Maximizing Influence Through Information-Overloaded Online Social Networks

Online social communities have become an important communication channel for people to share and discover information. Pieces of information spread within the community via the underlying social network, from one individual to another. However, with the unprecedented ease and low cost of communication provided by online systems, information overload emerges as a negative factor that potentially threatens the effectiveness of communication. As such, traditional single-message based diffusion models, such as Independent Cascade Model (ICM), are inadequate for describing the role of information overload. We then proposed an extended version of ICM (EICM) that explicitly takes the message multiplicity into account, after examining the message exchange patterns observed from a real online social community. We extensively evaluated this new diffusion model and compared with standard ICM in addressing one fundamental algorithmic problem related to viral marketing: How to select a set of network nodes/individuals to facilitate information diffusion and maximize influence? The evaluation results obtained from using both real and simulated data sets show that our approach results in node-selection heuristics outperforming well-studied notions of various centrality measures based on the ICM. The study and findings presented in this research have important managerial implications. In particular, from a viral marketing perspective, information overload effect should be recognized in order for campaign managers to build advantages in their influence maximization decisions.