IUI and agents for the new millennium (panel II)

Abstract

Advocates of intelligent user interfaces are used to fighting an uphill battle against more conventional approaches. Skeptics have been reluctant to accept intelligent tutoring systems, adaptive user interfaces, machine learning, predictive user models, anthropomorphic interaction, etc. as part of everyday interfaces because they have been suspicious of the feasibility of such techniques and fearful of the risk of possible mistakes. The good news is that we seem to be making progress in gaining acceptance. Past IUI conferences abound with examples of intelligent interface experiments that clearly demonstrate their feasibility. Limited examples of intelligent interfaces are actually starting to make their ways into commercial products. There is considerable evidence that opposition is softening. However, we'e not out of the woods yet. Many of the early examples of commercial IUI and agent software are positioned as "add-ens" to the more familiar directmanipulation interfaces, rather than playing a central role. We haven't yet reached the point where a new application is simply assumed, as a matter of course, to require all the representation, reasoning and learning features that IUl attendees advocate. But suppose we do? Suppose intelligence becomes such an integral part of the interface in the 21st century that we couldn't imagine applications without it? How will our software environment and the software industry change as a result? Will knowledge bases, inference engines, and learning algorithms become as much a part of the operating system as windows and menus? Will the idea of an "application", as a standalone, shrink-wrapped single-purpose interface, disappear'? Once the interface is intelligent, is there any point to having present-day concepts like "files" or "directories"? Will all interfaces become personalized to the extent that there won't be any more "generic" interfaces that remain the same across millions of users'! Will all information sources be interactive and customized, obsoleting paper books and linear movies'? Will that lead to a loss of shared context among users? How will different intelligent user interfaces interoperate and co-operate'? What, if anything, will be the next step beyond IUIs and agents? The panel will ask participants to speculate on how the widespread acceptance of intelligent user interfaces that we expect for the next millennium will transform our computing environments.