Word of Mouth Dynamics in Online Social Networks: Investigating Social Influence Cascades on YouTube

Introduction and Research Question With its user-friendly interface and the growth in popularity of online video, YouTube has catapulted to a dominant position on the Internet. While this model is extremely attractive for marketers and content creators, recent work has recognized the relatively ephemeral nature of popularity of videos on YouTube, where only a tiny fraction of videos managed attract most views. This makes it very important for researchers as well as practitioners to develop a deeper understanding of user behavior and the role of interpersonal influence on YouTube. One of the characteristics of YouTube is the ease of formation of social ties such as friendship and subscription networks. YouTube also offers several avenues for user interaction through commenting on and favoriting videos. Thus, another characteristic of YouTube is the publicly observable nature of interpersonal communications (through actions such as commenting) in addition to the social structure of interactions (through observable ties such as friendship and subscriptions). This paper explores how user interactions through commenting actions early in the life of a video generate cascades of word-of-mouth (WOM) communications that influence subsequent popularity of videos. Cascades are defined as topological patterns in information propagation graphs (Leskovec et al. 2006). In the YouTube context, word of mouth arises endogenously from patterns of interaction structured through a network that individuals are embedded in. Depending on the network topology, WOM from the local network travels beyond the immediate local network neighborhood, contributing to large-scale popularity of a video, similar to models of cascades in social networks (e.g., Watts 2002). We draw upon a stream of research in physics (e.g., Watts 2002) and computer science (e.g., Leskovec et al. 2006) to conceptualize the micro foundations of WOM generated by patterns of interactions in a cohesive group. The two questions we explore are: • How does network structure and network position impact the formation of WOM communication early in the life of a video? • How does WOM communication early in the life of a video impact subsequent popularity?