From: AAAI-86 Proceedings. Copyright ©1986, AAAI (www.aaai.org). All rights reserved. WHERE SHOULD IN INTELLIGENT THE INTELLIGENCE INTERFACES Robert Thomas USC / Information BE PLACED? Neches S. Kacemarek Sciences Institute 4676 Admiralty Way Marina Del Rey, CA 90292 Phone: PURPOSE Obviously, many human-machine language techniques solving, text understanding paradigms in intelligent seek to provide intentions have based emphasiee hands of the user. with, and the AI techniques If our understanding important detail. interaction more purpose paradigms specific alternative intelligent issues of in two directions: within interfaces and to What these of l the focusing kinds of in greater of to bring out by systems The kinds are the fundamental to of users l inputs handled so that knowledge, or by building systems could generate reasoning and behavior as l is built into knowledge- been interface chosen design. of interaction. represent Each There for which is a description to systems a wide has developed is also great the panelists have of the experience that range diversity created and of a uniquely in the interfaces. interests of the John Seely Brown has a long history interaction systems of involvement dating back with to his work support for communicative interaction. Phil Hayes has been involved repair with he is currently interfaces and in user-system the construction of cooperative user interfaces. His work on COUSIN is a pioneering effort in the construction of user interface management systems. It is a prime example of the a particular users can get by with less in more self-knowledge so that justifications of their own done in expert systems. that the Operability Project at Xerox PARC, concerned with the intelligibility of addressing? Correcting errors? Keeping track of the interleaved activities? how to use a system to accomplish those on SOPHIE in the early ’70s. Combining his interests in the semantics of procedures and his collaboration with on be goal? What are the general approaches towards addressing those problems ? For example, one might approach the understandability issue by trying to broaden the range of style intelligent including that realize panelist: will address include: problems have with of systems Following systems. panel members can the intelligence panelists experiences presenting attention How different it is discussion first, by trying and them. help Are there other kinds of knowledge-based have distinctive interface needs? in the grow, such distinctions by ought intelligent interfaces Understanding the system? Identifying referents of discourse? state of plans and goals during Planning is to paradigms second, a variety to helping techniques users need to build confidence in the system and to learn from it. Are there other needs or other consequences? set, but leave that users need help AI concerns that must be addressed when interfacing to intelligent systems ? Many expert system designers have added explanation capabilities based on the belief that tool” are some very different particular based systems be exploited to provide a better user interface? It would seem that sharing intelligence between the interface and the system would reduce cost and enhance consistency. Is this true? Are there special Consultant those commands is to extend l different “power command interfaces to examine models, of the user’s hand, the problems intelligent Thus, the kinds of questions l a powerful of this panel paradigms, knowledge-based other can having to learn a command language”, but rather at the level of particular solutions in particular systems. the knowledge like the UNIX that can be applied that we continue The the better particular On the one hand, and applying about The How approaches? This question is not to be answered at the cliche level of “natural language would save users from and natural led to two very Behind these distinctions assumptions systems, on an understanding On for selecting building have explored generation. have l search, discourse expert systems domain. systems like COUSIN the responsibility fundamental and apprentice” task heuristic chosen to interfaces interfaces. assistance and applicable representation, researchers “intelligent are Intelligent planning, knowledge techniques based, OF THE PANEL interfaces. use of problem user models, AI (213) 822-1511 l power tools approach of interface design. His current interests include interfaces to knowledge-based systems and combining natural language with other command modes. Tom Malone’s interests have been aimed at organieational needed in Information computing and the kinds of interfaces such environments. His work on the Lens exemplifies the power tools approach. He also influenced RABBIT, which is a blend of the power tools and intelligent assistant approaches. His current interests involve understanding the impact on the interface that results from the distinctions between expert support systems contrasting naturrll intelligence. INVITED and expert intelligence TALKS systems with AND and also in artificial PANELS / 115 1 l Dick Waters has contributed programming environment Programmer’s Apprentice significantly through his and KBEmacs. defined in this work combines work the The interface combines both and separating those aspects of the environment to provide for from those to provide tools for. l the on aspects of power tools and intelligent assistance. Since the interface aspects, there is an interest in discovering l to assistance Bob Wilensky has worked extensively on natural His work on the UNM Consultant language interfaces. The is a prime example of an intelligent assistant. emphasis in this work has been on the naive user. Mike Williams has contributed in the development of intelligent interfaces to data bases that do not involve The descriptive retrieval paradigm natural language. found in the RABBIT system employs an interface that that the user assistant intelligent presents an communicates with through a set of power tools. It provides the feeling of direct manipulation, power tools, but uses an understanding to provide the power. 1152 / INVITED TALKS AND PANELS as found in of the data base