Appended papers

Appended papers
Appended papers
Paper 1
Paulsson, F., & Naeve, A. (2007). Establishing technical quality criteria for
Learning Objects. In P. Cunningham & M. Cunningham (Eds.), Exploiting the Knowledge
Economy: Issues, Applications, Case Studies (Vol. 3, pp. 1431-1439). Amsterdam: IOS Press.
Paper 2
Paulsson, F., & Naeve, A. (2006). Virtual Workspace Environment (VWE): A
Taxonomy and Service Oriented Architecture Framework for Modularized Virtual Learning
Environments - Applying the Learning Object Concept to the VLE. International Journal on
E-Learning, 5(1), 45-57.
Paper 3
Paulsson, F., & Berglund, M. (2008). Suggesting a SOA-framework for modular
virtual learning environments: comparing two implementation approaches. International
Journal of Web-Based Learning and Teaching Technologies, 3(1), 43-57.
Paper 4
Paulsson, F., & Berglund, M. (2006). A service oriented architecture-framework
for modularized virtual learning environments. In A. Méndez-Vilas, A. Solano Martin, J.
Mesa Gonsalez & J. A. Mesa Gonsalez (Eds.), Current Developments in Technology-Assisted
Education (1 ed., Vol. 3, pp. 1935-1939): FORMATEX.
Paper 5
Paulsson, F., & Engman, J. (2007). Treating metadata as annotations: separating
the content mark-up from the content. International Journal of Emerging Technologies in
Learning (iJET), 1(3), 1-7.
Paper 6
Paulsson, F., & Engman, J. (2005). Marking the National Curriculum - a new
model for semantic mark-up. In P. Cunningham & M. Cunningham (Eds.), Innovation and the
Knowledge Economy: Issues, Applications and Case Studies (Vol. 2, pp. 1731-1738).
Amsterdam: IOS Press.
1
2
1 Introduction and background
The research area, together with the background of my research and the problems studied are
presented in this chapter. Some basic concepts that are used in this thesis are introduced. The
structure, the research questions, the purpose and delimitations of this thesis are also
described.
1.1 Technology Enhanced Learning and learning technology
The use of technology to support learning has been around for quite some time now. The kind
of technology, and the ways technology have been used has varied over time. Nowadays,
when Technology Enhanced Learning (TEL) is discussed, it is usually learning with the help
of ICT that is targeted.
The use of computers for learning and education started to emerge in the 1950’s, emanating
from efforts to train soldiers in a fast and efficient way during the Second World War (see,
e.g., Saettler, 2004, pp. 178-196). Behaviorists like Skinner and Thorndike were in the lead of
the early efforts that eventually led to the idea of instructional design. Instructional design
consists of several related theories and concepts that are rooted in learning theories.
Instructional design theory deals with the methods of instruction; while instructional models,
instructional strategies, etc., represent the practice of applying the theory to instructional
development or instructional development processes (Reigeluth, 1999c).
The instructional design movement continued to evolve during the later half of the 20th
century and it has made deep impressions in the TEL history with ideas such as the Teaching
Machine, first introduced by Skinner (1958), later re-launched by Gagné (1965). Other
milestones were the concepts of programmed instruction that were intended to be realized by
the Teaching Machine, based on Skinners frames of instruction, described by Skinner (1958)
and programmed learning that was an evolvement of programmed instructions. Later on,
during the sixties, seventies, and eighties, programmed learning evolved into computer-based
training and computer-based learning, and during the seventies learning theory and
pedagogical practice started to shift from behavioristic theories to cognitivists’ learning
theories (see, e.g., Gagné, 1977).
One of the most significant technological developments during the eighties was the
introduction of affordable desktop computers that opened up for the use of technology in most
organizations – including schools. Desktop computers made it possible to introduce
pedagogical software at the classroom level.
The development and use of TEL exploded during the nineties and the breakthrough of
multimedia and CD-ROM technology in the early nineties opened up new possibilities for the
delivery of digital learning content. In parallel, the development of hypertext - that started in
forties with the work of Bush (1945) and gained speed in the sixties, when Nelson (1965)
coined the term hypertext – eventually led to the development of the World Wide Web (the
web), (see, e.g., Nielsen, 1995), which revolutionized TEL. Suddenly TEL was a global affair,
and it became possible to distribute collaborative learning and digital learning content in ways
that were unmanageable just a decade before. The web shifted the focus from ComputerBased Training (CBT) to Web-Based Training (WBT) or Web-Based Learning (WBL) –
depending on the pedagogical point of view. During the eighties and nineties, the pedagogical
focus shifted once again, this time to the benefit of constructivism, influenced by the work of
researchers like Vygotskij and Piaget (see, e.g., Mylov, 2000; Dusek, 2006, pp. 198-210).
The breakthroughs of the Internet and the web marked the beginning of a general
globalization that changed the way ICT-systems and system borders are perceived. ICT3
systems and their functionality have become a global issue that goes beyond organizational
boundaries. Consequently, system functionality becomes distributed and intertwined in a way
that creates dependencies that are beyond the control of a single organization. This has
triggered a development of new types of flexible systems and services, and has at the same
time created a need for common standards and frameworks for open interfaces. Altogether,
those developments have lead to a modularization of ICT infrastructures and systems.
New ideas and strategies have been applied in order to meet the new demands for flexibility
and adaptability, and this has triggered new ways of implementing systems. One of the most
significant changes during the last decade is an increased level of system modularity. Modular
web-based applications are often accomplished by exposing (providing) functionality as
services to the outside world, such as Web Services in a web-based Service Oriented
Architecture (SOA). Services can subsequently be used (consumed) by other systems or
services as loosely coupled functionality-components within a larger pool of services. The
usage contexts are provided by the way in which services are being used, and a service may
be used for different purposes in different contexts. This way it is possible to create flexible
and dynamic application-like functionality on the web using modular approaches where
services interrelate through the utilization of each other, creating a global grid of functionality
and information.
Web Service-based SOA-approaches have mainly been applied to legacy systems, but are also
slowly changing how new systems are designed and implemented. Modular infrastructures
and system architectures are becoming more common in new development as well, and
general reference models and modular frameworks are emerging within several application
areas, including TEL, where these create the foundation for new kinds of distributed Virtual
Learning Environments (VLE). Examples of such SOA-based frameworks are the eFramework, described by Wilson, Olivier, Jeyes, Powell, and Franklin (2004), and the Open
Knowledge Initiative’s (OKI) Open Service Interface Definitions (OSIDs) that are used in the
Sakai project, described by Thorne and Kahn (2003).
In addition to Web Service-based approaches there are other more lightweight approaches to
exposing or syndicating functionality and data that are already used in VLEs. Examples are
technologies such as XML-RPC or RSS1, which are used for simpler, but yet powerful,
services in isolated contexts. Those technologies support modularity as well, as they can be
used to syndicate a specific service or specific data to be used by other services or
applications.
There is a comprehensive amount of work that implies that increased modularity is a path to a
considerably higher level of flexibility, from the technological point of view, as well as from
a usage point of view in terms of how functionality and information can be accessed,
assembled and utilized in different contexts (see, e.g., Williams, 2001; Baldwin & Clark,
2000; Nierstrasz & Dami, 1995). Flexibility can also be contemplated from a pedagogical
point of view, and pedagogical flexibility is something that is needed in order for technology
to meet new and changing pedagogical requirements.
Modular approaches for TEL have mainly been applied to digital learning content, and
Learning Objects is one of the most widespread and widely adopted concepts for delivery of
such content. Learning Objects represent a modular approach that aims at facilitating
interoperability, flexibility and reuse of digital learning content. The positive relation between
modularity and flexibility can be derived from research on learning objects as well.
Abernethy, Treu, Piegari, and Reichgelt (2005), uses the case of introductory ICT courses to
1
See Appendix 1 for explanations of technical terms.
4
show how enhanced flexibility is achieved using modular learning content. They emphasize
the relation between the level of granularity of learning objects and the achieved flexibility.
Rehak and Mason (2003), point out the benefits of modularity, while stressing the problems
caused by differences in how the learning object concept is perceived in the learning
community, (see also, e.g., Gibbons & Nelson, 2002; Silveira et al., 2007). These citings are
merely examples of work that emphasizes the relation between modularity and pedagogical
flexibility. The dilemma is not to establish that there is such relation, the problem is rather
that the views of what flexibility and adaptability are and how it can be achieved and utilized
are ambiguous.
The emergence of the learning object concept has triggered a change in how VLEs are
perceived and implemented. The idea of Learning Management Systems (LMS) is intimately
connected to learning objects, and LMSes were originally a way of placing learning objects
into a clearer and more suitable context for education and training in trade and industry. It is
therefore common for LMSes to focus on training and performance rather than on learning
and education. Many LMSes are more focused on providing functionality for administration
and management of learning, courses, and learning objects, rather than on providing advanced
features for pedagogical activities (see, e.g., Ismail, 2002). The pedagogical context is often
provided as a part of the learning content governed by the way in which the content is
sequenced and packaged. The result is that much of the application logic providing
pedagogical support is coded inside digital learning content.
The term LMS is sometimes used synonymously with the term VLE, and this causes a lot of
confusion. As an example, Paulsen (2002) defines a Learning Management system as:
“Learning Management System is a broad term used for a wide range of systems that
organize and provide access to online learning services for students, teachers, and
administrators. These services usually include access control, provision of learning
content, communication tools, and organization of user groups.”
In this thesis LMSs are regarded as being subtype of VLEs. LMSs are further discussed in
chapter 3.
In recent years, the focus has shifted toward social interaction, user-created content and
different forms of user participation. The view of the learner is transforming from being a user
of the VLE to being a participator in distributed VLE communities, combined with
technological developments, changing attitudes, open services on the web, etc, a change that
is sometimes referred to as Web 2.0 (see, e.g., O'Reilly, 2006; Anderson, 2007). This
evolvement toward social interaction and participation corresponds well to the shift of
pedagogical focus toward constructivist learning theories that has occurred during the last 1020 years (see, e.g., Jonassen, 1991c; Dempsey & Van Eck, 2007).
1.1.1 The future is semantic
Parallel to the development described in the previous chapter, the Semantic Web envisioned
by Berners-Lee, Hendler, and Lassila (2001), is evolving. The Semantic Web is changing the
ways in which information management and system interaction can be done, creating new
exciting possibilities for knowledge management and TEL that are likely to completely
change the way in which we regard information searching and browsing. One such example is
the concept browser Conzilla, described by Naeve (2001). The semantic web will most
definitely be essential for TEL and it will have implications for how Web Services and
modular systems can be assembled and kept together (see, e.g., Nilsson, 2001; Nilsson,
Naeve, Duval, Johnston, & Massart, 2008; Naeve, 2005; Naeve & Sicilia, 2006; Dietze,
Gugliotta, & Domingue, 2007; Richter, Nejdl, & Allert, 2002; Koper, 2004; Devedzic, 2004).
5
Semantic Web Technology is already used to create Semantic Web Services, i.e., selfdescribing services that can be automatically discovered and matched with other services and
information for automatic web service composition without intervention of humans (see, e.g.,
Nagarajan, 2006a; Lama, Arroyo, Sicillia, & Lópes-Cobo, 2006; Dietze, Gugliotta, &
Domingue, 2007).
The technical foundation for the Semantic Web is an information description language called
Resource Description Framework (RDF) (Decker et al., 2000). RDF’s abilities to represent
data and information semantics make it possible to transfer some of the programmatic and
logical capabilities of application software to be represented by information semantics using
RDF, and be completely detached from application software. Instead it can be interpreted by
RDF-aware software and incorporated in different services. One consequence of this is that
software components and software services can act (and re-act) differently depending on the
semantics expressed by RDF-data. These are especially interesting capabilities for dynamic
and modular environments, and for representing dynamic and generative processes
independently of software and application logics. Cardoso and Sheth (2006a), and Fensel and
Bussler (2002) describe the concept of combining Semantic Web technology and Web
Services into Semantic Web Services.
1.2 Emerging problems
The technical developments leading up to new ways for delivering TEL, described in chapter
1.1, together with a shift in pedagogical focus have led to a veritable revolution of learning
and education. New possibilities have arisen, but also new challenges and problems.
A change in pedagogical focus has occurred parallel to the technological development. While
Behaviorists’ learning theories were predominant during the early to mid-20th century, there
has gradually been a shift toward cognitivistic and constructivistic learning theories during the
later half of the 20th century (see, e.g., Schunk, 2003). Constructivistic approaches and
associated pedagogical methods, such as problem-based learning and case-based learning, are
becoming increasingly popular among teachers and pedagogues in all parts of society. The
correspondence between behavioristic learning theories and the learning technology used in
computer-based training and other similar approaches is quite good, while the pedagogical
shifts during the later half of the 20th century have contributed to the creation of a gap
between the pedagogical requirements and expectations and the supporting technology.
Many pedagogues discern interesting possibilities in the use of technology, but the change in
pedagogical focus and the variety of pedagogical requirements introduce new demands on
how technology must be implemented, as well as on how pedagogical design must be realized
and represented by technology. Much of the technology in use today still bears a legacy from
its early behavioristic roots, focusing on instruction and training rather than on learning. The
consequence is that a large portion of the learning technology in use does not match
pedagogical expectations and requirements. Learning technology needs to become more
flexible, both from a pedagogical point of view and from a technological point of view.
Pedagogical flexibility is dependent on technological flexibility, which means that learning
technology must be able to support and adapt to a variety of pedagogical approaches, as well
as to different functional requirements provided by these approaches.
While an increased level of modularity provides a way of increasing flexibility, LMSes often
fail to meet the requirements for flexibility through modularity. LMSes are often implemented
in a silo-like fashion, unable to adapt to a modular approach. As information silos, they do not
interact well with the surrounding world and this problem becomes worse as much of the
surrounding world consists of other information silos. Those information silos often contain
6
information and functionality that would be useful for the LMSes, and this leads to a need to
duplicate data and functionality, creating unwanted redundancy (see, e.g., Downes, 2002).
The situation becomes even more complicated as learning objects provide a modular concept
for content, while there is no corresponding model or concept for the VLE, which means that
the potential flexibility provided by modularity cannot be utilized to its full potential. Hence,
modular concepts aiming for reusable and flexible digital learning content are being
implemented in non-modular and inflexible VLEs. For many reasons this is a non-fit, and
much of the pedagogical benefits that could be obtained through modularity are lost when
using non-modular VLEs. Current practices with LMSes merely use the VLE as a container
for unpacking and displaying digital learning content that is instructionally designed and
sequenced beforehand, while missing out on many of the possibilities that are created by
modular content in combination with modular functionality (see, e.g., Dodani, 2002;
Abdallah, El Hajj, Benzekri, & Moukarzel, 2002; Downes, 2002; Edwards, Rai, Philips, &
Fung, 2007).
Both within the fields of computer science and TEL, modularity is often ‘sold’ using
arguments like reuse, flexibility and adaptability (see, e.g., Silveira et al., 2007; Polsani, 2003;
Dodani, 2002). At the same time learning objects are criticized for not meeting the
expectations of modularity, reusability, and platform and pedagogical neutrality. Learning
objects simply do not deliver what they are expected to deliver (see, e.g., Dodani, 2002;
Polsani, 2003; Downes, 2002).
