Intelligent systems should be able to initiate and understand social interactions, which are always brought about by physical actions. Thus, all physical actions need to be interpreted for their social effects. This paper gives an analytical framework that identifies the main challenges in modelling social interactions. Such interpretation is subjective and might lead to misunderstandings between participants in the interaction. In real life, people use rituals, such as greetings, as a method for common grounding. Greeting as an essential prerequisite of joint activity is taken as case study to illustrate our approach.