The role of the system image in intelligent user assistance

Many researchers have demonstrated the ways in which well-designed graphical interfaces allow users to acquire conceptual models of the interfaces and the application programs behind these interfaces. It is also clear that the users’ models are initially flawed and incomplete, and the problems that users have with these systems revolve around the misconceptions and alternate conceptualizations in these models. Our work indicates that graphical interfaces may be especially sensitive to misconceptions, and that advisory systems for these kinds of systems must anticipate and resolve such problems. In particular, they must be able to understand alternative conceptual models of the system, and may need to diagnose and remediate these misconceptions. We will describe our work on direct manipulation interfaces and intelligent advisory systems, focusing on the problems people encounter when using interfaces and the ways in which our advisor's design is being driven by the properties of graphical interfaces.