In this thesis, I argue that there is a good and simple reason for that: the expected
characteristics of learning objects are well expressed by the learning object community but
they have failed to express how those characteristics should be implemented in order to be
accomplished. This becomes even more evident when comparing learning objects to
component technologies within computer science, where modularity is essential for flexibility
in terms of reuse, dynamics and adaptability.
Ideas and concepts from computer science were the original source of inspiration for the
learning object community when the concept was first invented and are often referred to when
discussing learning objects and modularity, such as by Song and Andersson (1999) and
Dodani (2002). The definitions of modularity and related concepts, as well as how they
should be implemented are much clearer within computer science. There are clear rules,
models and frameworks that make it all work. Szyperski (2002c) defines a software
component (i.e., a type of module) as:
“… a unit of composition with contractually specified interfaces and explicit context
dependencies only. A software component can be deployed independently and is subject to
composition by third parties.” (p. 36)
A problem that the learning object community is struggling with is that some of the most
important characteristics that define components have been left unspecified. The result is that
ideas are falling apart and concepts and definitions are becoming fuzzy and straggly. Those
issues can be regarded as problems that concern the technical quality, and there are many
lessons to be learned from computer science.
However, modularity is only one piece of the puzzle. In general learning is not a linear and
homogenous process, which means that there is also a need for new ways of representing
pedagogical processes in VLEs, processes that utilize the potential flexibility that is created
by modularity. Current practice of sequencing, such as provided by SCORM Sequencing and
Navigation (Capone, Christensen, Falls, & Fletcher, 2004) is not powerful enough. One of the
main reasons for this is that sequencing assumes that the pedagogical sequence and process
7
can be pre-determined. Attempts have been made in order to develop more sophisticated
models for representing pedagogical process. The most common approach has been through
the use of meta-languages, so-called educational modeling languages (EML) (see, e.g.,
Rawlings, Koper, van Rosmalen, Rodrígues-Artacho, & Lefrere, 2002). One of the more
recent attempts is the IMS2 Learning Design (Koper, 2005). Even though educational
modeling languages provide considerably more flexibility, they only solve a portion of the
problem, as the pedagogical process still needs to be pre-determined to a large extent.
A conclusion that can be drawn from the reasoning above is that a high level of modularity
provides a sound basis for achieving technical flexibility, and that technical flexibility can be
utilized to achieve pedagogical flexibility. Furthermore, there is a need to establish an
infrastructure that can provide the conditions needed to reach the same level of modularity for
the virtual learning environment as learning objects are intended to provide for digital
learning content.
My hypothesis is that a reasonable approach would be to incorporate digital learning content
and virtual learning environments into the same conceptual space, using a common
framework and a common abstract model. To utilize the potential flexibility that is provided
by a modular infrastructure there is a need to represent pedagogical design and pedagogical
process in ways that are generative rather than normative, which is the case with most current
practice.
2
IMS Global Learning Consortium, http://www.imsproject.org/
8
1.3 Research questions
Based on the problems outlined above, the following research questions have been identified:
1. How can a common abstract model for modularity, which covers both the VLE and
digital learning content, be accomplished without creating incompatibilities with
existing models?
2. How can a technical infrastructure be designed and implemented according to a
common abstract model, in order to provide the necessary supporting infrastructure,
and without disqualifying legacy?
3. How can pedagogical designs and processes be represented in order to utilize
modularity to increase pedagogical flexibility and adaptability?
1.4 The purpose of this thesis
The overall aims of this thesis are:
To examine how, and to which extent, modularity can help in enhancing pedagogical and
technical flexibility.
To show how experience from computer science and system architecture can be applied to
learning objects and the VLE, by establishing links between modularity in computer science
and modularity in pedagogical practice.
These aims will be accomplished using an experimental approach, based on the reasoning in
chapter 1.2, by examining modularity through implementation in software systems and
infrastructures.
9
1.5 Thesis structure
The first three chapters provide the background, context, and theoretical framing for the
research presented in this thesis, while the last two chapters account for the results,
conclusions and discussion of the work.
Figure 1-1. The structure of this thesis.
The first chapter gives an introduction to the research field and a description of the problem
area. The research questions and the aim of the thesis are also presented in chapter 1. Chapter
2 summarizes the papers and gives an overview of the projects that constitutes a large part of
the empirical context of this thesis. Chapter 3 contains an overview of the research field and
provides the theoretical framing. The research results and my contributions to the field are
presented in chapter 4, after which my conclusions and discussion of the results follow in
chapter 5.
The structure of the thesis is shown in Figure 1-1 and the mapping of the research questions
and the appended papers are shown in Figure 1-2.
10
Figure 1-2. The three research questions and how they are addressed in the appended papers.
1.6 Delimitations
Even though this thesis emanates from a pedagogical point of view, includes pedagogical
contexts and discusses pedagogical theories, the research has a strong focus on technical and
technological consequences of pedagogical requirements. Hence, the pedagogical aspects are
to be regarded as the empirical context of the technological research presented in this thesis.
The main focus is on infrastructure, modularity and system architecture for virtual learning
environments and digital learning content used for TEL.
The purpose is neither to push any pedagogical values nor to examine the implications of
different pedagogical approaches. The only relevant aspect for the context of this thesis is that
different pedagogical theories and approaches exist, and that this undisputable fact has
consequences for the design and development of the learning technology to be examined.
Furthermore, this thesis does not compare general technical characteristics (such as
performance) or detailed technical differences in the implementation strategies of
technological frameworks, other than in cases where this is relevant in a TEL context or from
a modularity point of view.
11
1.7 Concepts and terms
This thesis is interdisciplinary in many ways, and I reserve for myself the use of some terms
and concepts in a slightly unorthodox way depending on the context. This has its explanation
in the fact that some of the concepts and terms are used differently in the TEL community
than within computer science. Some terms have a slightly different meaning also within
computer science, depending on whether they are used in, for example, component-based
software engineering, or in a web application context, such as with web services or in a Web
2.0 context.
The rest of this chapter is devoted to briefly explaining some basic terms that will be further
discussed in the coming chapters, while other terms and concepts will be explained as they are
introduced.
1.7. 1 Learning technology
The term learning technology will be used for referring to the technology that is used for
learning. The term learning technology puts the focus on learning that can be either formal or
informal, and not on education, which can be interpreted as a more formal form of learning,
as in the term educational technology, used by Januszewski and Yeaman (2001). The term
learning technology can be interpreted as being fairly pedagogically neutral as it avoids the
tension of terms like instructional, as in instructional technology that leads the thoughts to
instructional approaches, such as instructional design or individual instruction. Seels and
Richey (1994) use the term instructional technology.
Learning technology is a general term that covers all aspects of technology used for learning
in all contexts, including technology that is mainly intended for a specific context or
approach, such as approaches with an instructional focus. For the purpose of this thesis, the
term learning technology only refers to digital technology, i.e., ICT, even though learning
technology could be argued to embrace analog technology as well.
1.7.2 Technology Enhanced Learning
While the term learning technology is used to refer to the technology used for learning, the
term technology enhanced learning (TEL) will be used to refer to the actual activity of
learning with the support and enhancement of technology. The reasons for using the term
TEL are similar to the reasons for using the term learning technology: it is an allencompassing term that covers most aspects of learning with the support of technology,
independent of the pedagogical approach. As with learning technology, the term TEL is used
in the sense that it limits the use of technology to the use of ICT, excluding analog
technology, such as radio, videocassette recorder (VCR), telephones, and so on. The term
covers special cases where TEL can be used, such as for distance learning, flexible learning,
blended learning, etc.
Examples of other terms that are sometimes used synonymously with TEL are ComputerBased Training (CBT), Computer Assisted Instruction (CAI), and e-learning. In this thesis,
those terms are regarded as specialized directions that are all covered by the term TEL. The
expression ICT and learning is commonly used, especially in the Nordic countries (see, e.g.,
Söderlund, 2000), when addressing TEL in schools and primary/secondary education. ICT
and learning is about as broad as TEL and is here regarded as synonymous.
1.7.3 Object – as in learning object
The term ‘learning object’ is used when learning objects are addressed as a phenomenon and
concept. The term is used in the same sense as a large part of the TEL-community uses the
12
term, that is, addressing the concept of platform and context independent, self-contained
pieces of reusable digital learning content, which can be assembled to form a larger unit of
digital learning.
The notion of an object is quite different when addressing learning objects compared to an
object in Object-Oriented programming (OOP). An object in OOP is a clearly defined
instance of a class with explicit properties, operations and attributes. Learning objects on the
other hand are actually more similar to software components, but often without “contractually
specified interfaces and explicit context dependencies only” (see the quotation of Szyperski
(2002c), in chapter 1.2).
1.7.4 Digital learning resources
A digital learning resource is an arbitrary digital resource that is used for learning, or a digital
resource intended to be used for learning. The reason for using the term digital learning
resource is that the term learning object is often misused, as discussed in chapter 3, which
easily leads to misunderstandings and confusion. The term learning object is still too vague
and undefined, and is therefore merely used in contexts where learning objects are discussed
as a phenomenon, while the term digital learning resources is used when such resources
(including learning objects) are discussed in a general context.
1.7.5 Component
The term software component as defined by Szyperski (2002c) will be used when for
addressing components from a software architecture and software development point of view.
The term component will be used in a general sense to describe a constituent part (module) of
a composite and modular architecture or infrastructure.
13
14
2 Projects and paper summaries
This chapter describes the projects in which the studies presented in this thesis were carried
out. The purpose is to establish a better understanding of the context and background. This
chapter also contains descriptions of my contributions to the appended papers in relation to
my co-authors and paper summaries.
2.1 Projects
The studies and data collection were carried out within three different projects. These were
three separate projects but related to each other in different ways, and their existence
depended on each other. Altogether the outcomes of the projects make an important
contribution to the research field and to this thesis by addressing research questions 1 and 2,
presented in chapter 1.3, in suggesting a system-design approach that promotes modularity
and unifies the predominating model for digital learning content, the learning object model,
with a corresponding modular architecture-model for the virtual learning environment.
Research question 3 was addressed through the suggestion and implementation of a prototype
of a semantic web-based annotation model for representing pedagogical processes through
dynamic linking of curricula and components in a modular learning architecture.
These projects and the studies presented in this thesis should be regarded in the light of my
background of being a computer scientist on the one hand and being a teacher, working with
teacher training, on the other hand.
2.1.1 ICT and schools – a complicated relationship
The Virtual Workspace Environment (VWE) project, described in chapter 2.1.2, has provided
the empirical backbone of this thesis.
The VWE-project started in 1998, and was born out of a desire to change the way in which
virtual learning environments were perceived. ICT seemed to be apprehended differently in
many educational institutions (especially in schools) than in other sectors of society. This is,
for example, indicated by Papert (1994), as well as by Pedersen (1998). This phenomenon
revealed itself in the way that teachers often responded to the introduction of ICT as yet
another problem, rather than the solution to pedagogical problems, or at least being a
supporting tool for pedagogical practice and learning.
Figure 2-1 illustrates this situation. The figure was originally designed for a presentation
(Paulsson, 1998) preparing for the initiation of the European Schoolnet (EUN) in 1998.
15
Figure 2-1. The view on ICT of many teachers, as understood in 1998. Unfortunately, this
figure is still valid in many respects.
In Sweden, teachers also had (and still have) a rather complicated relationship to the use of
ICT. This complicated relationship is essentially related to previous experiences from national
ICT programmes and ICT projects. Söderlund (2000, p.9, pp. 48-50) examines the long-term
relationship between schools and ICT in detail. Söderlund discusses how ICT has been
introduced in Swedish schools between 1980 and 2000, and how teachers have understood
ICT. Söderlund also points out that the adoption of technology is much slower in schools than
in other parts of society. According to Söderlund, one reason for this may be that ICT has a
strong symbolic value and the school system has been the subject of a number of politically
initiated measures. The purpose and objective of the chosen political measures have not
always been clear, except for the sense that there has been a belief that the use of ICT will
lead to positive change. The result has often been that ICT has been introduced and
implemented based on political conditions rather than on the conditions of teachers and
students, and this is also likely to be one of the reasons for the attitude illustrated in Figure 21. Those circumstances are emphasized by Söderlund (2000, pp. 179-183), where he discusses
the difference between the “political” approach described above, and the bottom-up approach
taken in some cases at the local level, where ICT has been implemented based on the
conditions and needs of pedagogues in combination with knowledge building for teachers.
Hence, this complicated relationship between pedagogues and ICT together with the topdown perspective seem to have been strong influential factors for the common belief that the
design and implementation of ICT and ICT tools is something that cannot (and according to
some - should not) be influenced by active involvement by teachers or students. There is an
obvious risk that this may strengthen the attitude illustrated in Figure 2-1. The VWE project,
described in the following chapter, was set out to involve pedagogues in the design and
implementation of ICT.
16
2.1.2 The Virtual Workspace Environment3
The VWE project was aimed at changing the view on ICT, described above, by shifting some
of the control to the users, that is teachers and students, and at the same time making it easier
for teachers and students to participate in the actual design and implementation of their VLE
and the ICT tools that they are supposed to use. This was motivated by an expectation that it
would allow teachers and students to set the boundaries for the pedagogical activities to a
higher extent, rather than allowing the technology to do that.
The basic idea of the VWE project was to develop a modular VLE that could be controlled
and managed by teachers and students to a much higher extent than what was (and still is)
usually the case.
The VWE philosophy was inspired by concepts and ideas from component technology in
computer science and ideas similar to the ones leading up to learning objects, while
maintaining a strong focus on modularity, adaptability and reuse. The rationale behind the
VWE project was the assumption that modular content, that is learning objects, are not
enough to obtain the level of adaptability and flexibility needed to support the variety of
learning contexts, pedagogical approaches and the diversity of needs of teachers and students.
Learning objects only contribute to the adaptability and flexibility of digital learning content,
but the pedagogical freedom is still restricted by the VLE. This is a problem that relates to
research questions 1 and 2 that were presented in chapter 1.3.
The idea constituting the basis for the VWE is that teachers and students should be able to
compose and aggregate (as well as disaggregate) their personal and shared VLEs. The
composition of a VLE is carried out using a “toolbox” containing components, in a similar
way as content is supposed to be aggregated and disaggregated using learning objects.
Such modular concepts lead to a level of flexibility and adaptability that allows teachers and
students to better adapt and form their (personal and shared) VLEs to their own needs, as well
as to the needs of a group. As a consequence, this also affects the ability to adapt to the needs
of a specific pedagogical approach, or a specific learning situation. An important condition
for this kind of adaptability is the possibility of dynamically adding, as well as removing,
functionality.
2.1.3 The Standardized Contextualized Access to Metadata (SCAM)4
The need for a general-purpose system for managing advanced metadata was born out of
needs that arose within the VWE project, together with experiences and need from other
projects at the Royal Institute of Technology (KTH), and the Knowledge Management
Research group (KMR)5.
The more modular the VLE and the digital learning content became, the greater was the need
for sophisticated metadata management. The need for metadata that kept things together grew
as the VLE transformed into a complex web of loosely coupled and (in terms of functionality)
simple components, which interacted in a rather complex system of loose and dynamic
relations. It was realized that machine-readable and machine-processable semantics would be
needed in order to not put too much cognitive burden on the users, and to make it possible to
locate components for different purposes and within different contexts, as well as for
3
The VWE project was funded by the European Commission, and by the Swedish National Agency for
Education as a part of the European Schoolnet (EUN) project. The VWE project was lead and hosted by the
department of Interactive Media and Learning (IML) at Umeå University.
