Online services heavily rely on the contributions of active and well-behaved users. Thus understanding online user behavior is important for the design, deployment and management of online services. However, user behavior is complex and difficult to understand. Traditionally, researchers study user behavior primarily using user studies including surveys and interviews. These are typically performed on small user populations, but can provide deep insights on the causes and motivations behind user behavior.
More recently, the problem space has been dramatically altered by the increasing availability of large datasets that capture user behavior, e.g. time stamped sequences of user actions in online social networks, or detailed server logs of user queries from search engines. The availability of such data provides new opportunities to observe user behavior at a wider scale. In this talk, I will present research on understanding online user behavior. I will discuss how researchers study user behavior patterns by leveraging large-scale datasets of user actions, how user studies help to explain the underlying mechanism of user behavior, and how such understanding is critical to improving user experiences in the physical world. Finally, I will discuss open research problems and potential future directions.