Robots that work close to humans need to understand and use social cues to act in a socially acceptable manner. Social cues are a form of communication (i.e., information flow) between people. In this paper, a framework is introduced to detect and analyse social cues and information transfer directionality using an information-theoretic measure, namely, transfer entropy. We demonstrate the framework in three settings involving social interactions between humans: object-handover, group-joining and person-following. Results show that transfer entropy can identify information flows between agents, when and where they occur, and their relative strength. For instance, in a person-following scenario, we find that head orientation of a predictor is particularly informative, and the different times and locations that this is used to convey information to a leader influences their behaviour. Potential applications of the framework include information flow or social cue analysis for interactive robot design, or socially-aware robot planning.