Content uploaded by Dietrich Albert
Author content
All content in this area was uploaded by Dietrich Albert
Content may be subject to copyright.
S. L. Wong et al. (Eds.) (2010). Proceedings of the 18th International Conference on Computers in Education. Putrajaya, Malaysia:
Asia-Pacific Society for Computers in Education.
Towards Generic Visualisation Tools and
Techniques for Adaptive E-Learning
Dietrich Albert
a
, Alexander Nussbaumer
a
, Christina M. Steiner
a
a
Department of Psychology, University of Graz, Austria
dietrich.albert@uni-graz.at
Abstract: In this paper, we describe the work on the visualisation tools and techniques
currently developed in the GRAPPLE research project. Since the GRAPPLE project aims at
developing a generic solution for adaptive e-learning, also the visualisation tools need to be
as generic as the GRAPPLE approach and its data models are. This paper also discusses
related work in this field and outlines differences and advantages of the newly developed
visualisation tools and techniques. Especially comparisons are made with a tool which relies
on visualizing competence-based knowledge structures.
Keywords: adaptive e-learning, open learner model approach, visualisation technique,
meta-cognitive support
Introduction
Visualisation has often been used in e-learning to stimulate meta-cognition by providing
feedback to the learners about their learning process. In order to provide visual information
to learners, learner models are needed which can be used as basis for this information. This
approach is well known as Open Learner Model (OLM) approach and has been often
described in literature (for example in [4] and [8]). In adaptive e-learning systems user and
adaptation models are used to achieve adaptation of learning resources to the learners'
characteristics. Traditionally, these models are not revealed to the learners, but used for
adaptation algorithms only. Especially adaptive systems make use of user models which
often are available at a detailed level. Therefore, opening up these models can provide rich
information which can support the learner in his or her self-reflection activities.
Obviously, presenting user model data to the learner highly depends on the available
information of a learning system and how this information is organised and structured.
Based on that data visualisation techniques are be employed to make accessible the user
models for the learner. Making understandable the presented data is a key factor for
achieving reflective and meta-cognitive activities of learners.
This paper presents a new visualisation approach which aims at being more generic
than existing visualisation strategies in adaptive e-learning systems. This approach is
currently developed in the context of the GRAPPLE (Generic Responsive Adaptive
Personalized Learning Environment) research project [6] which aims at delivering generic
adaptation functionality for various Learning Management Systems. To this end, flexible
learner models are used for adaptation functionality. Therefore, also the visualisation
technique has to cover the same flexibility, which brings new possibilities to the learner
regarding the visually accessible information.
S. L. Wong et al. (Eds.) (2010). Proceedings of the 18th International Conference on Computers in Education. Putrajaya, Malaysia:
Asia-Pacific Society for Computers in Education.
1. Visualisation in Adaptive E-Learning
In order to outline some of the key features of visualisation approaches in adaptive
e-learning systems, an example from previous work in this field is introduced in this section.
The visualisation tools for making visible and accessible learner models based on
Competence-based Knowledge Space Theory (CbKST Tools) [3][9] has been developed in
the research project iClass [7]. Instead of providing a general overview of visualisation in
adaptive e-learning and OLM, the newly developed approach is compared with the CbKST
Tools, which reveals the advantages, but also limitations of the GRAPPLE visualisation
approach.
The CbKST Tools are based on a domain and user model which both follow the
mathematical-psychological approach of Competence-based Knowledge Space Theory [2].
The central element is the set of skills and prerequisite structure on skills which arise due to
psychological dependences (see Figure 1). In order to model a domain, the skills necessary
to cover this domain are modelled and the structure of these skills is identified. Learning
resources are associated with skills, whereas learning objects can convey skills and
assessment items can test skills.
The user model relies on the domain model and can express different user
characteristics. Goals can be defined as a set of skills which should be attained. The skill or
knowledge state of a learner can also be defined as a set of skills which the learner already
has available. This is outlined in Figure 1 with red circles in the skill structure. The learning
history can be shown either as sequence of learning resources or as sequence of the skills
associated with the performed learning resources.
Figure 1: Prerequisite structure on skills (a) and possible competence state of a learner (b).
The logical structure of the prerequisite relations of the skills can be depicted as
acyclic directed graph with the special property that transitive relations are not drawn. In the
example in Figure 1 the skills below other skills connected with a line are prerequisite for
those skills. For example, skills S1 is prerequisite for skills S3, but also for skills S4, S5, and
S6. This structure can be visually displayed to the learner as it is shown in Figure 1. In this
way the conceptual part (skills, but not associated learning resources) are visualised and
opened up to the learner.
