Interactive systems are complex systems that allow operators to control and monitor other systems. The interactive systems offer one or several user interfaces, composed of hardware and software components, which are used by the operators to perform their tasks. The complexity of interactive systems lays in its nature (hardware and software integration) but also in the rapidly evolving technology (new input and output devices, new interaction techniques …). This paper argues that these interactive systems are the very place where human-system integration takes place. It is thus important to design, develop, test, certify and deploy them very carefully. Unfortunately, methods, techniques and tools from software and system engineering need deep tuning to be adapted to their characteristics. The paper presents a generic, customizable interactive system architecture and highlights how its components relate to the human operator. It also presents a taxonomy of possible faults that may degrade the behavior of the various components. We demonstrate that these two contributions can be jointly used to systematically identify possible faults (both in the operator and in the system) and proposes mechanisms to prevent, remove or tolerate them.