4
The SCAM project is an ongoing project that is hosted by IML at Umeå University together with the
Knowledge Management Research Group (KMR-group) at the Royal Institute of Technology (KTH).
5
http://www.kmr.nada.kth.se/
17
providing the VLEs with the potential to react, change and evolve based on feedback from
components and actors within the system. This includes VLEs that are adaptive as well as
adaptable. A VLE can be made adaptable by modularization, but it became apparent that
machine-processable metadata were needed in order to also achieve adaptivity6.
The SCAM system, which is a result of the SCAM project, is an RDF repository that can be
plugged in for use by (or as a part of) other software or services on the Semantic Webb
(Paulsson, 2003). SCAM was originally intended to be used as a development platform and
code library for developers, but has over time evolved into two branches of stable, industrystrength applications as well. These are packaged as SCAM Nimble, for the development of
learning object repositories (LOR) that conforms to standards such as Dublin Core (DCMI,
2004), IMS Metadata (IMS, 2001) and IMS Content Packaging (IMS, 2004), and the SCAM
ePortfolio7. The SCAM ePortfolio is an ePortfolio for the Semantic Web, which can be
integrated into VLEs as a service.
SCAM is currently distributed using a combined GPL/LGPL Open Source License (OSI,
2007).
Issues related to SCAM and the storage and management of RDF in relational databases are
discussed in detail by Palmér, Naeve, and Paulsson (2004). However, these issues are outside
the scope of this thesis.
2.1.4 The Annofolio project8
The Annofolio project can be regarded as a sub-project of the SCAM ePortfolio project. The
objective of the Annofolio project was to explore the possibilities of creating dynamic
relations, based on machine-processable semantics, between digital resources, such as
information objects, curricula, learning object and other components in a modular learning
architecture.
A prototype application, called the Annofolio, was produced by the project. The Annofolio is
basically an ePortfolio with sophisticated annotation functionality, implemented as an add-on
to the SCAM ePortfolio. The Annofolio was tested in a case where learning objectives and
central concepts in the national steering documents were dynamically connected to learning
objects in a LOR. The connections were made using semantic relations in the respective set of
annotations that were added using the Annofolio tool (see paper 6 for detailed examples).
This kind of semantically-based interlinking adds a certain amount of adaptivity, where
digital learning resources can be found and explored based on the relation to corresponding
learning objectives in the curriculum, expressed through semantic web annotations. Research
question 3, presented in chapter 1.3, was addressed in the Annofolio project through the
development of a prototype for connecting digital learning resources to the curriculum.
6
I make a distinction between adaptable and adaptive. Adaptable means that the VLE can be adapted through
active actions by users, while adaptivity means that the VLE can automatically adapt and react to changes or
inputs that occur in the VLE based on generated data, external input, learner preferences, attention metadata, and
so on.
7
Also called Confolio.
8
The Centre for IT in Northern Sweden (CINS) hosted the Annofolio project, together with IML at Umeå
University. The Annofolio project was funded by the Swedish Agency for Innovation Systems (VINNOVA) as a
part of a project called Tools and methods for markup of educational content, lead by the research programme
Digital Literature at Uppsala University.
18
2.2 Paper summaries
2.2.1 My contributions to the papers
I am the main author and have written all of the papers that are appended to this thesis. I have
also been responsible for and carried out the major part of the scientific work presented in the
papers.
The co-authors’ roles have mainly been to assist with the implementation, design and coding
of prototypes, which are indeed valuable contributions to the projects described above.
The work presented in the papers was carried out within the projects described in chapter 2.1.
I have worked as the project leader in all three projects.
The nature of my contributions is presented below. I have, for the sake of clarity, chosen to
explicitly state the contributions of each co-author as well. Observe that the co-authors have
agreed to the descriptions below.
Paper 1. I wrote the paper and was responsible for carrying out the scientific work, including
the presented studies that resulted in the suggestion of six areas for quality criteria for
learning objects. I presented the paper at the eChallenges 2006 conference in Barcelona.
Ambjörn Naeve assisted me with feedback and suggestions during the study and assisted in
delimiting and formulating the six areas for quality criteria.
Paper 2. I wrote the paper and was responsible for the scientific work that resulted in the
VWE Learning Object Taxonomy, which was also the main result presented in the paper.
Ambjörn Naeve provided feedback and suggestions regarding the framing of the VWE
learning object taxonomy. He also assisted in modeling and visualizing the VWE learning
object taxonomy using the Conzilla concept browser9. The first VWE prototype and SOA
framework were also briefly presented and discussed in paper 2. My contributions to that
work are described below.
Paper 3 and paper 4. I wrote the papers and was responsible for carrying out the main part
of the scientific work as well as for the development of the VWE concept. I was responsible
for the design of the VWE SOA architecture framework as well as parts of the Java RMIbased SOA-architecture. I also contributed to the coding of the prototypes and development
(and adaptation) of VWE Tools. Mikael Berglund contributed to the VWE project with
substantial amounts of code and he had the main responsibility for the implementation design
of the Web Services architecture.
Paper 5 and paper 6. I wrote the papers and was responsible for executing the main part of
the scientific work. I was the originator of the Annofolio markup concept, which was first
described in, (Paulsson, 2002). Jonas Engman, at CINS, made important contributions to the
transformation of the Annofolio concept into a working method. He also contributed to the
system design and the further development of the markup method, together with me. Engman
also contributed with substantial amounts of code for the prototype. I contributed to the
coding as well. My main focus was on the code related to XML and XSLT, and the
integration with the SCAM ePortfolio. I carried out the study and the scientific work
connected to the markup of the national steering documents, presented in paper 6.
9
http://www.conzilla .org/
19
2.2.2 Paper 1: Establishing technical quality criteria for Learning Objects
Published in the proceedings of the eChallenges 2006 conference (25-27 October 2006), in
Barcelona, Spain.
Paper 1 presents a technical evaluation of learning objects from two Swedish learning object
repositories. The paper also reviews some of the literature and research that addresses
learning object models and definitions in order to propose a number of areas for working with
technical quality criteria for learning objects.
The absence of common low-level definitions and abstract models for learning objects
presents a threat to the interoperability and technical quality of learning objects - as well as to
learning objects as a concept. Most of the vision associated with learning objects has yet to be
fulfilled, because most current implementations of learning objects do not fulfill this vision. It
is therefore essential to establish criteria that secure the technical quality of learning objects as
one step in addressing those issues. As a result, six areas for technical quality critera were
suggested.
The quality criteria were tested as a basis for the VWE learning object taxonomy, and by
implementation in the Virtual Workspace Environment (VWE), described in paper 2, paper 3
and paper 4.
2.2.3 Paper 2: Virtual Workspace Environment (VWE): A Taxonomy and Service
Oriented Architecture Framework for Modularized Virtual Learning Environments
- Applying the Learning Object Concept to the VLE
Published in the International Journal on E-Learning, 5(1), 45-57, (2006).
The term learning object is used for referring to digital learning resources that are supposed to
be modular, fairly context independent and have a high level of reusability. The basic idea is
that several learning objects can be assembled to form a larger pedagogical unit (a learning
module) that can be used for a specific learning context. Even though learning objects have
been around for more than a decade, and the concept is widely used and implemented, there is
no common or broadly accepted definition.
The atomic learning object taxonomy, suggested by Wiley (2002a), works fairly well for
defining the characteristics of learning objects at some levels; especially on the functional and
usage levels. Wiley’s taxonomy is based on the level of complexity and interactivity of
learning objects, but it fails to define learning objects at the architectural and interface levels,
i.e., how to actually accomplish the functional characteristics that are expressed in the
taxonomy.
An alternative abstract model and learning object taxonomy (the VWE learning object
taxonomy) was presented in this paper. The VWE learning object taxonomy is compatible
with Wiley’s atomic learning object taxonomy, which means that it can essentially be
regarded as a complementing taxonomy rather than a competing taxonomy. All the taxonomy
levels of Wiley’s taxonomy can be implemented using the VWE learning object taxonomy,
and the objective of the VWE learning object taxonomy is to add an abstract model for
learning objects where data, application logics, and presentation are separated into different
layers. The VWE learning object taxonomy does not only address modular content, it also
addresses functionality components in VLEs using the same modular concept as learning
objects normally represent for content. This becomes possible when application logics, data,
and presentation are separated so that a learning object can represent application functionality
as well as digital content independently of each other.
20
Even though this paper briefly describes the VWE SOA architecture as well, the major
contribution of this paper is the VWE learning object taxonomy.
2.2.4 Paper 3: Implementing a Service Oriented Architecture for modularized
Virtual learning Environments: a comparison of two approaches
Published in the International Journal of Web-Based Learning and Teaching Technologies,
3(1), (2008).
This paper compares the implementations of two different Service Oriented Architecture
approaches for modular virtual learning environments. Both prototypes implement a similar
system design, using two different approaches: the first prototype is implemented using Java
and Remote Method Invocation (RMI), while the second prototype is implemented using Java
and web service technology. The objective of the system design was to implement an
architecture that corresponded to the VWE learning object taxonomy in order to provide an
architecture framework that facilitates the implementation of modular virtual learning
environments.
Some design issues are discussed in the paper together with a comparison of the implications
of using Java RMI versus web service technology. The comparison is made from the point of
view of modular VLEs, and the main purpose is not to compare general technical qualities,
such as differences in performance, but to demonstrate that the architecture framework can be
implemented using different approaches and technology platforms, and to point out
similarities and differences, as well as advantages and disadvantages.
The comparison showed that some minor changes had to be made in the system design when
moving from Java RMI to web services. The results were in favor of the web service
approach, although it is technically inferior in some ways and trade-offs were made in terms
of performance and efficiency. The advantages of web services are especially related to the
use of web service standards and the use of the http-protocol together with the use of an
XML-based protocol, i.e., SOAP. However, some of the biggest disadvantages are also
related to the use of the http-protocol and to the use of an XML-based protocol, as the
combination of http and XML create overhead that influence several layers in the architecture.
The most important results presented in paper 3 are that the prototype implementations of the
VWE learning object taxonomy proves that an approach where digital learning content and
the VLE are incorporated into the same abstract model works, and that it can be implemented
using different technology frameworks.
2.2.5 Paper 4: A Service Oriented Architecture-framework for modularized Virtual
Learning Environments
Published in the proceedings of the m-ICTE2006, (22-25 November), in Seville, Spain
Paper 4 is a conference paper that describes the final VWE web service-based architecture in
detail. The web service architecture, which was compared to the Java RMI approach in paper
3, was the final outcome of the VWE-project.
Even though papers 3 and 4 have some overlap, they each have a different focus. Paper 3
compares two approaches focusing on the comparison and on showing that the VWE learning
object taxonomy can be implemented using several approaches. Paper 4 focuses on the
resulting web service architecture, which was also, considering the current state of SOA,
suggested as the most reasonable approach to implement modular VLEs in accordance to the
VWE learning object taxonomy.
21
2.2.6 Paper 5: Treating metadata as annotations: separating the content mark-up
from the content
Published in the International Journal of Emerging Technologies in Learning (iJET), 1(3),
2007.
Structural markup and metadata are becoming increasingly important as the use of digital
learning content increases. Metadata are needed for discovery of digital learning resources as
well as for describing different aspects of digital learning resources, such as intended use,
technical constraints, relations to other resources, context of use, and so on.
Structural markup, commonly XML-syntax, is needed to describe the structure of the digital
learning content. Structural markup can be used for presentation, transformation, and for
processing of digital learning content in different ways. There are, however, increasing needs
to add metadata and annotations with machine-processable semantics that describe the
internal structure of a resource. This way, parts of a digital resource can be described in the
same way as metadata are used to describe the entire resource in a repository. In many cases,
only a small part of a resource is of interest, such as an excerpt of a text or a clip from a
movie. It is sometimes desirable to add annotations that emphasize a certain perspective, a
certain usage, or in other ways comments on a part of a resource. This can, for example, be
used to describe relations to other resources or to other sources of information, as is discussed
in paper 6.
Traditionally, markup languages such as Standard Generalized Markup Language (SGML)
and XML have been used for describing both structure and semantics, often combined in the
same schema, such as in the Text Encoding Initiative (TEI) 10, which is one of the most used
schemas for annotation of literary and linguistic texts. This mix of syntax and semantics is
problematic for several reasons: firstly, XML was developed for expressing syntax and
structure not semantics, secondly it is not possible to build a metadata ecology (M Nilsson,
Palmér, & Naeve, 2002) on a principle where metadata have to be added inside a resource
internals, that is into the XML-file in the case of TEI. Besides causing problems concerning
access and rights, the XML-file would simply become too messy and grow uncontrollably as
new metadata are added. There is no discernible realistic way of controlling the integrity of
the added markup, as anyone would be able to edit the XML-file. In addition, adding
descriptive semantic data into XML-files creates an unwanted processing overhead.
The problems described above were addressed by developing a method and a set of
corresponding tool-prototypes for “pointing-in” annotations into the right section of an XMLfile. Annotations are represented using RDF, which (in theory) means that any metadatamodel or application profile11 can be represented and used for markup. Metadata are stored
externally, separated from the resource, in a SCAM-repository. Anyone can add metadata and
in the case of the Annofolio, metadata are stored in user portfolios in the SCAM ePortfoliosystem. The method creates a layered and ecology-like metadata system, where different
layers of metadata are separated, stored and managed at an individual level by the provider.
The purpose of the Annofolio method is to use the structural markup of XML-resources to
“point-in” descriptive annotations that are stored in the SCAM-based Annofolio ePortfolio.
The main findings presented in paper 5 are a general-purpose method and architecture for
adding annotations to existing XML resources. The method relies on existing technologies,
such as XML for structure and syntax, XPointers for identifiers, the eXtensible Stylesheet
10
http://www.tei-c.org/
In a metadata context, an application profile is a defined set of elements and guidelines for metadata for a
specific application that can be put together by combining one or more metadata schema.
11
22
Language (XSL) for XML transformation, the Document Object Model (DOM) for managing
the hierarchy of XML-documents, JavaScript for manipulation, and semantic web
technologies (RDF and Annotea), for expressing machine-processable semantics.
2.2.7 Paper 6: Marking the National Curriculum - A New Model for Semantic
Mark-up
Published in the proceedings of the eChallenges 2005, (2005, 19-21 October), in Ljubljana,
Slovenia.
Paper 6 describes a case where the Annofolio was used to add annotations to the Swedish
national steering documents12. The method, which was described in detail in paper 5, was
developed as an alternative to XML-based markup methods, such as the Text Encoding
Initiative (TEI), where metadata semantics are expressed within the XML-document
Parts of the Swedish national steering documents were annotated using the Annofolioprototype. Central concepts and learning objectives were furnished with machine-processable
metadata in a selected number of course plans. A basic layer of curriculum metadata was
provided by indexers at the Swedish National Agency for Education in order to form the basis
for a curriculum service that was built based on the course plans in the national steering
documents.
Two demonstrators were developed to illustrate how the machine-processable metadata and
the curriculum service can be used. The first demonstrator used the markup tool for compiling
digital compendiums, consisting of excerpts from course plans that were chosen by the user
by using annotations to mark the starting points and end points of the sections of interest. The
markup was used as the basis for the generation of a digital compendium using the eXtensible
Stylesheet Language: Transformation (XSLT)13. The second demonstrator illustrated how the
annotated steering documents could be dynamically linked to learning object metadata. The
later presumes that learning object metadata are expressed using RDF (cf. Nilsson, Palmér, &
Brase 2003). Suitable learning objects were located based on the semantics expressed by the
annotations describing, for example, a specific learning objective within a course plan. In
addition, the connection could also be made in the opposite direction; that is, metadata
associated with a learning object could be used to locate relevant sections in the steering
documents.
