This short paper presents the work which has been carried out since 1989 in IRPEACS for developing both software tools and applications in the domain of adaptive hypermedia systems for education and training. IRPEACS is a research unit which is part of the French National Research Center (CNRS). It is a pluri-disciplinary research laboratory centred on Educational Technology. The researches address several inter-dependant dimensions concerning both initial and adult training, and particularly focus on : a) social,individual and collective practices of technologically mediated communication and training. b) production and uses of services of new technologies as "economic products" for training. c) human computer interaction , interfaces and software tools to support the design of multimedia products for training.
Since 1989 IRPEACS has been involved in four European Community DELTA projects (DELTA stands for Development of European Learning through Technology Advance).Two of these projects are still running : JITOL "Just in Time Open Learning" and DISCOURSE "Design and Interactive Specification of COURSEware". In 1991 , one of these projects (D1010 AAT "Adavanced Authoring Tools") produced "SHIVA" : a multimedia ITS authoring environment.
SHIVA resulted from an attempt to 'bridge the gap' beetween well understood CAL educational technologies and research tools in ITS (Intelligent Tutoring Systems) [Elsom-Cook 1990] by implementing the 'ECAL' system (the Open University , GB) as a mechanism for making pedagogical decisions within existing multimedia CAL (Computer Assisted Learning) materials. SHIVA cannot not be considered as an ITS because it is not capable of generating nor modifying its teaching materials. It only performs an adaptive sequencing of those materials. It fits with the definition of "one-on-one tutoring systems" according to G. McCalla [McCalla 1992] : SHIVA has access to a representation of the domain knowledge (a semantic network); it maintains a numerical overlay student model ; its tutoring strategy takes advantage of this student model to produce an adaptive curriculum of instructional activities. The domain knowledge is authored as a semantic network relating ILOs (Intended Learning Outcomes) of the domain, with ULMs (Units of Learning Material). The ULMs are "chunks" of teaching material stored in the base of teaching operations in the form of frames. There are two kinds of frames: presentation and problem frames. Both kind of frames may involve multimedia information (text , image , sound). ILOs may be linked with pre-requisit relations. Every ULM may be linked to several ILOs, thus making the ULM embedded task a multiple-concept task.
The system knowledge about the student knowledge is supported by counters associated to every ILO. The value of these counters reflects the extent to which the student has mastered the ILO. The system updates this student model every time when a ULM teaching operation has been completed. The update operation is done according to an assessment of the student activity based on some simple fixed rules.
The teaching strategy is to minimize the changing of teaching focus. The system decision making mechanism about "what to present to the student next" , is supported by a set of fixed rules taking into account the student model and the domain knowledge.
The result is an adaptive curriculum of teaching operations. There is no adaptive presentation within an ULM. There is no remediation of an ULM.
The SHIVA model, methodology and tools have been applied to two specific domains of teaching : Geography and Language. The SHIVA-Geography application concerned a complex knowledge domain : introduction to the systemic approach in geography with respect to the functioning of the water table. The SHIVA-Language application was aimed at providing a training tool in the use of commercial english in the Banking sector. The results showed that SHIVA was not suitable for developing courses with a strong procedural component, usually taught by drill-and-practice, and where there are multiple and hierarchical ways of categorising the domain. It was rather suitable for 'encyclopedic domains', where a large quantity of factual information is to be learned.A more general point concerned the fact that SHIVA incorporates a single interaction style - the system makes decisions about which declarative knowledge should be communicated - whereas learning in most complex domains requi
Since 1992 SHIVA has been re-implemented at IRPEACS and the current system runs now on Microsoft Multimedia Windows 3.1.
FFrom 1992 to 1994 it has been improved , enhanced and turned into a more generic and more open software tool which can be classified as a "Rapid Prototyping Tool to design Adaptive Navigation Support" in Hypermedia Systems for education and training (AHS). This new system is named PHASME. PHASME has been designed with the following aims and features :
In this approach, the learning environment is viewed as an Hyper-media system whose nodes are small software modules supporting learning activities. Some of the links are static (built-in), some are dynamically computerised at runtime. The former adpative curriculum sequencing problem of SHIVA has moved to a matter of adaptive navigation support.This support is provided by an intelligent agent (in fact a small multi-agent system) whose goal is to suggest the most relevant link to follow.
