In this paper, we explore the use of abductive reasoning to control knowledge base interactions. This control is achieved by means of the maintenance of dynamically changing knowledge bases. Such knowledge bases occur in interactive systems, where they contain information about the changing context of the interaction or dialogue. Abduction is used in a number of ways. It is used to assimilate or explain new information and update the context accordingly. This is already a well-known application of abduction. In addition, it can be used to perform various kinds of reasoning such as default reasoning, and planning of the next knowledge state, via the use of additional predicates such as "abnormal" and "anomalous". The paper contains an explanation of the reasoning patterns involved and an extended example relating to dialogue management in the ESPRIT PLUS project.