The user model consisting of goals, skill state, and learning history is directly depicted
on the skill structure. Since all of these elements are related to skills, the respective skills
can be marked and highlighted (Figure 1b). In this way the learner is always presented with
the same structure (for a specific domain), but the user model values are changing on this
visual structure.
In addition to use these visualisations as display for domain and user model values, the
same visualisation can be used to guide the learner though the learning process. Since it is
meaningful to sequence the skills to be attained according to the prerequisite structure, a
S. L. Wong et al. (Eds.) (2010). Proceedings of the 18th International Conference on Computers in Education. Putrajaya, Malaysia:
Asia-Pacific Society for Computers in Education.
learning path reveals as learners should start with easier skills at the bottom of the structure
and continuing with higher levelled skills. Learners see their skills state easily, so they can
choose skills one level higher as their available skills. The visualisations have been made
interactive, so that by clicking on a skill, the associated learning resources are offered to the
learner. Hence, the learner gets interactive navigational support by using this visualisation.
Using that visualisation technique learners are supported to perform meta-cognitive
aspects on their own learning process. They can set goals by picking skills, they can make
plans by choosing learning resources associated with the goal, and they get feedback and
orientation about their current learning state and progress.
Other approaches also take into account group work and collaboration and reflect
them in visualisations. For example the technique presented in [8] aims at mirroring the
activity of small teams engaged in a task. Each individual is contributing to the group and
the ways that team members interact with each other are displayed in a so-called Wattle
Tree visualisation.
2. Domain and Learner Model in GRAPPLE
The models in GRAPPLE follow a different approach than described in Section 1, except
that there are also domain and user models (see Figure 2). The domain model [5] basically
consists of a concept map of the learning domain. The relations between concepts can
express semantic relations between concepts (as usually done in concept maps), but also
hierarchical relations between concepts can be expressed. In addition to the domain model,
the Conceptual Adaptation Model (CAM) is the basic model where adaptive lessons are
defined by using concepts of the concept map and connecting them with pedagogical
relations. Pedagogical relations can freely be defined and used in adaptive lessons to
indicate the sequence of concepts for the adaptive engine.
Figure 2: GRAPPLE domain model (concepts with related learning resources) with user model
variables defined on these concepts.
The user model [1] is totally flexible, since every kind of user model variable can be
defined upon a domain model. A user model variable is a variable of any data type and is
associated with all concepts of an adaptive lesson. For example the variable knowledge can
be defined as integer with a range from 0 to 100, so that the user model can express the
knowledge level for all concepts. Another example is the variable visited defined as boolean,
which is used to express all concepts a user has visited with associated learning resources.
The user model of the CbKST tool has a simpler structure, since the skills can be seen
as the knowledge dimension of concepts. Hence, in the notation of GRAPPLE there is only
S. L. Wong et al. (Eds.) (2010). Proceedings of the 18th International Conference on Computers in Education. Putrajaya, Malaysia:
Asia-Pacific Society for Computers in Education.
one predefined user model variable knowledge, which cannot be altered. Also the relation
between concepts of the CbKST Tool can be seen as a special case of the GRAPPLE
approach, since in GRAPPLE every kind of relationship can be defined. However, defining
such relations in GRAPPLE is rather a pedagogical design than psychologically proven
dependences between skills.
3. Visualisation Tools and Techniques
Following the flexible user and domain model approach described in Section 2, also the
visualisation techniques have to be flexible in order to capture the information provided in
these models. According to the domain and especially user model, there are several
dimensions which can be displayed to the learner:
• a distinction between a single and a multiple learner view
• different user model variables defined on concepts
• performed activities in terms of learning resources
• goals in terms of concepts and user model variable values on concepts
In order to achieve flexibility also for the visualisation technique, standard
visualisation techniques have been developed which can capture all or most of the
information dimensions described above. For example, Figure 3 shows two of the
developed visualisation techniques. Figure 3a depicts the knowledge user model variable
for the concepts of a lesson for one learner (purple bars). Furthermore the average values of
the other learners are shown (red bars) and the expected level (goal) is also shown for each
concept (top black line). Figure 3b outlines which learner has performed which activity
(purple circle on crossings in matrix). Furthermore, the average values for learners and
activities are expressed with the small bar diagrams. Both representations can be employed
for other representations, whereas the information to be represented can freely be chosen.
Figure 3: Two different visualisations: (a) knowledge level of the concepts of a lesson for a single
learner, and (b) activities performed for a class.
