the precision of identifying all observed real-world strong ties is 45%. Finally, we empirically highlight the practical relevance of this finding by demonstrating that people react stronger to suggestions of a close friend compared to an acquaintance in a social advertisement experiment.