This document discusses several research problems related to analyzing social communication in online media: 1. The first problem examines why people repeatedly view the same YouTube video and proposes that it is the interesting conversations around the video, not just the video itself. An approach is presented to model conversational themes and interestingness over time. 2. The second problem introduces a framework for predicting "social synchrony" in social media by learning user actions and evolving models of the social network over time. 3. The third problem examines how to characterize communication at both the individual and group levels in order to extract and analyze prototypical social groups within a topic in the blogosphere.