Expectation-based user interaction

Multimedia and multimodal interfaces reflect the growing technological possibilities of computer-based systems for interaction with the user. The ongoing increase in communication bandwidth and the growing variety of communication channels enable further improvement in the user interface. However, how this increased communication capacity can optimally be exploited is as yet unknown. Since the functionality of these computer-based systems also continues to grow, the increased complexity of interaction procedures and the difficulty of mastering them are prime issues in the design of "easy to use" multimodal user interfaces. In order to appreciate more fully what is involved in self-evident and at the same time efficient interaction between user and system, we will first briefly describe the layered-protocol model of computer-human dialogue as proposed by Taylor (1988a). This conceptual framework emphasizes the relevance of layered feedback for the efficiency of communication. As indicated by Engel & Haakma (1993), in particular early feedback about the system's interpretation of the message part already received (I-feedback) as well as on machine expectations about message elements still to be received (E-feedback) are of relevance for the system's ease of use. Thereafter, as an interesting example of improved human-computer interaction through layered multimodal I- and E-feedback, an experimental trackball device will be described. It provides the user, in addition to the standard visual I-feedback about the current cursor position, with tactile E-feedback about the expected cursor target position. Lastly, our running experimental exploration of the possibilities for automatic cursor-endpoint prediction will be described, this research being of relevance for the further improvement of interaction with the mentioned trackball device with expectation-based force-feedback.