JAIED v7n2

The three papers comprising this special issue of the Journal ofArtificial Intelligence in Education are based, on edited transcripts of invited talksgiven at the AI-ED 1995 Conference in Washington, DC. It seemed to us that the invitedpresentations at conferences are often the most influential in affecting the directions ofa research field, and yet they often do not exist in a referential form. For example, KurtVan Lehn's talk at AI-ED 1993 is often quoted via the short paper in the proceedings,which in fact, differed substantially from the talk.

For these papers, we asked the speakers-cum-authors to resist thetemptation to embark on a wholesale revision of the transcripts. We wished to retainsomething of the style, the informality, and the spontaneity of a talk, as opposed to afinely honed paper. We have deliberately left in various comments and asides which willremind readers that it is indeed a transcript, not a normal paper, they peruse. We askreaders to read the paper in this spirit (and not to quote any slight infelicity ofexpression).

I would like to thank the three authors-Tak-Wai Chan, PierreDillenbourg, and Johanna Moore-for allowing their talks to be published in this way. It issometimes not easy to face up to a transcript of what one has really said especially,perhaps, for those such as Tak-Wai and Pierre for whom English is a second language. Ofcourse, these written versions lose something of the vitality of the presentations,especially without the visual demonstrations. Nonetheless, we are sure this issue providesa record of events which have been and will be influential in the AI-ED field.

The Editor

Return to Contents

Learning Companion Systems, Social LearningSystems, and the Global Social Learning Club

Tak-Wai Chan

Department of Computer Science and Information

Engineering

National Central University, Chung-Li

Taiwan, 32054 R.O.C

This talk reviews our work in the past and presents a vision of futureresearch based on our current approach. I take a rather personal perspective in describinghow the idea of the learning companions emerged. Then I discuss how, after the developmentof a series of experimental systems over some years, the idea of the learning companionswas extended to a class of systems which we call social learning systems. Failures orsuccesses of some of these systems are discussed. After that, I describe the recentreadjustment of our approach and take a leap to aim at a global social learning club. Itis becoming clear that such a virtual learning community will be typical or prevalent inthe next century.

Return to Contents

Some Technical Implications of Distributed Cognition on the

Design on Interactive Learning Environments

Pierre Dillenbourg

TECFA, Faculte de Psychologie et des Sciences de l'Education

University of Geneva, 1227

Carouge, Switzerland

Edited transcript of "Diagnosis is mutual: A distributed cognitionapproach," an invited talk at the World Conference on Artificial Intelligence inEducation, Washington, DC, August, 1995.

This talk prompted two arguments. First, I contend that the three mainfunctions in an interactive learning environment are intrinsically collaborative. Thefunctions are: (a) Learning modeling is a collaborative process. This is not a new idea,but rather the natural evolution of research on learning modeling; (b) explanation andeven (c) tutoring are described also as collaborative processes. My second claim is thatcurrent knowledge-based techniques are not appropriate to develop collaborative agents. Isuggest that collaborative engines should merge dialogue operators with reasoningoperators. For instance, dialogue operators could be used during the instanciation ofrules in order to enable agents to negotiate the scope of variables.

Return to Contents

Making Computer Tutors More Like Humans

Johanna D. Moore

Learning Research and Development Center

University of Pittsburgh

Pittsburgh, PA, 15260, USA

If computer-based instructional systems are to reap the benefits ofnatural language interaction, they must be endowed with the properties that make humannatural language interaction so effective. In this paper, I describe several features thatdistinguish human tutorial explanations from those produced by today's computer-basedtutors. I argue that in order to build systems capable of structuring explanationsnaturally and appropriately interpreting and generating context-sensitive utterances, itis necessary to model and reason about the plans the speakers in a discourse areattempting to carry out. I then present a plan-based approach to natural languagegeneration, and provide examples from several implemented systems that illustrate how theapproach enables systems to participate effectively in natural language dialogues withtheir users.

Return to Contents