Human behavior is highly complex and dependent on the interplay of various intrinsic and extrinsic variables making its study extremely challenging. In the era of big data, exponential growth of data across multiple domains has provided a unique opportunity for us to study human behavior, actions and relationships at scale by combining data from diverse sources.
This dissertation is guided by the following principles: 1) leveraging rich situational and interaction context 2) fusing information from heterogeneous data sources using efficient and effective computational models and 3) studying human behavior across multiple application domains to draw actionable insights to address real world challenges. To that end, the goal of this dissertation is to design data driven, context aware and user centered computational models that help us to understand, analyze, model and infer a range of individual and collective human behavior aspects from preference to engagement; from personality traits to political behaviors; from public opinion to relationship formation.
We primarily study three aspect of human behavior: 1) preference, engagement and recommendation 2) personal traits, sentiment and opinions and 3) opinion and network evolution. In particular, we ask the following questions: How do personal factors such as age, gender and location impact content preference, engagement and consumption? How personal attributes, physical and social context guide individuals to navigate information, select friends, and grow their social networks? How do individuals express themselves? How do opinions evolve over time? How are friendships formed? By answering such questions, this dissertation aims to move one step closer to understanding human behavior.