Decoding human behaviors from social media

Thesis (Ph. D.)--University of Rochester. Department of Computer Science, 2019.Given its large user base, deep user engagement level, and comprehensive coverage on different human activities, social media offers us a novel and effective lens for observing, analyzing, and gaining better understanding on human behaviors. Our work leverages the data collected from various social media platforms, and attempts to decode human behaviors at both individual and collective levels. From a mobility perspective, we study the relations between human mobility and other aspects of our lives, such as shopping patterns and circadian rhythms. Furthermore, we draw an analogy between human activities across online communities and movements in the physical world. The analogy leads us to a series of findings on the striking similarities between the physical and cyber spaces, and suggests promising new research directions. From a language perspective, we investigate how people from different social groups express themselves. The discovered divergent language patterns reveal the unique features of social groups, such as their interests, perception, personality traits, and so on. We then broaden our study on language to emoji, a type of commonly used non-verbal cue in Computer Mediated Communication, and reveal intentions and effects of using emojis. Our work validates the effectiveness of using social media as the data source for decoding human behaviors. More importantly, our work reveals insightful findings on human behaviours in various aspects, and sheds newer light on our own lives