Volume 7, Number 3/4 1996
Tom Conlon and Helen Pain 219
Wouter R. van Joolingen and Ton de Jong 253
Ana Arruarte, Isabel Fernández-Castro, and Jim Greer 277
Steven Ritter and Kenneth R. Koedinger 315
Gilbert Paquette, Francois Pachet, Sylvain Giroux, and Jean Girard 349
Abstracts
Persistent Collaboration: A Methodology for Applied AIED
TOM CONLON
Moray House Institute of Education, Heriot-Watt University Holyrood Road, Edinburgh, ScotlandHELEN PAIN
Department of Artificial Intelligence, University of Edinburgh 80 South Bridge, Edinburgh, ScotlandWithin the artificial intelligence in education (AIED) community there are clear signsof dissatisfaction with existing research approaches. We suggest that AIED projects are ofvarious kinds and that the research methods that are appropriate will vary betweenprojects. Our focus is on projects in the area of "applied" AIED-those that aimto produce practically useful classroom tools. Following a review of studies of researchin education, theories of educational change, and approaches to computer systems design,we propose that the most appropriate methodology for applied AIED is one that is derivedfrom a combination of action research with user-centred design. We present our version ofsuch a methodology, which we have named Persistent Collaboration Methodology (PCM) onaccount of the central role that it gives to collaboration among researchers, teachers,and technology experts. The methodology, which we illustrate at length by reference to anongoing project in the area of knowledge-based modelling, appears to offer a suitablybalanced marriage between the "technology push" and the "learningpull." Because of this, and because it also supports a productive relationshipbetween theory and practice, the use of our methodology should enhance the prospects ofindividual projects and of the AIED field in general.
Design and Implementation of Simulation-Based Discovery Environments: The SMISLESolution
WOUTER R. VAN JOOLINGEN AND TON DE JONG
Faculty of Educational Science and Technology University of Twente, P.O. Box 217 7500 AE Enschede, The NetherlandsThis article describes the design of interaction with the learner in simulation-baseddiscovery environments. This interaction is by definition of a mixed initiative: Indiscovery learning it is the learner who takes the lead, but also the learning environmentshould take an active role in shaping itself on the basis of monitoring the development inthe simulation and the actions by the learner. The SMISLE system, which is an authoringsystem for integrated simulation-based learning environments, aims at offering designersof learning environments a generic solution for defining the structure of the learningdialogue. The solution chosen is one of integrated, object-oriented design. This meansthat a designer can specify elements of the dialogue ("instructional measures")taken from a generic library and alter their characteristics. Moreover, the designerspecifies for each instructional measure under which conditions it may become active, thusspecifying control over the complete learning dialogue. The current article describes theway instruction is designed in SMISLE and presents five applications that were createdusing the SMISLE system.
The CLAI Model: A Cognitive Theory of Instruction to Guide ITS Development
ANA ARRUARTE AND ISABEL FERNÁNDEZ-CASTRO
Department of Computer Languages and Systems Computer Science Faculty University of the Basque Country UPV/EHU 649 Postakutxa, 20080 Donostia, SpainJIM GREER
ARIES Laboratory, Department of Computer Science University of Saskatchewan Saskatoon, Canada S7N 0W0In this paper, the authors present a pragmatic cognitive theory of instruction, theCLAI Model (Cognitive Learning from Automated Instruction), which is intended to be atheory practically useful for developing real instructional systems. Their main goal wasto embed this theory in an automatic instructional system in order to help initiate in thelearner the appropriate cognitive processes to acquire new knowledge. This theoryintegrates cognitive processes, instructional events, and instructional actions within athree-level framework. The three levels are connected by several relationships adoptedfrom various successful instructional systems and based on educational instructionaltheories and empirical studies. Moreover, the authors define a set of supported actionsthat will be used by the learner as additional ways of initiating the cognitive processes.The approach is flexible enough to modify, adapt, or amplify the existing links among thethree levels.
The theory is intended to serve as a guide to define and restrict the pedagogicaldecisions to be considered in a tutoring system. However, in order to build an automaticsystem, it is necessary to support the theory by a central operative kernel able to selectand execute different possible instructional actions.
An Architecture For Plug-In Tutor Agents
STEVEN RITTER
Department of Psychology, Carnegie Mellon University Pittsburgh, PA 15213, USAKENNETH R. KOEDINGER
Human-Computer Interaction Institute School of Computer Science, Carnegie Mellon University Pittsburgh, PA 15213, USAThis paper outlines the authors' efforts to build new learning environments thatincorporate tutoring elements into pre-existing software packages. Two systems aredescribed; one provides tutoring support in the Geometer's Sketchpad and the othersupports students using Microsoft Excel. Although the implementation of these two systemswas somewhat different, they share many basic components. An analysis of theirsimilarities and differences allows us to move toward a set of standards for tutor agentsthat interact with complex tools. By constructing learning environments in this manner, wecan leverage the power of existing workplace software and educational microworlds tocreate more powerful learning environments.
ÉpiTalk: Generating Advisor Agents for Existing Information Systems
GILBERT PAQUETTE
LICEF, Télé-université 1001 Sherbrooke East Street Montréal H2X 3M4, CanadaFRANCOIS PACHET
LAFORIA, Université Paris 6, E.4 Place Jussieu, 75252 Paris, cedex 05, FranceSYLVAIN GIROUX AND JEAN GIRARD
LICEF, Télé-université 1001 Sherbrooke East Street Montréal H2X 3M4, CanadaAdvisor components play an important role in intelligent tutoring systems and learningenvironments, as well as in electronic performance support systems and help components ofcommercial applications. A complete advisor should go beyond mere contextual help wired inan information system. It should achieve good balance between the user's and the advisor'sinitiative. It should be able to offer suggestions at different levels of abstraction,from very domain-specific suggestions to more abstract generic problem-solving methods.Finally, it should use various viewpoints on the user's tasks and support collaborative,as well as individual, activities. Indeed, this is not trivial.