The architecture of the Annofolio-system is briefly described in this paper. However, the
main relevance, in the context of this thesis, is to illustrate how the Annofolio method can be
used to generate semantic relations between curricula and digital learning resources; that is
between pedagogical instructions and other components in the learning environment.
In the following chapter I will present my theoretical framework. The theoretical framework
will serve as the point of departure where my research results and my contributions to the
research field are discussed in chapter 4 and 5.
12
The Swedish national steering documents (http://www3.skolverket.se/ki03/front.aspx?sprak=EN) are the
Swedish national curriculum, regulations, laws relating to public education, and course plans.
13
XSLT is a stylesheet language designed for transforming the structure of XML documents. XSLT is also used
to transform XML documents into other file formats, such as html.
23
24
3 Research field and theoretical frame of reference
This chapter describes the research field and the theoretical framework in which the research
presented in this thesis was carried out. Because the focus of this thesis is on modularity and
the development of virtual learning environment, this chapter will span both pedagogical
theory and theory from computer science. This width, involving several research fields, is also
the reason why this chapter may appear as unusually extensive. The purpose is to illuminate
the interrelations between technology and pedagogy.
3.1 Theoretical perspectives
The research presented in this thesis concerns the interrelations between learning theories,
instructional design14, and the role of modularity in the design and development of learning
technology. The understanding of these relations is important for the understanding of TEL as
well as for the understanding of many of the issues involved in the design and development of
VLEs. The use of ICT is, in general, becoming more and more applied and process focused,
and by that also more interdisciplinary. TEL is a typical application field of ICT and it is
important to regard TEL and learning technology from a holistic point of view. DodigCrnkovic, and Crnkovic (2003) express this point of view from a general perspective:
“The development of human thought parallel to the development of human society has led
to an emergence of Sciences that do not belong to any of the classic types we have
described earlier, but rather share common parts with several of these. Many of the
modern Sciences are of interdisciplinary, eclectic type. It is a trend for new Sciences to
search their methods and even questions in very broad areas.” (p. 1)
Even though this thesis focuses on the technological (computer science) aspects of TEL and
the design and development of VLEs, the reasoning above implies that a synthetic worldview
is needed in order understand the issues raised in chapter 1, and in order to address the
research questions formulated in chapter 1.3.
My theoretical framework needs to be rooted in several research fields, spanning from
psychology and pedagogy to computer science. I have adopted an eclectic approach to my
research in order to cope with this.
The objective of this chapter is not to explain different theoretical standpoints in detail, but to
increase the understanding of, and to some extent establish the relations (the main threads and
connecting thoughts) that more or less explicitly exist between pedagogy and technology.
These are relations that are sometimes overlooked or not understood. The focus is on the
mainstream theories, and I have taken the freedom to focus on the aspects that are of special
relevance for this thesis, well aware that many aspects of the theories are excluded. The
reason for this is that all of the theoretical fields are extremely extensive and delimitations
need to be done. In spite of this it is my intention to illuminate a couple of different
viewpoints within the different fields. It should also be noted that some of the phenomena
described in this chapter are not scientific theories in a literal sense, but are nevertheless
relevant for this thesis and for establishing my theoretical framework.
I have structured this chapter into three parts: pedagogical theory, TEL theory, and computer
science theory. This structure could be questioned for several reasons, some of the
14
I use the term instructional design in the first section (pedagogical theory) as it refers to an established
research field. The term learning design will be used in relation to Educational Modelling Languages (EML)
(see chapter 3.10), where it is appropriate. The term pedagogical design will be used as a general, collective term
to address the activity of planning and designing pedagogical activities - including instructional design and
learning design.
25
pedagogical theories, such as that learning theory could be argued to be a part of psychology.
There are, of course, no clear boundaries between the three fields and there are overlaps and
grey zones, especially in the TEL section, which is situated somewhere in-between
technology and pedagogy, but this disposition suits the purpose of this thesis in terms of
clarity and structure.
Pedagogical theory deals with theory that is of general relevance for learning and teaching,
TEL theory is “in-between theory” that is not solely pedagogy, and not solely computer
science or technology, but that has a clear TEL focus. Computer science theory deals with
theories that are of relevance for the design and implementation of learning technology, in
terms of software, software architecture and infrastructure - with a specific focus on
modularity. In some cases I have chosen to emphasize the historical perspective. The reason
for this is that I find the historical viewpoint relevant for determining and understanding the
current position and state of TEL, as well as for understanding the origins of my research
questions (chapter 1.3) and the emerging problems in the research field (chapter 1.2). This is
also the reason why the ordering is chronological.
3.2 Learning theory
This chapter gives an overview of the learning theories that are indicated by the literature to
have been the most influential during the 20th century. The purpose is to establish a better
understanding of the relation between learning theories, instructional design, and the design of
learning technology, together with chapter 3.3 and 3.4. Learning theory is also used as a tool
to encircle some of the emerging problems described in chapter 1.2, as well as for the
derivation of the origins of my research questions, in particular research question three.
The literature indicates, consistently, that the three dominating directions of learning theory
during the 20th century were behaviorism, cognitivism and constructivism (see, e.g., Schunk,
2003; Ertmer & Newby, 1993; Mylov, 2000). Each learning theory has different implications
for the implementation of pedagogical instruction (the result of instructional design) and by
that also for the implementation of learning technology (see, chapter 3.4.1, chapter 4, and
chapter 5).
The purpose of chapter 3.2.1 to 3.2.4 is to briefly discuss the origins and the main features of
the different learning theories in order to establish a basic understanding of the similarities
and differences of behaviorism, cognitivism, and constructivism.
3.2.1 Behaviorism
Learning theories based on behaviorism were dominating during the first 60-70 years of the
twentieth century (see, e.g., Schunk, 2003; Saettler, 2004). The basis for behavioristic
learning theories is the principle of stimulus-response. According to behavioristic learning
theories, behavior is triggered by external and environmental stimuli. Skinner (1976) refers to
this mechanism as operant conditioning. The difference between classical conditioning
(Pavlov, 1927), and operant conditioning is mainly that operant conditioning addresses
voluntary behavior and is thereby better applicable to human learning. Skinner and Pavlov are
likely to be two of the most well-known behaviorists and their respective work has been
influential in the field of behavioristic learning theory. Many commonalities can be found in
the work of Skinner and the work of Watson, such as their view of conditioning (see, e.g.,
Walker, 1990). Watson is, together with Thorndike, often regarded as two early and
influential behaviorists.
Thorndike established a behavioristic branch called Connectionism (e.g., Walker, 1990). He
regarded the mind as an interconnected network of small and simple units that are associated
26
to each other. Much of his work was carried out in relation to testing the intelligence and
learning ability of children. Thorndike formulated three basic laws that established the
relations between stimulus and response (e.g., Walker, 1990): 1) the law of effect, which
states that positive response will enhance behavior and negative response will usually weaken
behavior; 2) the law of readiness, which states that a series of responses can be chained
together in order to achieve goals; and 3) the law of exercise, (which was later withdrawn)
which claims that the connection between stimulus and response is that behavior is
established by repeated connections of stimulus and response. Like Pavlov (1927), Thorndike
mainly based his research on animal behavior.
Watson transferred his research based on animal experiments to human learning and behavior
as well. He argued that psychology should only pay attention to observable behavior and not
worry about mental processes (Andersson, 2000, pp. 63-75). Watson (1920) even carried out
an experiment involving a child (Albert) with the purpose of proving that classical
conditioning (conditioned fear in Albert’s case) was transferable to humans. Watson (1920)
meant that all humans are born with a couple of basic reflexes, and that the rest of our
knowledge and behavior is the result of stimulus-response. Watson is sometimes referred to as
the father of behaviorism, and is maybe most famous for his idea of being able to shape
children to whatever he wanted.
“Give me a dozen healthy infants, well-formed, and my own specified world to bring them
up in and I'll guarantee to take any one at random and train him to become any type of
specialist I might select—doctor, lawyer, artist, merchant-chief and, yes, even beggar-man
and thief, regardless of his talents, penchants, tendencies, abilities, vocations, and race of
his ancestors.” (Watson, 1925)
According to pedagogical literature, Skinner is the researcher who is most commonly
associated with behaviorism from a learning theory point of view. Even though Skinner seems
to have based much of his research on the work by Watson, Skinner studied the stimulusresponse pattern from a human learning perspective, establishing the principle of operant
conditioning (see, e.g., Andersson, 2000, pp. 63-75). The basic mechanisms of operant
conditioning differ from the classical conditioning in that learners operate on the environment
and are rewarded for the right behavior with some form of feedback (reinforcement) that can
be either positive or negative. Skinner (1976) believed that no reinforcement leads to that
certain behavior is not repeated and that punishment can repress behavior. Skinner used the
concept of chaining to explain how small learning steps could be chained together as
sequences in order to learn sequences of behavior. Chaining later became one of the corner
stones of instructional design (see chapter 3.4), and is sometimes referred to as chunking and
sequencing.
A distinguishing characteristic of behaviorism is that the learner is given a rather passive role
in the learning process: the learner is an object for teaching (see, e.g., Andersson, 2000). The
learner is viewed as tabula rasa (i.e., blank state), an expression coined by the 18th century
philosopher Locke (see, e.g., Nilsson, 2001), which implies that humans start out with no
mental content. Knowledge is gained by modifying behavior by stimulating change of
behavior, using positive and negative reinforcements. The mind is a black box and cognitive
processing is generally not considered in the learning process.
Behavioristic learning theories provided a distinct and clear basis for the early attempts of
using instructional design with technology in that it was – and still is – quite straightforward
to represent using technology. According to behaviorists, every step in the learning process
can be predicted and by that described and programmed beforehand. There are no individual
differences or unexpected turns that need to be considered. Behaviorism was the dominating
27
learning theory during the main part of the 20th century. According to Mylov (Mylov, 2000),
behavioristic learning theories started to be challenged first in the sixties, even though
alternatives existed long before that. The first direction to challenge behaviorists’ learning
theories was cognitivism (see, e.g., Schunk, 2003, pp. 83-135; Saettler, 2004; Marton &
Booth, 2000, pp. 25-29; Bjerg, 2000).
3.2.2 Cognitivism
A shift in favor of learning theories based on theories and models from cognitive science
seemed to have occurred between the mid-fifties and sixties, (see, e.g., Schunk, 2003, pp. 83135; Saettler, 2004; Marton & Booth, 2000, pp. 25-29; Bjerg, 2000). Partly, this shift seems
to have been an evolvement of behaviorism that led to theories such as the social learning
theory, described by Bandura and Walters (1963) and (cf. Bandura, 1977), and some regard
social learning as the bridge between behaviorism and cognitivism (Schunk, 2003, pp. 83135). Bandura and Walters (1963) argued that learning could be accomplished through
modeling, and by observing and imitating other people, as well as through the traditional
behavioral mechanisms. Bandura (1977) found the idea that the environment is the cause of
all behavior to be too simplistic to explain the complexity of human beings. He argued that
the environment and individual’s behavior cause each other. Bandura claimed that an
individual’s personality is created out of interplay between the environment, behavior, and
internal psychological processes. Bandura also found that learning could be achieved merely
by watching others, emphasizing social aspects in the cognitive process (Bandura, 1986), also
referred to as Social Cognitive Theory (see, e.g., Schunk, 2003, p. 83). This represents a view
of learning that stands out as being quite different, and to some extent the opposite of the
ideas represented by behaviorists, like Watson, Skinner, and others.
From the above point of view, cognitivism can be seen as a reaction to behaviorism – a
reaction that moved the focus from observable behavior to more complex cognitive processes
such as thinking, memory, and problem solving, by shifting the focus to internal cognitive
processes (see, e.g., Snelbecker, 1974). The increased focus on cognitive processes was a
result of the progresses within cognitive psychology15 that were made during the fifties,
referred to as the cognitive revolution (see, e.g., Saettler, 2004, p. 318; Marton & Booth,
2000, pp. 25-29).
Within cognitivism, knowledge is regarded as a schema that is considered to be an internal
structure of mental constructions. The schema changes when learning occurs. New
information is compared to the learner’s existing schema, which may be altered or extended
when new knowledge is gained. Typical for cognitive learning theories is that the learning
process is regarded as a three-stage information-processing model, where input is first
registered by the learner’s senses, then entered into the short-term memory, and finally
transferred to the long-term memory for storage and retrieval, referred to as information
processing, described by Schunk (2003, pp. 136-189). Hence, cognitive processing and
previous knowledge are given a lot of attention in cognitive learning theories. Learning is
regarded as being a mental activity that concerns the receipt, organization and storage of
information (Jonassen, 1991c), also taking higher mental processes into account. Cognitivism
changed the view of the learner from being passive to being active, even though the role of
the environmental conditions was still emphasized. Clark (2006), summarizes cognitivism as
“…learning is about active construction of new knowledge by interacting with new
information, and instruction is about promoting the psychological processes that mediate that
construction”. (p. 5)
15
Nowadays cognitive psychology is commonly referred to as being a part of cognitive science.
28
Examples of other influential contributors to cognitive learning theory are Piaget, introducing
the Stage Theory of Cognitive Development (Piaget, 2006; Piaget & Inhelder, 1997) (see also
chapter 3.2.2), Bandura, (see above), Merrill (1983), with the Component Display Theory
(CDT) Reigeluth (1999b), who introduced Elaboration Theory and Bruner (1961), whose
theory of Discovery Learning, an inquiry-based instructional method that emphasizes
personal discovery of knowledge, which are sometimes regarded as a bridge to constructivism
(see, e.g., Schunk, 2003, pp. 241-284).
The pedagogical change of focus, from behaviorism to cognitivism, influenced the practice of
instructional design (Ertmer & Newby, 1993), which in turn had implications for the
representation of instructional designs and pedagogical processes. For those reasons, the work
of Merrill and Reigeluth is further discussed in chapter 3.4. This also led to new instructional
models that were set out to adapt pedagogical instructions to cognitive learning theories for
use with TEL. One such example is the Inquiry Teaching model by Collins (1977), which was
a question-based model, used within Computer Assisted Instruction (CAI) (see also chapter
3.5).
For the understanding of learning theory, I find it important to emphasize that there is a
difference between cognitive science and cognitive learning theories. While cognitive science
is the science practiced within psychology, cognitive learning theory (cognitivism) is the
application of theories from cognitive science, addressing a specific branch of epistemology.
The reason that this distinction needs to be made is that cognitive science is important for
other learning theories as well, such as for constructivistic learning theories that eventually
challenged cognitivism.
3.2.3 Constructivism
Simpson (2002) argues that constructivism is a psychological and philosophical perspective,
and an epistemology rather than a learning theory. However, constructivism is often referred
to as a learning theory in line with behaviorism and cognitivism (see, e.g., Schunk, 2003, pp.
285-328).
Compared to cognitivism, constructivism emphasizes personal experience, and learning is
best accomplished by doing (see, e.g., Dewey, 2004). The role of the teacher tends to
transform into a guiding and mentoring role, as knowledge transfer from teachers to students
is not considered to occur. As with other learning theories, the roots of constructivism can be
traced back in history. Rousseau, for example, criticized that children were taught about
things, rather than experiencing the things themselves. Rousseau said that: “You think you are
teaching what the world is like; he is only learning the map”, cited by (Subramaniam, 2002).
