General User Modelling: a Facility to Support Intelligent Interaction

A strawman conceptual architecture is proposed for knowledge-based systems that design their own graphical interfaces. The architecture is conceived of partly as an elaboration of the Seeheim model for user interfaces, but differs from it in several fundamental ways. The Seeheim model describes a class of fully realized, modular interfaces, without addressing how the interface itself is created. In contrast, we make explicit provision for an initial phase in which the system creates its user interface. Application domain-specific knowledge is used to influence interface design and to determine both the content and style of the information presented. Unlike the Seeheim model, we also suggest that media independence is not desirable in the architecture's early stages.Two testbed systems that we have developed are discussed in terms of the proposed architecture. APEX designs and renders sequences of pictures that illustrate how to perform simple actions in a world of objects. GRIDS lays out displays containing text and pictures with which it is provided. Each emphasizes different components of the architecture.