UC Riverside

This dissertation focuses on understanding user behaviors from online digital footprints like online social networks, social media, and smartphones. My central hypothesis is that studying online activity of users allows us to analyze and detect interesting real-life events, creating a bridge between online and offline activities. I observe users at three levels: group level, individual level, and interaction network level. My approach to developing a comprehensive understanding of these aspects in this dissertation is essentially computational as well as empirical: I present computational models, characterization techniques, and finally discuss large-scale quantitative observational studies to understand users at each of the levels.

Understanding user behaviors has always been of interest to researchers from multi-faceted domains over the past few decades, especially in a social and psychological context. Present day has several modern capabilities encompassing a host of social media websites. These sites feature variegated interactional affordances, ranging from blogging, sharing social media elements as well as a rich set of social actions such as liking, commenting, tagging and so on. Consequently, these communication tools have begun to redefine ways in which individuals interact with one another and how we study their societal behavior.

There are three main research components in this thesis and I present the contributions in three chapters. First, we observe that the group behavioral dynamics of users that follow a skewed distribution with long tails and their online activity can be characterized using the PowerWall distribution. Second, we study the temporal timeline of user activity to observe changes in time. While it is obvious that different users behave differently, we present the iNᴇᴛ methodology that quantifies change to identify interesting behaviors in time. We present how several exogenous influences like elections and sports affect user activity. Third, we present MIMiS, a framework for characterizing recurrent behaviors in smartphone users while preserving their privacy. We are able to identify groups of these users with similar mental health states like depression and stress and how it correlates with their academic performance. In a nutshell, this research makes significant contribution into a better understanding of users, and bridging the gap between online and offline human behaviors.