Rousseau argued, like modern constructivists, that the emphasis should be on learners’
construction of their own mental models, rather than on teachers conveying theirs. Like
Piaget, Rousseau devoted his attention to the developmental stages of children. He introduced
a theory that explained the development of a child in four stages (see, e.g., Egidius, 2003, pp.
17-24). Rousseau’s stage theory was later revised by Piaget (2006) and presented as the Stage
Theory of Cognitive Development.
Schunks overview ( 2003, pp. 285-328) shows that the individual construction of knowledge
is one of the core values of constructivism, and knowledge is regarded as something that is
individual in the sense that everyone constructs his or her own representation of the world and
there is no objective truth. Learning occurs in a context and the more authentic the context is
the more efficient learning becomes, and in social constructivism the social context is
especially emphasized, “Constructivism highlights the interaction of persons and situations in
the acquisition of skills and knowledge” (Schunk, 2003, p. 287). The active nature of learners
29
is strongly emphasized by constructivists and by necessity these are assumptions that have a
strong influence on didactics, instructional design, and pedagogical methods. They also affect
the teachers’ role in the way that they put the focus on creating and structuring situations that
stimulate learners to actively engage with content and act in a social context, rather than
traditional teaching. The teacher becomes a mentor and is expected to provide guidance rather
than on acting in the traditional role of the lecturing teacher. According to Schunk (Schunk,
2003, pp. 285-328) this has also influenced curriculums toward synthetic and integrated
approaches.
Constructivistic learning theories challenge the traditional, lecture and teacher-centered
teaching approaches and put the learner and the construction of knowledge and personal
mental models in the centre. Learning becomes a personal as well as a social process of
assimilation and accommodation when the learners assimilate new knowledge into previous
experience, changing mental models, testing them against social interactions, assimilating
again and so on (Piaget, 2002).
Constructivism is sometimes divided into three main directions: exogenous constructivism,
endogenous constructivism, and dialectical constructivism (see, e.g., Schunk, 2003, pp. 285328). Exogenous constructivism refers to knowledge as a construction of the external world,
endogenous constructivism refers to knowledge as being derived from previous knowledge,
and is a cognitive abstraction of the external world, and dialectical constructivism refers to
knowledge as the product of the interaction between the learner and the environment.
Knowledge is influenced by the external world as well as by the mind (see, e.g., Schunk,
2003, pp. 285-328; Piaget, 2002).
Two of the major contributors to constructivism, as it is perceived today, are Piaget and
Vygotsky (see, e.g., Schunk, 2003, pp. 285-328; Marton & Booth, 2000). Piaget’s research
showed that children learn from doing, not from being fed with information. He argued that
children are testing their mental models against new experiences, and new knowledge occurs
through assimilation and accommodation, in which new experience is assimilated into old
mental models (Piaget & Inhelder, 1997). This triggers an accommodation process that
changes mental structures to accommodate new experiences. Piaget meant that we constantly
refine (construct) our mental model of the world, adapting it to new knowledge and
experience. Piaget’s work focused on children, and children’s cognitive development. Even
though Piaget is often considered to be a cognitivist (see, e.g., Egidius, 2003), his
contributions to constructivism are important and I have, like Schunk (2003, pp. 285-328) and
(Marton & Booth, 2000, pp. 21-29), chosen to discuss Piaget in the context of constructivism
as well.
Vygotsky (1978) developed a sociocultural constructivist theory that emphasizes social and
cultural factors in cognitive development. Vygotsky argued that social interaction is essential
for the construction of knowledge, and that language is our most important instrument for
this. Like Piaget, Vygotsky’s research was mainly focused on the cognitive development of
children. The fundamentals of Vygotsky’s theory state that social interaction and its
associated tools, are essential to learning, and that self-regulation is developed through
internalization (Vygotsky, 1978), see also Meece (2002, pp. 169-170). Human development is
a result of cultural transmission of tools (language and symbols) where language is the most
critical tool related to the zone of proximal development (ZPD), which is the difference
between what can be done by the individual, and what can be done by help of others
(Vygotsky, 1978).
Another central researcher within the field of constructivistic learning theory was Dewey,
who is most famous for his contributions to a branch of constructivism called pragmatism.
30
Dewey coined the phrase “learning by doing”, a phrase that summarizes the core of
pragmatism as a learning theory.
3.2.4 Comparing learning theories
A framework for comparing learning theories is needed in order to understand the differences
and similarities of the learning theories that have dominated the twentieth century. Schunk,
(1991, pp. 17-21), Schunk suggests five questions16 that can be answered based on different
learning theories, and that help in illustrating and understanding the differences and
similarities. Table 3-1 shows a simplified comparison, based on the work of Schunk (1991,
pp. 17-21; Schunk, 2003, pp. 17-21), and the work of Ertmer and Newby (1993). Table 3-1
illustrates the key differences of each learning theory.
Behaviorism
Cognitivism
Constructivism
How does
learning occur?
- Learning is a change in observable
behavior.
- When a proper response to an
environmental stimulus is given,
learning occurs.
- Reinforcement of responses
increases the likeliness for a behavior
to be repeated.
- Learning is discrete changes
between states of knowledge.
- The acquisition of knowledge is a
mental activity that entails internal
coding and structuring by the learner.
- The learner is perceived as active in
the learning process.
Which factors
influence
learning?
- Environmental factor are
emphasized.
- The arrangements of stimuli and
consequences is critical
- Environmental factors are
important.
- Mental activities that lead to a
response are emphasized, as well as
mental planning, goal-settings and
organizational strategies.
- Learners’ thoughts, attitude and
values are considered as important.
- Encouraging learners to use
appropriate learning strategies.
What is the role
of memory?
- Is not considered.
- Forgetting is caused by non-use of a
response over time.
- Learning occurs when information
is stored in the memory and
organized it in a meaningful way.
- Forgetting is the inability to retrieve
information from the memory.
- Both the knowledge and the use of
knowledge is stored in memory
How does
transfer occur?
- Generalization causes transfer
- Transfer occurs in identical or
similar situations, across common
elements.
- Transfer depends on how
information is stored in memory.
When a learner understands how to
apply knowledge in different
contexts, transfer has occurred.
- Prior knowledge is used to establish
boundary constraints for identifying
similarities and differences.
- By constructing meaning from
experience.
- The learner constructs his/her own
reality interplaying with the
surrounding world, based on
experience and interaction.
- Knowledge is not mind-independent
and cannot be transferred to learners.
- The internal representation of
knowledge is under constant change
and there is no objective reality.
- Both learner and environmental
factors are critical, with a focus on
the interaction between those.
- Learning contextual and content
knowledge should be embedded in
the situation where it is used.
- Learning must occur in realistic
settings using tasks that are relevant
to the experience of the individual.
- Learning is depending on activity,
concept and culture.
- The memory is always evolving and
changing.
- Representations of experiences are
not stored as formalized declarative
knowledge.
- Focus is on flexible use of
preexisting knowledge rather than
recall of pre-packaged schemas.
- The focus is on the creation of
cognitive tools in the sociocultural
context of the individual.
- Learners create situation-specific
understanding by assembling prior
knowledge from multiple sources,
supported by experience.
- By involving authentic tasks in
relevant contexts.
- Transfer is facilitated by
contextualization and experience.
- By emphasizing the portraying of
tasks rather than defining the
structure of learning required.
16
I have used the 1st edition of Learning Theories: an Educational perspective (Schunk, 1991), as this is the
edition that is used by Ertmer and Newby (1993). The reason is that the list needs to be consistent with the work
of Ertmer and Newby that I use in chapter 3.4.1. In the latest edition (4th) (Schunk, 2003) the list has been
changed to: How does learning occur, which factors influence learning, what is the role of memory, what is the
role of motivation, how does transfer occur, which processes are involved in self-regulation, and what are the
implications for instruction?
31
What types of
learning are
best explained
by the theory?
- Simple learning involving
discriminations, generalizations,
association and chaining.
- Basic skills training.
- Complex learning involving
reasoning, problem solving and
information processing.
- Knowledge that can be sized and
chunked so that it can be easily
assimilated and/or accommodated.
- In constructivism, it is impossible to
identify types of learning
independent of content and learning
context.
- It is not possible to isolate units of
information or knowledge domains
according to hierarchical analysis of
relationships.
- Complex knowledge in illstructured domains.
- Knowledge at the advanced and
expert level (Jonassen, 1991a)
Table 3-1. A comparison illustrating the key differences between learning theories. Based on
the research by Ertmer & Newby (1993).
Ertmer and Newby (1993) point out that there is no single “best” learning theory. Different
learning theories support different types of learning and learning objectives, and it depends on
the situation, on the learners, and on the learning objectives which theory is the most suitable.
In most cases the learning process is constantly changing, and within one learning situation
more than one learning theory may be involved. Learning strategies may be subject to change
as the learner progresses and new levels of knowledge are reached and the complexity
increases (Shuell, 1990). Hence, different types of knowledge are best taught and learned in
different ways. For example, facts and procedural knowledge are not thought of in the same
way as knowledge and concepts that depend on problem solving, complex mental processes,
and social interaction (see, e.g., Ertmer & Newby, 1993; Shuell, 1990; Schunk, 2003).
Each learning theory has its implications and limitations regarding what type of knowledge
that can be targeted and how learning is accomplished. In the context of this thesis it is mainly
the implications for instructional design and the representation of instructional design in
technology systems that are of interest in order to address my research questions. Instructional
design and the connection between instructional design and learning theories are discussed in
chapter 3.4.
As discussed above, different learning theories support and explain different types of learning
and learning objectives, which imply that there may be a need to classify knowledge and
learning objectives. This is discussed in the following chapters.
3.3 Classifying knowledge and learning objectives
There are several different ways of describing and classifying knowledge and learning
objectives. Such classifications can be used as a tool to analyze the implications of learning
objectives for the use of learning theories, instructional design strategies, and the design and
development of learning technology. Jonassen (1991a) made such a classification where he
describes three categories (levels) of knowledge: introductory, advanced and expert levels.
Jonassen’s categorization may seem a bit rough, but can be useful to describe the connection
between learning theories and their pedagogical implications at a general level, such as in
Table 3-1. However, for more in-depth analysis there is a need for sharper tools, and one such
tool is Bloom’s taxonomy (Bloom & Krathwohl, 1956), or the revised version of Bloom’s
taxonomy that is more focused on teaching and learning practice (Anderson & Krathwohl,
2001). In this thesis, Bloom’s revised taxonomy is used for illustrating the connections
between learning theories, different types of knowledge and learning objectives, learning
design, modularity, and the system representation of instructional design in learning
technology.
32
3.3.1 Bloom’s Taxonomy of Educational Objectives
In 1956, Bloom and Krathwohl (1956) published their Taxonomy of Educational Objectives
for the Cognitive Domain, commonly referred to just as Bloom’s taxonomy. Bloom’s
taxonomy has been (and still is) influential as a classification schema for educational
objectives. It is also used as a reference and tool by a wide variety of pedagogical
practitioners (Anderson & Sosniak, 1994). Bloom’s taxonomy has made it possible for
instructional designers to better analyze and specify instructional outcomes and objectives.
The basic idea of Bloom’s taxonomy (Bloom & Krathwohl, 1956) is to arrange educational
objectives in a hierarchal structure. The taxonomy is divided into three domains: Affective
(attitude), Cognitive (knowledge), and Psychomotor (skills). It is mainly the cognitive domain
that is of relevance for this thesis. According to Andersson and Sosniak (1994), the cognitive
domain is also the part of Bloom’s taxonomy that is most frequently used by practitioners (see
also Anderson & Krathwohl, 2001).
The cognitive domain consists of six levels that are hierarchically dependent on each other.
The classification is based on the level of required cognitive processing and on the cognitive
complexity. Higher levels represent a deeper understanding of a topic and a higher level of
thinking, which in turn represents a higher level of knowledge (Bloom & Krathwohl, 1956).
The six levels of the cognitive domain in the original Bloom’s taxonomy are illustrated in
Figure 3-1.
Figure 3-1. The six levels of Bloom’s taxonomy. The original taxonomy at the left and the
revised at the right. The figure is based on the work of Schult (2007).
In 2001, a revised version of Bloom’s taxonomy (Anderson & Krathwohl, 2001) was
published. Several researchers who were part of Bloom’s original workgroup were involved
in the 2001 revision - including Anderson.
The main differences between the old and the new taxonomy are a change of terminology and
a change of structure (Forehand, 2005), illustrated in Figure 3-1. These changes reflect some
of the changes in pedagogical practice, focus and terminology that have occurred since the
first version of Bloom’s taxonomy was published. Based on experiences from the use of the
old Bloom’s taxonomy, the revised Bloom’s taxonomy has a stronger emphasis on the
applicability as a tool for practitioner’s work with instruction and curricula (see, e.g.,
Anderson & Krathwohl, 2001; Forehand, 2005).
33
A new knowledge dimension has been added in the revised version of Bloom’s taxonomy.
The knowledge dimension has four different levels: factual knowledge, conceptual
knowledge, procedural knowledge and meta-cognitive knowledge, which represent the types
of knowledge to be obtained. The different levels of the knowledge dimension are connected
to different aspects of the cognitive process dimension through a set of categories and subcategories (Anderson & Krathwohl, 2001). Together this forms a grid that can be used by
practitioners to apply Bloom’s taxonomy to their practice. Table 3-2 illustrates this.
The reason for using the revised version of Bloom’s Taxonomy is that it better reflects
modern learning and teaching practices. The new knowledge dimension in the revised
taxonomy can be used to better understand which cognitive activities are involved at different
levels of the cognitive process (see Table 3-2). This cognitive mapping can be used to analyze
the requirements that must be put on the instructional design, and with that also the
requirements for its representation in systems.
The Cognitive Process Dimension
The Knowledge
Dimension
Remember
Understand
Apply
Analyze
Evaluate
Create
Factual
knowledge
List
Summarize
Classify
Order
Rank
Combine
Conceptual Knowledge Describe
Interpret
Experiment
Explain
Assess
Plan
Procedural Knowledge Tabulate
Predict
Calculate
Differentiate
Conclude
Compose
Construct
Achieve
Action
Actualize
Meta-Cognitive
Knowledge
Appropriate Use Execute
Table 3-2. The different levels of the knowledge dimension and their connections to the
cognitive process dimension. Based on Anderson & Krathwohl (2001, p. 40 pp. 67-68) and
Fisher (2007) .
Table 3-2, together with the illustration in Figure 3-1 illustrate the relation between different
types of knowledge, the cognitive activities involved and the associated cognitive complexity.
These figures may also be compared to the work of Ertmer and Newby (1993) (see Figure 32), where instructional strategies associated with different learning theories are compared
based on the level of the required cognitive processing and learner’s level of task knowledge.
34
Figure 3-2. A comparison of the associated instructional strategies of the behavioral,
cognitive and constructivist viewpoints based on the learner’s level of task knowledge and the
level of cognitive processing required by the task (Ertmer & Newby, 1993).
3.4 Instructional design and instructional design theory
Apart from learning theories, the development of TEL has been under strong influence of
instructional design and instructional design theory.
