Using Big Data and Agent-Based Modeling to Understand Social Media Diffusion

With the increasing abundance of `digital footprints' left by human interactions in online environments, e.g., social media and app use, the ability to model such behavior has become increasingly possible. Many approaches have been proposed, however, most previous model frameworks are fairly restrictive, and often the models are not directly compared on a diverse collection of human behavior. We will explore a new modeling approach that enables the creation of models directly from data with no previous restrictions on the data. We develop three non-parametric models for exogenously-driven, self-driven, and socially-driven behavior in digital social networks, and compare their predictive and descriptive abilities on a heterogeneous catalog of human behavior collected from fifteen thousand users on the microblogging platform Twitter over the course of a year. We find that despite the popularity of renewal process-type models, based on exogenous drives, for explaining digitally-mediated human behavior, most users are better modeled using self- or socially-driven models. Our work highlights the importance of a flexible modeling approach when attempting to explain and predict human behavior in digital environments