A set of visualisation widgets has been implemented where each widget uses a specific
visualisation technique or user model representation respectively. Some of them are simpler
in terms of the presented information and others are rather complex connected information
is displayed in one widget. Furthermore, some of the widgets are intended to be used by
learners and others are rather suitable for teachers or tutors.
A data format has been defined (in JSON format) which can contain all the
information and each visualisation tool gets the same data for a lesson. Depending on the
visualisation technique and the chosen information dimensions to be visualised, the tool
selects the respective parts and renders them. These tools have been implemented as Flash
S. L. Wong et al. (Eds.) (2010). Proceedings of the 18th International Conference on Computers in Education. Putrajaya, Malaysia:
Asia-Pacific Society for Computers in Education.
objects (using Macromedia Flex), which get the data over HTTP from a Web application
having access to the user and domain model data. The visualisation tools can be included in
Web pages of Learning Management Systems connected to GRAPPLE. According to the
configuration settings different information dimension can be displayed.
4. Conclusion and Outlook
In this paper visualisation techniques and according tools have been presented which are
capable of rendering flexible user model data. These tools visually open up the data used for
adaptation of learning resources, which should help learners to get an overview on their
current learning progress. Furthermore, they can compare themselves with other learners,
which should have positive effect on their motivation. A limitation can be identified, that no
meaningful guidance can be provided with these tools as it easily could be done with the
CbKST Tools.
An initial evaluation has been conducted with 43 students and 32 university lectures.
The overall result of the student and teacher visualisations indicated a medium to good
quality in all aspects (suitability for the task, self-descriptiveness, usability, meta-cognition,
cognitive load, benefits for instructors, and acceptance). This result suggests that these
visualisations are suitable for their intended purpose and also largely self-descriptive and
understandable. Learners think that this visualizations are suitable for getting an overview
of the current status in learning process. The result of the more complex visualisation
(Figure 3b) is significantly inferior to those of the simpler visualisations (e.g. Figure 3a).
The reason might be that it is more difficult for students to understand the complex
information.
Acknowledgements
This paper and work presented in this paper is part of the ongoing research and development in the EU FP7
project GRAPPLE (Project Reference: 215434) and could not be realized without the close collaboration
between all 15 GRAPPLE partners, not listed as authors, but nonetheless contributing to the ideas described
here
References
[1] Abel, F., Heckmann, D., Herder, E., Hidders, J., Houben, G.-J., Krause, D., Leonardi, L., & Sluijs, K. van
der (2009). A Framework for Flexible User Profile Mashups. Paper at the International Workshop on
Adaptation and Personalization for Web 2.0 (APWEB 2.0 2009).
[2] Albert, D., & Lukas, J. (Eds.) (1999). Knowledge Spaces: Theories, Empirical Research, and
Applications. Mahwah, NJ: Lawrence Erlbaum Associates.
[3] Albert, D., Nussbaumer, A., & Steiner, C. (2008). Using Visual Guidance and Feedback Based on
Competence Structures for Personalising E-Learning Experience. Proceedings of the 16th International
Conference on Computers in Education (ICCE 2008), 27-31 October 2008, Taipei, Taiwan.
[4] Bull, S., & Kay, J. (2008). Metacognition and Open Learner Models. In: The 3rd Workshop on
Meta-Cognition and Self-Regulated Learning in Educational Technologies, at ITS2008.
[5] De Bra, P., Smits, D., Sluijs, K. van der, Cristea, A., & Hendrix, M. (2010). GRAPPLE: Personalization
and Adaptation in Learning Management Systems. To appear in: Proceedings of the 22nd World
Conference on Educational Multimedia, Hypermedia & Telecommunications (ED-MEDIA 2010)
[6] GRAPPLE (2010). Generic Responsive Adaptive Personalized Learning Environment. EC-funded FP7
research project, http://www.grapple-project.org/
[7] iClass (2010). Intelligent Distributed Cognitive-based Open Learning System for Schools. EC-funded
FP6 research project, http://www.iclass.info/
[8] Kay, J., Kalina, Y., & Reinmann,.P. (2007). Visualisations for Team Learning: Small Teams Working on
Long-term Projects. Proceedings of the Conference on Computer-supported Collaborative Learning
(CSCL 2007), International Society of the Learning Sciences.
[9] Steiner, C., Nussbaumer, A., & Albert, D. (2009). Supporting Self-Regulated Personalised Learning
through Competence-Based Knowledge Space Theory. In Policy Futures in Education, Vol 2009/7/6.