There are huge amounts of published work on instructional design and instructional design
theory, representing different theoretical viewpoints and different pedagogical approaches. As
a research field, instructional design is very broad and full of nuances and directions. This is
illustrated by the extensive work presented in several anthologies (see, e.g., Reigeluth, 1983;
Reigeluth, 1999; Wiley, 1999; Reigeluth, 2008). This makes it hard to establish a clear and
explicit definition of what instructional design is, and what it is not. I have made a selection
based on an estimation of how influential the work has been, how often it has been cited, and
(obviously) based on the relevance for this thesis.
This chapter is intended as a brief orientation in the field of instructional design theory and
instructional design. The objective is to provide a framework for understanding and
explaining some of the issues involved when practicing instructional design in connection
with TEL and the representation of learning design in learning technology. This
understanding is necessary in order to address research question 3. Furthermore, an
understanding of instructional design and its practice, in relation to learning theory and the
evolvement of TEL, places modularity into context. This is relevant for addressing research
question 1 and 2, and for understanding why modularity is important, what kind of modularity
is needed, and how modularity should be implemented.
Instructional design and instructional design theory deal with the theories of designing
pedagogical instruction and the structuring of learning material. According to Reigeluth
(1999c), instructional design theory is actually a collection of theories that has evolved during
the 20th century and that deals with methods of instruction. Instructional models, instructional
strategies, etc., represent the practice of applying the theories to instructional development
processes. However, among pedagogues and in everyday speech instructional design theory,
35
instructional theory, instructional models etc., are often referred to as just instructional
design. Reigeluth (1999c) also points out that instructional design theory differs from many
other theories in that it is design-oriented and probabilistic rather than deterministic - a
characteristic that is suitable for the nature of learning.
Reigeluth (1999c) summarizes instructional design theories as:
“… Instructional-design theories are design oriented, they describe methods of instruction
and the situations in which those methods should be used, the methods can be broken into
simpler component methods, and the methods are probabilistic.” (p. 7)
Merrill (1999) explains instructional theory as being concerned with two primary
considerations: What to teach and how to teach. This definition is actually similar to some
definitions of didactics. Didactics is concerned with the same questions, and sometimes also
the question of why something should be taught, (see, e.g., Jank & Meyer, 1997).
I will use the term “instructional design” to describe the process of applying instructional
design theory to learning through the analysis of learning needs for the purpose of planning
and designing pedagogical instructions for implementation with learning technology. A
definition that is compatible with Reigeluth’s definition, cited above. I limit the use of the
term “instructional design” to the use in connection with ICT.
There are an abundance of instructional design theories that are influenced by, and adapted to,
different learning theories. The development of the instructional design research field has
followed (or rather interplayed with) the pedagogical developments and trends in general.
A couple of examples of instructional design theories are: Gagnè’s (1977) Hierarchical Task
Analysis (see also Saettler, 2004, p. 346), which provides a model for sequencing based on
eight levels of learning sorted by increasing complexity, CDT (see also chapter 3.2) by
Merrill (1983). CDT places the emphasis on learner control. The Instructional Transaction
Theory (ITT), also discussed by Merrill, Li, & Jones, (1991) and Merrill (1999), is set out to
be an alternative to the early frame-based approaches (Merrill, 1985). ITT is a pattern-based,
algorithmic system based on the concept of instructional transactions rather than the
traditional sequential and branched approaches of frame-based systems. ET (see also chapter
3.2), described by Reigeluth (1999b), is an attempt to step away from teacher and contentcentered teaching, in favor of learner-centered instructions. This shift was accomplished by
focusing on guidance in the process of selection and sequencing of content. Several overviews
of instructional design theories and methods are available (see, e.g., Reigeluth, 1983;
Tennyson, Schott, Seel, & Dijkstra, 1997; Charles M Reigeluth, 1999a; Wiley, 2002b;
Merrill, 2002; Reigeluth, 2008).
Merrill (2002) argues that there are a number of principles that are common to all
instructional design and he refers to them as first principles of instruction. Merrill identifies
five common principles: 1) Learning is promoted when learners are engaged in solving realworld problems, 2) learning is promoted when existing knowledge is activated as a
foundation for new knowledge, 3) learning is promoted when new knowledge is demonstrated
to the learner, 4) learning is promoted when new knowledge is applied by the learner; and 5)
learning is promoted when new knowledge is integrated into the learner’s world. However,
many implementations of learning technology fail to implement all of those principles.
Learning Object Design and Sequencing Theory (LODAS) is a theory that was introduced by
Wiley (2000). LODAS is of relevance to this thesis as it specifically addresses the
instructional use of learning objects. Wiley (2000, pp. 8-11) criticizes the work on learning
technology standards (see chapter 3.6), and especially the LOM-standard, for not considering
the instructional design aspects of learning objects. Wiley (2000, p. 9) argues that if the
36
standard bodies do not consider instructional design, neither will teachers. He points out that:
“…instructionally grounded sequencing decisions are the heart of the instructionally
successful use of learning objects”. Beside the problems that may be caused by not
considering the sequencing aspects (i.e., the instructional combination of learning objects) of
instructional design, there will be problems regarding the scope of learning objects (i.e., how
big should a learning object be). Wiley (2000, p. 11) claims that the granularity is a key factor
for learning objects and he points out that this problem is in principal caused by the lack of a
general taxonomy for learning objects. Learning objects and learning object taxonomies are
discussed in chapter 3.6, where Wiley’s Atomic learning object taxonomy is also discussed.
The primary goal of LODAS is “… to provide scope and sequencing guidance for the design
of learning objects.” (Wiley, 2000, p. 25).
The LODAS theory is a synthesis of four existing theories: Elaboration Theory, Work Model
Synthesis, Domain Theory, and the Four-Component Instructional Design model. Wiley
claims that LODAS is primarily targeting complex cognitive problem solving and is at the
same time compatible with several existing and established instructional theories and is
especially adapted for use with learning objects. Wiley argues that the ability to adapt to
individual differences have been left out in some of the theories, but that the combinations of
instructional design theory and learning objects is a way of addressing individualization issues
and adaptability. The adaptation to existing instructional design theories in combination with
the focus on learning objects lead to a theory made up of two distinct parts: instructional
design prescriptions and learning object prescriptions. The first part considers instructional
design issues and the second part consider how those can be applied to learning objects in
creating modules for learning (Wiley, 2000).
The LODAS theory illustrates an attempt to address learning that contains a high level of
complex cognitive processing using instructional design and learning objects. As with most
instructional design it is dependent on modularity, and the more complex the cognitive
processing becomes (see chapter 3.3), the more sophisticated modular models are needed.
Merrill’s first principles of instruction (Merrill, 2002) implies that there is a need to support
“everything” in order to cover all learning needs and all situations that may occur. The
implication of this is that the models and taxonomies for modularity need to be sophisticated
and flexible, both from the pedagogical view-point, in order to support all kinds of
instructional approaches, and from the technical view-point, in order to represent all possible
instructional approaches and all types of learning. Even though the LODAS theory consists of
one part for instructional design prescriptions and one part for learning object prescription, it
does not consider the technical properties of learning objects. This is an issue that was
addressed in paper 1 by suggesting six work areas to establish technical quality criteria that
deals with technical properties of learning objects.
As previously indicated, instructional design theory, and especially the pedagogical practice
of instructional design is often considered in relation to learning theory (see, e.g., Schunk,
1991; Merrill, 2002; Ertmer & Newby, 1993; Snelbecker, 1999).
3.4.1 Learning theories and implications for instructional design
As illustrated by Figure 3-1 and Figure 3-2, learning theories support some types of learning
better than others. The differences are related to the character and level of the cognitive
processing involved and to which factors that influence learning. According to Ertmer &
Newby (1993), this has implications for the practice of learning design as well as for teaching
and learning in general. Different learning theories were compared in Table 3-1, and Ertmer
and Newby (1993) compare the effect of different learning theories for the practice of
37
instructional design. In order to do this they added two questions to Schunk’s list (see chapter
3.2.4):
- What basic assumptions/principles of this theory are relevant to instructional design?
- How should instruction be structured to facilitate learning?
The answers to those questions, based on the study presented by Ertmer & Newby (1993)
extend Table 3-1 to include the questions above (see Table 3-3).
The comparison in Table 3-3 is mainly relevant in order to determine the required properties
for representing instructional design (or equivalent) in learning technology, addressing
research question 3. This is discussed in chapter 5. It also has some relevance for addressing
of research questions 1 and 2 as well, in that it indicates the diversity that needs to be covered
by a modular approach.
Behaviorism
Cognitivism
Constructivism
What basic
assumptions/
principles of this
theory are
relevant to
instructional
design?
- Many characteristics and
assumptions are embedded in current
instructional design practice.
- Emphasis on producing observable
outcomes in students.
- Pre-assessment of students to
determine where to begin.
- Mastering early steps before more
complex, sequencing of instructional
presentation.
- Reinforcement to affect
performance.
- Use of cues, shaping and practice to
secure stimulus-response.
- Many principles are similar to
behaviorism but for different reasons.
- Feedback for guiding and
supporting accurate mental
connections.
- Analyze learner to determine
predisposition, and learner’s mental
structures for designing instructions
for ready assimilation.
- Focus on active involvement of
learner and learner control.
- Emphasis on structuring, organizing
and sequencing information to
facilitate optimal processing.
- Encourage linking to prior
knowledge.
- Understanding that students’
previous knowledge influences
learning outcomes
How should
instruction be
structured to
facilitate
learning?
- Presentation of the target stimulus
and provision of opportunities for the
learner to practice giving proper
response.
- Cues are used to trigger the proper
response and for linking stimulusresponse pairs.
- Arranging the environmental
conditions to stimulate the right
responses.
- Use reinforcement to strengthen the
association.
- Make knowledge meaningful and
help learners organize and relate new
information to existing knowledge in
memory.
- Instructions must be based on
students existing mental structures.
- Analogies and metaphors are used.
- Arranging practice with feedback to
enhance assimilation and
accommodation.
- Relating new tasks to similar
procedures.
- Knowledge is linked to the context
in which the skills will be learned and
applied, as well as to previous
experiences.
- Focus on assisting learners in
actively exploring complex topics to
construct their own understanding
that is validated through social
negotiation.
- Learner control, and manipulation
of information.
- Content is not specified and may be
derived from many sources. Not
necessary controlled by the teacher.
- Presenting information in a variety
of ways.
- Supporting problem solving that
allows learners to go beyond the
presented information and content.
- Performance goals are related to the
process rather than to content.
- Assessments focus on transfer of
knowledge and skills.
- Instruct students on how to
construct meaning and model
construction of knowledge.
- Promote collaboration, design of
authentic learning environments.
- Cognitive apprenticeship,
collaborative learning, cognitive
flexibility, social negotiation.
- Anchoring learning in meaningful
contexts through modeling and
coaching.
- Learning objectives are not prespecified and instructions are not predesigned.
Table 3-3. A summary of the main conclusions of Ertmer and Newby (1993) on the effects of
different learning theories on instructional design practice.
Ertmer and Newby (1993) suggest that there is a continuum of learning theory that ranges
from behaviorism to constructivism where the focus gradually shifts from teaching to
learning. This is a key issue because teaching can be fairly predictable while learning is not.
The sequence of instructions can be determined beforehand when using a behaviorist
approach, while that is not possible using a constructivist approach (see, e.g., Greeno, 1989).
In fact, in the latter case it is often not even possible to predict what content will be used,
38
which path learners will take, or when and how social interaction and collaboration will occur
(See, e.g., Schunk, 2003, pp. 285-328).
Most learning situations address several types of knowledge, leading to tasks that require
different levels of cognitive processing, and that span over different dimensions of knowledge
and cognitive processing, as illustrated in Table 3-2. As different learning theories support
different types of knowledge better than others this means that learning technology must be
able to support all approaches at once. This requires the technology to be flexible as well as
adaptable from the computer science perspective.
A key factor for achieving flexibility and adaptability is modularity, and a key factor for
adaptivity is the ability to represent instructional design and pedagogical process in such way
that it can be regenerated based on input from users, preferences or feedback generated by the
system (see e.g., Koper, 2004); M. Nilsson, 2001; Devedzic, 2004; Richter, Nejdl, & Allert,
2002). Alteration of the pedagogical representation may also result in that the VLE needs to
be altered at run-time.
3.4.2 Instructional design and the effects on TEL
When moving toward constructivism in the learning theory continuum (see Figure 3-2), the
learning process becomes personal and learner-centered. This makes it harder to determine
the pedagogical process in advance and to outline a generic instructional design that
represents learning as a generally applicable process. The situation becomes even more
complicated when social interaction and social knowledge building are considered as a part of
the pedagogical process. All participants in a learning situation have different background
knowledge, different mental models and assimilate knowledge in different ways. This creates
a process of joint and mutual “knowledge shaping” that involves learning content, social
interaction, testing and rejecting, and consensus building. These activities can be supported
using a number of knowledge management techniques combined with different tools for
collaboration and social interaction, but it will be virtually impossible to predict how, where,
and when the different parts and components of this process will be applied. Consequently, it
becomes impossible to describe the pedagogical process as a predetermined sequence of
activities and interactions involving digital and human resources. It will only be feasible to
design for small parts of the overall pedagogical process, suggesting best practice, while the
order of activities, the learning content, other data, and details of how interactions and
activities are managed need to be generated at runtime based on feedback from the VLE, from
the actors in the system, and from input originating from the deduction of existing and
generated data.
The concept of instructional design may appear to be overwhelmingly complicated for many
practicing teachers and pedagogues, but in its essence it can actually be compared to a wellstructured and well-written lesson or course plan. A lesson plan can be a useful tool for
teachers, as well as for students, as long as it is used in a pragmatic way. However, when
teachers teach in a classroom and realize that the situation is not the expected or that the
situation is changing for some reason, there is a possibility to step away from the lesson plan,
which constitutes the instructional design, and improvise. That is, based on input and
feedback, the instructional design can be changed at run-time to adapt to the new situation.
The new in the situation can, for example, be that students do not react as anticipated, they
turn out not to have the expected prior knowledge, or it simply turns out that the learning
objectives have to be changed as other objectives have become more important for the
students in the current context. The need for changeability may also arise as a consequence of
individualization.
39
This kind of adaptability, based on sensibility and perceptiveness, is one of the things that
characterize a good pedagogue. Adaptive and responsive behavior is feasible in a classroom
situation, where the instructional design is represented by a lesson plan (that sometimes only
exists in the head of the teacher), and the semantics of the instructional design are interpreted
and implemented at run-time by the teacher.
The situation is different in a VLE where instructional design is represented using technology
that may only support sequencing that is either hard-coded or implemented using an XMLbased meta-language, such as SCORM Simple Sequencing and Navigation (Capone,
Christensen, Falls, & Fletcher, 2004) or IMS Learning Design (Koper, 2005). Meta-languages
representing instructional design are discussed in chapter 3.10. Such approaches require the
pedagogical process to be predetermined in order to be representable in a VLE. This makes
pedagogical instruction much more inflexible as it becomes fixed or at least severely limited
and cannot be altered at runtime based on feedback from the actors in the system or based on
feedback from the system itself. However, the way that pedagogical process and instructions
are represented by technology is only one part of the problem. VLEs also need to be able to
adapt to changes that are inflicted by changes of pedagogical instructions. This affects the
architecture and design of the underlying infrastructures and systems (see discussion in
chapter 5). These are issues addressed by research question 1 and 2.
