Software agents have been introduced into information systems in recent years. As the capabilities of software agents become more complex, it becomes more difficult to understand and design such systems. In order to make these agents behave as they are supposed to so as not to infringe on the rights of their human users, it becomes necessary to consider the social context in which the agents have to operate. This social context consists of the human actors, the social constraints and institutional facilities associated with them, such as roles and norms. This calls for new methods which consider these factors that affect the proper use of agents in information systems. This paper aims at presenting a semiotic approach to the modelling and design of agent-based information systems. It has made explicit links between roles and norms, which aids a successful design of agent-based information system.