INFORMATION MAPPING IN PRACTICE Rule-based multimodal interface design

The design of a user interface can be considered an information mapping task. The task consists in mapping (a) information from the task domain of the artefact being designed as well as other relevant constraints on the design process, onto (b) a set of input/output modalities which, when fully specified, will constitute the user interface of the artefact. Interface designers thus essentially solve information mapping problems whether or not they think of themselves in those terms. This raises the question if it might be possible to provide knowledge-based support of the information mapping process. At a recent workshop at CHI’95 on “Knowledge-Based Support for the User Interface Design Process”, it was proposed that current human-computer interaction research adopt a weak sense of the term ‘knowledgebased support’ according to which not only rule-based AI systems but also systematically developed hypertext and hypermedia training material and sets of guidelines, walkthrough methodologies, and other similar approaches would count as knowledge-based support. In this sense, several examples of knowledge-based support of the information mapping process already exist. The domain of static graphic graphs is well explored [13,18,19]. Hovy and Arens [15] wrote a seminal paper which called for a more general approach to information mapping. The Namur group has worked on AI knowledgebased support of information mapping in the domain of business oriented applications for nearly a decade [14,20]. Over the last 2-3 years we have been developing an approach to this question, called modality theory. This overview paper presents the research agenda of modality theory (Sect. 2), introduces modality theory and the information mapping methodology (Sect. 3), describes recent results on knowledge-based support of the information mapping process (Sect. 4), and concludes by indicating some core questions for future work (Sect. 5).