Knowledge about the history and origin of instructional design and TEL may provide an
explanation to why pedagogical instruction and TEL are implemented as they are and to why
today’s learning technology has the characteristics that it has. It may also contribute to the
understanding of some of the implications and potential problems of instructional design in
relation to its application to learning technology. It should be emphasized that it is not mainly
instructional design itself that causes the problems. The main problems are caused by how
instructional design is represented in systems, which makes it hard to represent instructional
design in VLEs while still maintaining pedagogical neutrality (see discussion in chapter 5).
Pedagogical neutrality becomes an issue in order to support learning theories that are not rulebased (algorithmic) and predictive in the way that behaviorism and to some extent also
cognitivism are. Pedagogical neutrality and flexibility are qualities that are often emphasized
and discussed in the learning object community (see, e.g., Quinn & Hobbs, 2000; Friesen,
2004a; ADL, 2004; Bratina, Hayes, & Blumsack, 2002). However, Friesen (2004b) argues
that pedagogical neutrality is hard, or even impossible to achieve with today’s technology.
Others express similar views (cf. Parrish, 2004; McCormick, 2003).
The development paths and current states of learning technology implementation are
discussed in the following chapters.
3.5 Technology enhanced learning
This chapter provides an overview of TEL and its different development paths. The purpose is
to emphasize the context within which learning theories, instructional design, and learning
technology interplays.
Much of the TEL that we see today still seem to be influenced by the ISD approach that was
used by the US military during the Second World War (see, e.g., Gagne, 1962; Saettler,
2004), which continued to evolve and develop after the war. Even though there were several
earlier attempts to use technology for learning, such as Pressey’s (1927) teaching machine
(see also Skinner, 1960), I have chosen to start with instructional system design as it marks a
turning point that led on to the branches of TEL that we are still struggling with.
The rationale for the development of ISD (later instructional design) was a need for methods
that facilitated fast training of military personnel. The training was first carried out using non40
digital technology, such as instructional films (Saettler, 2004, pp. 184-194). The more
complex the weapon systems got, the more important appropriate training became. Learning
theories originating from behaviorism (see chapter 3.2.1) were appealing, as well as suitable,
for the learning needs of the army. Most of the skills needed by army personnel were of a
typical drill and practice nature, composed out of a series of predestined, relatively simple and
straightforward tasks that could be sequenced together to form larger units of training. An
example of how this was implemented in its early days was the machines for individualized
self-instruction called “phase checks” (see, e.g., Saettler, 2004). English (1942) invented a
similar device already in 1918. The approach of decomposing learning into smaller modules
that are later sequenced together (also referred to as chunking and sequencing) is in many
ways still representative for how learning and the delivery of digital learning content is often
implemented with technology, one such example is SCORM (Capone, et al., 2004).
3.5.1 Programmed instruction and programmed learning
The military needs for education were still the driving force that gave stimulus to the
developments of instructional design and learning technology long after the war (see, e.g.,
Saettler, 2004; Snelbecker, 1999). However, ideas and methods from instructional design
were slowly transferred to education and training outside the military sector and instructional
design started to spread to other parts of society in the fifties (see, e.g., Snelbecker, 1999).
This development was still led by behaviorists such as Skinner, Gagne and others (see, e.g.,
Saettler, 2004, pp. 394-401; Snelbecker, 1999). Eventually, this led to the birth of
programmed instruction, which was implemented by the teaching machine (see, e.g., Skinner,
1958; Homme, 1960; Skinner, 1960; Meyer, 1960). Programmed instruction was based on a
modular approach where training materials were decomposed into small modules referred to
as frames. The frames were then sequenced together in a linear fashion where all learners are
obligated to follow the same linear path. Later implementations of programmed instructions
used a slightly more sophisticated, intrinsic (branched) approach supported by simple
interactivity. The consequence was that several different paths could be programmed and
learners can be directed to different paths depending on input, such as prior responses (see,
e.g., Norman A Crowder, 1960; Pask, 1960). This creates a higher degree of interactivity.
Cowder (1964) discusses the differences between linear approaches and intrinsic approaches
in detail.
Programmed instruction emphasized self-teaching by providing atomized, self-contained and
self-paced training using technology (see, e.g., Homme, 1960) even though Lysaught and
Williams (1963) also emphasized the important role of teachers. The basic principle of
programmed instruction is, in general terms, to walk through instructions in an incremental
way where learners progress is successively tested and proper responses are reinforced (see,
e.g., Orlich, 2000).
Programmed instruction later evolved into programmed learning, which could be regarded as
a variation (or a specialization) of programmed instruction. The evolvement of programmed
learning came out of a desire to implement a model that was better adapted for use in schools.
The core ideas of programmed learning were in most respects similar to the ideas of
programmed instruction, but one of the differences between programmed instruction and
programmed learning was that programmed learning had a more expressed focus on the user.
User characteristics, such as prior knowledge, were considered while programmed instruction
focused on modifying behavior. This development seems to mark a move toward cognitivistic
views of learning. A typical programmed learning application would be learning material
consisting of small chunks, still referred to as frames, and the pedagogical activities were
usually of the “fill in the blanks” types. Learning was reinforced through immediate feedback
41
on learner activity (e.g., Saettler, 2004). The distinction between programmed instruction and
programmed learning is often not made and the terms are often used as synonymous. The
differences are often unclear and ambiguous, and the ideal application of both is the automatic
teaching machine.
3.5.2 Computer-assisted instruction and computer-based training
Eventually, starting in the sixties with projects such as the PLATO-project (Bitzer, Lyman, &
Easley, 1965), programmed instruction and programmed learning evolved into computerassisted instruction (CAI), also referred to as computer-based training (CBT) or Computerbased Learning (CBL). This development appears to have been triggered by the adoption of
more advanced instructional theories that were beneficial for instructional design, such as
Gagne’s (1977) nine events of instruction that provided a tool for designing better
instructions, and Bloom’s Taxonomy (Bloom & Krathwohl, 1956) (see chapter 3.3.1) that
provided a tool for defining and categorizing educational objectives. However, according to
Saettler (2004, pp. 307-311), most of the early CAI implementations arose from the
behaviorist tradition, typically of drill-and-practice type, while CAI gradually slid toward
cognitive learning theory during the seventies.
The shift from programmed instruction to CAI also indicates a shift from mechanical learning
machines to modern computer-based approaches that were more similar to the kind of TEL
used today. Compared to what was commonly the case within programmed instruction, CAI
made better use of computer power, and used more sophisticated, intrinsic systems, for
organizing and managing content, utilizing technologies such as branching, interlinking and
sequencing. While still emphasizing characteristics like self-paced and self-contained
instructions that are managed and delivered using computers, CAI emphasized specialized
training software and interactivity. CAI incorporated a higher number of different types of
learning-related activities, such as diagnostic testing, game-like training, quizzes, student
monitoring and likewise, and was usually considered to be more efficient than previous
approaches (see, e.g., Fletcher-Flinn & Gravatt, 1995; Johnson, 1985).
During the nineties there was a shift from specialized educational software, as the carrier of
CAI, to CD-ROM and multimedia. Focus was turned toward the delivery of interactive
multimedia learning content. A significant difference was that the development process
changed and multimedia experts, rather than software developers, were producing the
learning content. Multimedia content was also less expensive to produce – even though it was
still very expensive. It also became possible for skilled pedagogues to develop their own
learning material using relatively uncomplicated multimedia development software, such as
Multimedia Lab and HyperStudio (see, e.g., Havice & Galloway, 1998; Haviland & McCall,
1999). The breakthrough of multimedia on CD-ROM also meant the breakthrough of digital
media, such as digital images, digital audio and digital video. Audio and video have
traditionally been used in education (e.g., Saettler, 2004, pp. 123-129), and the introduction of
digital media in combination with interactivity created new pedagogical opportunities – as
well as challenges. Many of the concepts from the multimedia and CD-ROM era were
transferred to the web where it created new pedagogical and technical challenges that were
bound to occur in a networked environment (see chapter 3.8.3).
The term (as well as the idea of CAI) is still in use, even if the focus is gradually changing.
Generalized VLEs are taking over more and more of the tasks of CAI, and specialized
training software is becoming generalized and incorporated into VLEs and learning content.
42
Learning content of the type that was previously delivered as specialized training software is
now being delivered as learning objects, intended for use in VLEs and especially in LMSes17.
One of the major drawbacks of specialized training software is that it is (and always has been)
expensive to develop, which makes it hard to gain profitability in the production of digital
learning content. This is also an important reason for the emphasis on reusability within the
learning object community. Learning objects are further discussed in chapter 3.6.2.
3.5.3 Web-Based Training and Web-Based Learning
In the mid-nineties, the breakthrough of the web led to an evolvement that took TEL to the
Internet in the shape of Web-Based Training (WBT), Web-Based Instruction (WBI), WebBased Learning (WBL), and similar concepts. I will use the term Web-Based Learning, for
similar reasons that I use the term TEL (see chapter 1.7.2).
The move from CAI (and similar approaches) to web-based approaches is likely to be one of
the most important occurrences in the evolution of TEL. Most significant in this development
is that the “Computer” was exchanged with the “Web” as the mediator of education and
learning, which indicates learning is now delivered via the Internet and the web. That is, TEL
in all its forms moved to the Internet. This transition marks a change in view from the local
perspective, with the personal computer in focus, to a global view with the Internet and the
web in focus. The breakthrough of WBL occurred in the latter half of the nineties and the use
of the Internet and the web as delivery channel has since then placed the focus on
collaboration and social interaction rather than on individual training, which in turn has
created new challenges for VLEs and learning technology in general (see, e.g., Piccoli,
Ahmad, & Ives, 2001). At the same time WBL is still struggling with some of the limitations
of web technology, which was not an issue before the web. A paradox of using the web for
TEL is that it introduces endless possibilities in terms of independence of time and place,
social interaction, delivery of content and services etc., but at the same time it introduces
some rather severe technical limitations that prevent TEL to actually gain full advantage of
all, in many respects still theoretical, possibilities (see chapter 3.8.3).
The evolvement of TEL and learning technology is illustrated in Figure 3-3. The figure also
illustrates how things are interconnected and how the technological development and the
pedagogical development are (from a TEL point of view) communicating vessels.
Figure 3-3. The evolvement of TEL and learning technology from a historical viewpoint.
17
Because there are many views of what is and what is not considered to be an LMS respectively a VLE I use
the term LMS as specialization or a sub-type (or branch) of VLEs.
43
3.6 Modularization of TEL
The interrelation between technology and learning, illustrated in Figure 3-3, is likely to be one
of the driving forces behind some of the recent trends within TEL. Some of those trends are
discussed from a modularity point of view in this chapter.
3.6.1 Learning Technology Standards and modularity
The increased awareness of the importance of standards has led to similar processes for
developing standards within the learning technology domain as within most other domains
with special needs. One of the cornerstones in the development of domain-specific standards
is the utilization of existing general technology standards as the basis for domain-specific
standards (Duval, 2004). While the Internet has changed system borders, and in some cases
even removed them, it is the use of open standards that has made it possible.
TEL benefits from this development as learning and TEL-services can flow more
independently of organizational and geographical boarders and vendor lock-ins. One example
is the rapidly increasing amount of free and globally available digital learning content18 from
prestigious and well-known educational institutions, a phenomenon that is analyzed and
discussed by Wiley ( 2007) and in OECD ( 2007).
Many organizations have started to realize that the management of information and ICTsystems is no longer only a matter for their own organization, it is rather a matter of being a
part of a global infrastructure where information and services are raw material that can be
used as building blocks for the creation and delivery of new services and refined information
(see, e.g., O'Reilly, 2006). The ability to share information, (or pieces of information) and
data through communication with other systems and services has become crucial for many
organizations. This is also one of the driving forces behind the so-called Web 2.0 movement.
However, Web 2.0 is an ambiguous term that has no commonly accepted definition (cf.
Anderson, 2006; Shaw, 2005; O'Reilly, 2006). In spite of this, the technologies and
philosophy that are usually associated with Web 2.0, such as social and collaborative
software, have already become essential for education (see, e.g., Salmenjoki, Paulsson, &
Vainio, 2007; Avram, 2006; Anderson, 2007).
The importance of open, international standards as a crucial component for modern
infrastructure in a global context is becoming more and more recognized and several
organizations are focusing on the development of standards for learning technology, which I
will refer to as learning technology standards. Among the most influential standard
organizations are the IMS Global Learning Consortium (IMS)19, IEEE20 and their LTSC
initiative21, Dublin Core (DC)22 for metadata, and the ISO-ITLET initiative (JTC-1 SC3623).
Most organizations, and the specifications they produce, have cross relations that are
discussed by Duval ( 2004).
One of the most influential organizations on the use of learning technology standards is the
Advanced Distributed Learning (ADL)24 (Masie, 2003). ADL is the maintainer of the
18
Often referred to as Open Educational Resources (OER) or Open Courseware.
http://www.imsproject.org/
20
http://www.ieee.org/portal/site
21
http://ieeeltsc.org/
22
http://www.dublincore.org/
23
Standards for: Information Technology for Learning, Education and Training, http://www.jtc1sc36.org/
24
http://www.adlnet.gov/
19
44
Shareable Content Object Reference Model (SCORM) (ADL, 2004). SCORM is a reference
model that prescribes how to implement standard specifications that are mainly developed and
promoted by other organizations, such as the IEEE and IMS. In addition, SCORM
implements its own application profiles of some specifications, such as metadata, and
sequencing and navigation, as well as some SCORM-specific technologies such as the
SCORM Run-Time Environment (RTE). SCORM is formally not a de jure standard, even if it
is often referred to as a standard (Masie, 2003). SCORM is important because it has a great
impact on the use of learning technology standards and on the way that TEL is delivered as it
influences how VLEs (especially LMSes) are implemented (See, e.g., Downes, 2003; Masie,
2003). The impact of SCORM has given it the status of de facto standard, resulting in that
SCORM has been implemented in many LMSes and is widely used for the packaging and
delivery of learning objects and training courses.
The main part of the standardization work has been focused on digital learning content, and
especially on standards for descriptive information (i.e., metadata), sequencing and
navigation, and packaging of digital learning content. Much of the standardization work
evolves around modular digital learning content, i.e., learning objects. The focus is slowly
changing as the content-centered standards are maturing, and VLEs, system architecture and
learning technology infrastructure in general are gaining more attention. Learner-related data
is another area that is getting more attention, and where standard specifications are being
developed, such as IMS LIP, which is an information model for learner-related data (Wilson
& Jones, 2005; IMS, 2005b).
Similar to learning objects, there are a couple of initiatives addressing infrastructure and
system architecture that have a certain focus on modularity. Most notable are the work by
IMS on General Web Services (IMS GWS) (IMS, 2005a), and IMS’ recent adoption of the
MIT’s Open Knowledge Initiatives (OKI) and the Open Service Interface Definitions (OSIDs)
(Thorne & Kahn, 2003), as well as the e-Framework25 initiative ( Olivier, Roberts, & Blinco,
2005), which was originally initiated by the British Joint Information Systems Committee
(JISC)26, and has now turned into an international project involving several countries and
organizations. The OKI OSIDs are implemented in the Sakai project27 (Counterman et al.,
2005). The emerging learning architecture models and frameworks also reflect the importance
of learner-related information, such as in the IMS GWS, as well as in the e-Framework.