The PHASME approach is currently in the process of being applied to already built HyperMedia educational software in the domain of Chemical Engineering and Surface Analysis. The already built software provides some built-in user-driven navigational features, while PHASME proposes an overhead navigational support based on a student model and teaching strategies (one of them being of the SHIVA type).
We believe that such a system falls within the educational AHS category as defined by [Brusilovsky,1993].
The problems we have met fall within most of the topics that have been suggested for the workshop :
The PHASME architecture is concerned with topic 6. We believe that the attempt to apply our approach to an already existing courseware falls in topic 4. We have tried to combine the tutoring approach of SHIVA with a more exploratory mode of learning - an approach that we have called "a SHIVA assistant" which comes under topic 5. One of the main issues of this approach was how to maintain consistency in the student model when both the student and the ITS share the initiative in the sequencing of the learning task (the "sequencing" must be taken here as a navigational path). We had also to deal with the updating of the student model connected to the assessment of the student activity involving graphical interaction.
We suggest that the consideration of issues concerning the HCI navigational aspects of AHS, particularly the use of relevant metaphors might be worth an additional topic. As stated by [K.V Vaananan, J.Schmidt,1993] the hypermedia metaphor requirements will be fulfilled by three basic dimensions : a) organizational : the metaphor manifests inherent structure that can reflect the node-and-link information. b) functional : the metaphor is represented through visually recognizable objects that allow direct manipulation. c) navigational : the metaphor suggests ways to move around in the hyperspace. Navigational metaphors include : guided tour, organized travel, maps... Considering the inherent difficulty of navigation in an hyperspace and the possibility of some adaptive navigation paths provided by the system (as an help in the learning process), we must not forget the issue of the usability of such a system. Is it not useful to find good navigational metaphors appropriate to AHS ?
Bessiere C.,Leonhardt J.L.,Zeiliger R.,1991, Multimedia authoring tools : atelier ORGUE, in Computer Networks and ISDN systems 23 pp 157-161, North Holland
Brusilovsky P.L., 1992, A Framework for Intelligent Lnowledge Sequencing and Task Sequencing, in Intelligent Tutoring Systems, proceedings of ITS'92, pp 499-506,Springer-Verlag.
Brusilovsky P.L.,Pesin L.,Zyryanov M.,1993, Towards an Adaptive Hypermedia Component for an Intelligent Learning Environment, Proceedings of EWHCI'93 East-West International Conference on Human-Computer Interaction, pp 27-38.
Cox,R.,Cumming,G.,1990,The role of exploration-based learning in the development of expertise,Proceedings of Computers in Education,pp 359-364, Elsevier Science Publishers B.V.
Elsom-Cook,M.,1990, ECAL : Bridging the gap between CAL and ITS, M. Elsom-Cook and C. O'Malley, in Computers & Education, 15, p 69-81.
Elsom-Cook,M.,1992, Belief-based bounded user modelling, in Learning Technology in the European Communities,pp 423-436, Cerri and Whitings eds, Kluwer Academic Publishers.
Participants of Delta AAT project,1991,in DELTA AAT Final Report ,annexe 4 : SHIVA,CNRS-IRPEACS 93 Ch des Mouilles 69130 Ecully, France.
Psotka,J.,Massey.L.D.,Mutter,S.A.,1988,Intelligent tutoring systems : lessons learned, Lawrence Erlbaum Ass Publishers.
Kobsa,A.,1994,User modeling and User-adapted Interaction,Tutorial at CHI94, Conference on Human Factors in Computing Systems,ACM press.
McCalla,G.I.,1992,The search for adaptability,flexibility and individualization : approaches to curriculum in ITS, in Adaptive Learning Environments : foundations and frontiers, NATO ASI Serie F, vol 85, pp 91-121, Berlin : Springer-Verlag.
Self,J.,1992, Artificial Intelligence: the key to open learning, in Learning Technology in the European Communities,pp 341-354, Cerri and Whitings eds, Kluwer Academic Publishers.
Zeiliger R.,1993,Adaptive Testing : contribution of the SHIVA model, in Item Banking : Interactive testing and Self-Assessment, edited by D.Leclercq and J.Bruno, NATO ASI Serie F Vol 112,pp 54-65
Zeiliger R.,1991, SHIVA : manuel de reference, Delta D1010 pp 1-115, CNRS-IRPEACS 93 Ch des Mouilles 69130 Ecully, France.