Abstract Our high-level goal is to answer questions concerned with social influence such as: “Who influences whom?”, “Who can be influenced?”, “Why is an individual attracted to a particular group?”, and “Who is the most influential individual in a particular social network?”. To ask these questions we need to define social influence. In this paper we provide a formal definition appropriate to our world of Big Data and automated reasoning. Despite the pervasiveness of influence throughout society and given the vast and disparate literature on the topic, we observe a dearth of work on formalising its semantics. To remedy this, based on the literature, we have categorised and formalised five essential types. To our knowledge this is the first attempt to implement a nuanced representation, and it provides us with a conceptual basis for automated reasoning about social interactions.