Computer systems supporting complex human activities or complex processes require user interfaces that can build and maintain explicit models of the user, the process, and the system itself.
Such interfaces could adapt to the user, and could provide focused advice and guidance support at any point during the process. We qualify such adaptive user interfaces as “intelligent”.
Benefits which are expected from intelligent interfaces include a drastic reduction in the learning time required to efficiently use the environment, and efficient teaching of how to conduct the process for novice users.
In this article, we propose a generic architecture for designing intelligent user interfaces1
The main components of this architecture are:
a task scheduler
an activity selector
an activity configurer
a process context modeler
the interface with the domain process specific modules and functions.
These components and their organization are described.
An implementation of an intelligent user interface using this architecture will be presented. This user interface was built for the Knowledge Engineering Workbench (KEW), developed within the Esprit project ACKnowledge [MSW89]. KEW is a workbench that integrates complementary tools and techniques dedicated to knowledge acquisition.