Both the OKI OSIDs and the e-Framework address SOA for TEL, but approach it somewhat
differently. The main difference between the e-Framework and the OKI OSIDs is that the
OKI OSIDs are Java-centric, implementing a service architecture based on Java RMI, while
the e-Framework uses web service technology. Another important difference is the abstraction
level of the service descriptions. The e-Framework is targeting specific and detailed
pedagogical services, while the OKI OSIDs specify more general-purpose services, and are
less specific on pedagogical services. However, this may change over time. Norton (2004)
compares the OKI OSIDs and the e-Framework approaches. Norton’s comparison gives a
good overview of the two frameworks, even though the comparison is three years old and
some things have changed, the basic principles and differences remain.
Reference models and frameworks for VLEs, like the e-Framework and the OKI OSIDs, are
of relevance for this thesis as they facilitate a breakdown of functions and processes within, as
well as outside, the VLE, which creates the basic conditions for a modularization of the
infrastructure and system architecture. Even though these initiatives are still in their infancy,
25
http://www.elframework.org/
http://www.jisc.ac.uk/
27
http://www.sakaiproject.org/
26
45
SOA combined with abstract models for modularity, create the potential conditions for
modularity and flexibility for VLEs – much like the learning object concept does for digital
learning content. By modeling processes and functionality, using established design principles
and paradigms, it becomes possible to determine a framework that breaks down and isolates
services and processes, and their associated functionality can be defined and described (Erl,
2007, pp. 26-30). In papers 2, 3 and 4, a modular taxonomy and a SOA framework were
suggested and implemented when addressing research question 1 and 2. These results are
described and discussed in chapter 4 and 5. Service-orientation is discussed in chapter 3.11.
3.6.2 Learning objects
Modularization of learning processes and learning content has actually been around for a long
time. Aristotle (Klinger, 1999) suggested that learning could be subdivided into small chunks
and then linked together in hierarchal structures, i.e., chunking and sequencing. He suggested
a systemic approach using a hierarchical tree structure that could be used to describe, organize
and visualize knowledge through classification. Those are ideas that have been present
throughout the evolution of instructional design (see chapter 3.4), and similar models for
structure, based on hierarchal classification systems, using a root-node and branches, is still
used by some instructional approaches. This idea is also similar to the idea that constitutes the
basis for representing information structures using techniques like XML. Modularity, and
especially content modularity, has been important during the whole evolution of TEL from
the first attempts with programmed instruction and teaching machines until today’s webbased approaches with LMSes and learning objects.
Learning objects represent a modular approach for digital learning content. It is also the
approach most discussed and, according to some, one of the most promising approaches for
achieving a higher level of pedagogical and technical flexibility, and much of the rhetoric that
surrounds the learning object phenomenon concerns the pedagogical freedom of teachers and
students, a topic that is frequently debated and discussed within the research community (cf.
Weller, Pegler, & Mason, 2003; Christiansen & Anderson, 2004; Schluep, Bettino, &
Guttormsen Schär, 2005; McCormick, 2003; Rehak & Mason, 2003; Gibbons & Nelson,
2002). Learning objects are often seen as a path to adaptability, adaptivity, reusability,
scalability, and instructional independence (see, e.g., Gibbons & Nelson, 2002; ADL, 2001).
However, this view is also questioned and some aspects of the learning object phenomenon is
criticized. The main reason for this is that learning objects, to a great extent, have failed to
live up to the expectations (cf. Dodani, 2002; Friesen, 2004b; Parrish, 2004; Wiley, 2003;
McCormick, 2003).
The learning object concept has evolved during the last decade and a lot of research has been
carried out on learning objects and related areas. Big issues concern learning object’s relation
to instructional design (see, e.g., Wiley, 2002b; Wiley, 2000; Bannan-Ritland, Dabbagh, &
Murphy, 2002; Rehak & Mason, 2003; Douglas, 2001), and learning object definitions and
taxonomies (see, e.g., Wiley, 1999; Gibbons & Nelson, 2002; Quinn & Hobbs, 2000; Song &
Andersson, 1999; Wiley, 2002b).
The term learning object occurred in the mid-nineties and was coined by the Learning
Architecture and Learning Object task force (LALO) within the American organization
Computer Education Management Association (CEdMA). The underlying ideas of learning
objects were inspired by the object-oriented software engineering approach within computer
science. Parts of the TEL community started to realize the potential benefits of a modular
approach to digital learning content, similar to the Object-Oriented approach that has become
a de facto standard within the software development community. This led to the formation of
46
the LALO taskforce (Hodgins, 2001). The objectives of the LALO Task Force were
expressed by CEdMA as:
“The vision of the LALO task force has been to enable new and existing learning content
to be created as independent Learning Objects, such that they can be assembled in any
combination to meet an individual’s learning needs, resulting in increased personal
productivity.” (CEdMA, 1995)
The key enabler to accomplish this was to facilitate the development and use of learning
technology standards (see, e.g., Masie, 2003; Duval, 2004; Hodgins, 2000).
The definitions and the meanings of the term learning object has changed over time. Changes
have occurred when standards have matured, when implementations have shown that some
things did not work as anticipated, and depending on differences in focus or theoretical
perspectives. Several learning object definitions exist alongside each other, some definitions
are compatible with each other and some definitions are not.
IEEE made one of the first and most basic definitions of a learning object. According to IEEE
a learning object is any entity, digital or non-digital, which can be used, re-used or referenced
during technology-supported learning. IEEE exemplifies learning objects as; including
multimedia content, instructional content, learning objectives, instructional software and
software tools, and persons, organizations or events referenced during technology-supported
learning (IEEE, 2002). The IEEE definition is very wide and can be interpreted to include just
about anything, which makes the definition somewhat hard to use, and it does not actually say
anything about the properties of learning objects. The IEEE definition, which was originally
adopted from the CEdMA LALO working group, has served, however, as a basis for other
refined definitions.
Sosteric and Hesemeier (2002) define learning objects as digital objects that have a formal
educational purpose within a predestined pedagogical context. Sosteric and Hesemeier
(2002) take on a rather traditional view on learning, and there is a risk that such a view may
limit the pedagogical choices as well as the innovative aspects of using learning objects. Like
the IEEE definition, it does not say anything about technical properties. McGreal (2004)
claims that the definition should be limited down to units that practitioners already prefer to
work with and suggests a definition where Learning Objects are “…any reusable digital
resource that is encapsulated in a lesson or assemblage of lessons grouped in units, modules,
courses and even programmes”. Song and Andersson (1999) approach Learning Objects
slightly differently. They argue that Learning Objects should be regarded as decomposable,
and that there must be a separation between data, operations and the carrier of the data. They
also claim that an object should be described using a set of attributes and relationships to
other objects. Song and Andersson focus mainly on the internal structure of Learning Objects
and their relations to other objects. The Song and Andersson approach rests heavily on
experience and concepts from OOP. A problem in the learning object community is that there
are about as many definitions of learning objects as there are people working with learning
objects.
Learning objects have often been compared to Lego building blocks. Wiley (1999) criticizes
this comparison for being oversimplified. Wiley (2002a) suggests that learning objects are
more like atoms, and argues that a learning object cannot be combined with any other learning
object, and that proper composition of learning modules requires pedagogical knowledge,
subject knowledge, as well as knowledge about instructional design (cf. Friesen, 2004a).
Wiley (2002a) argues that learning objects are like atoms in that they attract some learning
objects, while repelling other. According to the atomic model, the rules for how learning
47
objects can be assembled and used are stricter and their interfaces tend to be more complex.
However, it is not important whether learning objects are compared to Lego or atoms; the
important characteristic is that the totality (the module) of the assembled learning objects is
more than the sum of its parts. Hence, an added value is created when learning objects are
assembled and contextualized.
Wiley (2002a) suggests a taxonomy for learning object-types that complies with his atomic
view of learning objects. The taxonomy focuses on the characteristics of learning objects, and
mainly addresses the type of elements contained, the type of logics contained, and the
complexity of learning objects. The taxonomy makes sense from a pedagogical point of view
where learning objects are regarded as small pieces of digital (potentially interactive
multimedia) content for learning purposes. Wiley’s taxonomy is however too abstract and
high-level to be really useful from an object and component technology or technical
implementation point of view. The main problem is that (like atoms) learning objects do not
only need to attract or repel each other pedagogically – they need to interface as well. Wiley’s
taxonomy is described in more detail below because it is widely referred to, as well as used as
a comparison in papers 1 and 2.
Wiley’s taxonomy sorts learning objects into five categories (see Figure 3-4), ranging from
the most simple Fundamental learning objects (e.g., pictures, video, and texts), to the most
complex, Generative-Instructional Learning Objects. Generative-Presentation/GenerativeInstructional learning objects can also include web pages, which could in some cases also be
argued to be Combined-Open Learning Objects. This is a problem because they can hardly be
treated as components from a component perspective (see chapter 3.11) and they are also
likely to be inflicted with a high-level of pedagogical context dependence. Fundamental
learning objects usually have a high level of context independence, pedagogically as well as
technically, and can be regarded as data objects that can be used as building blocks for
composing composite units (i.e., modules). Technically, these are not limited to learning, but
are described using metadata that loosely place them in a learning context. Fundamental
learning objects are fairly uncomplicated to handle as they are mainly used as raw building
blocks, as in cases where they are non-decomposable media objects, such as JPEG images or
MP3 audio or similar.
Learning object Fundamental
characteristic
learning object
Number of
elements combined
Type of elements
contained
Reusable
component objects
Common functions
Extra-object
dependence
Type of logic
contained in object
Potential for intercontextual reuse
Potential for intracontextual reuse
Combinedclosed learning
object
Combined-open
learning object
Generativepresentation
learning object
Generativeinstructional learning
object
One
Few
Many
Few-Many
Few-Many
Single
Single CombinedClosed
All
Single, CombinedClosed
Single, Combined-Closed,
Generative-Presentation
(Not applicable)
No
Yes
Yes/No
Yes/No
Exhibit, display
Pre-designed
instruction or
practice
Pre-designed instruction
and/or practice
Exhibit, display
Computer-generated
instruction and/or practice
No
No
Yes
Yes/No
Yes
(Not applicable)
None, or answer
sheet-based item
scoring
High
Medium
Low
Medium
High
Low
Low
Medium
High
High
None, or domainDomain-independent
Domain-specific
specific instructional &
presentation, instructional,
presentation strategies
assessment strategies
& assessment strategies
Figure 3-4. Wiley’s atomic learning object taxonomy (Wiley, 2002a, p. 17).
48
Generative-Presentation/Generative-Instructional Learning Objects on the other hand are
much more complex to handle. Partly because they are a result of someone’s authoring efforts
where several elements are aggregated into a composite module – often treated as an
instructional unit based on an intended pedagogical use and a specific pedagogical approach
including a pedagogical design and accompanying sequencing.
Taken all together, the definitions discussed above show that the common underlying idea of
learning objects (irrespective of definition) is to provide a modular approach that facilitates
the reuse and interoperability of digital learning content. In theory, learning objects are small
and reasonably context-independent (technically as well as pedagogically), platform-neutral,
and reusable chunks of content that can be aggregated and disaggregated in order to meet the
requirements of a specific learning context. Learning objects produced by different producers
are intended to work together, and the overall pedagogical context is created by the way in
which learning objects are assembled, and by the way learning objects relate to other learning
objects or learning modules. The pedagogical context is supposed to be created using
instructional design and represented using some sequencing technology, such as IMS Simple
Sequencing (IMS, 2003b) or SCORM Sequencing and Navigation (Capone et al., 2004). There
is, however, an ongoing development toward more sophisticated and fine-grained approaches
implemented using educational meta-languages that are sometimes referred to as Educational
Modeling Languages (EML) (Rawlings et al., 2002). EMLs are discussed in chapter 3.10.
As discussed above, an important objective of learning objects is to establish the conditions
needed for a high level of flexibility in terms of the composition (as well as decomposition),
delivery, use, and reuse of digital learning content. There is, however, some serious issues
associated with the ways in which learning objects are implemented – issues that prohibit
learning objects to be fully functional (see, e.g., Wiley, 2002a; Rehak & Mason, 2003;
McCormick, 2003; Douglas, 2001; Gibbons & Nelson, 2002; Friesen, 2004b; McGreal,
2004). It becomes apparent, when studying different learning object definitions that a
common problem is that most learning object definitions and taxonomies express various
properties (pedagogical and functional), but say nothing about how those properties should be
obtained, or implemented in terms of software architecture, standards or APIs for interacting
with other learning objects and with the VLE. Such interactions are essential for realizing
some of the properties that are commonly assigned to learning objects. This makes learning
objects lose contact with the original source of inspiration: the ideas inspired by OOP, and
component technology within computer science (see, e.g., Song & Andersson, 1999; Friesen,
2004b; Schluep, Bettino, & Guttormsen Schär, 2005; Douglas, 2001; Sosteric & Hesemeier,
2002). When disregarding those properties and losing this connection, the concept also loses
some of its meaning as it becomes hard (if not impossible) to actually implement many of the
properties in a way that is generic enough to fulfill the characteristics that are given to
learning objects in terms of reusability, modularity and platform neutrality.
Especially problematic are the properties that are related to the idea of learning modules as
context-independent pieces of learning content that can be combined, reused and distributed
as “integrated” packages, composed out of the small building blocks that learning objects are
intended to be. I will return to this discussion in chapter 5.
49
3.7 Virtual Learning Environments
As discussed in chapter 1.7, the term VLE is used in a broad and all-encompassing sense in
this thesis. There are some specialized branches of virtual learning environments that have
had a significant impact on how TEL is being implemented and used, and one of the most
influential concepts from that perspective is LMSes.
3.7.1 Learning Management Systems
LMSes started to occur in the late nineties and were a direct result of the increasing
application of learning objects, together with the development of learning technology
standards.
LMSes were an attempt to put learning objects and learning technology standards into a
context that was clearer and better suited for the requirements of trade and industry. This was
also likely to be one of the main reasons for the focus on training and performance rather than
on learning and education. In this respect LMSes are a natural consequence of the evolution
of TEL and instructional design that was described in chapter 3.4 and chapter 3.5. The wide
adoption of the LMS-concept has given SCORM (see, e.g., Masie, 2003; Duval, 2004) a
unique position as a de facto standard. The reason for this is most likely that SCORM
succeeded in presenting an all-encompassing, standard-based framework that was ready to be
implemented and used together with learning objects and LMSes. However, SCORM is
criticized for contributing to the preservation of some of the aggravating legacy of TEL by
being too strict and limited, at the same time it is not regarded as suitable for all learning
situations, accused of not being pedagogically neutral and focusing mainly on training and
performance (see, e.g., Kraan & Wilson, 2003; Downes, 2002; Friesen, 2004b). Even though
SCORM has evolved and become more flexible in later versions, many of the basic problems
still remain. Examples of such problems are the way instructional design is represented using
SCORM Sequencing and Navigation (Capone et al., 2004) and the way learning objects (or
Sharable Content Objects (SCO) in SCORM lingo) is defined and implemented. See chapter
3.10 and 5 for an in-depth discussion.
There are other terms that are used more or less synonymous with the term LMS. One
example is Learning Content Management Systems (LCMS). The term LMS is sometimes
even used synonomously with the term VLE.
As with most other terms related to TEL there are several definitions. Paulsen defines LMSes
in the following way:
“Learning Management System is a broad term used for a wide range of systems that
organize and provide access to online learning services for students, teachers, and
administrators. These services usually include access control, provision of learning
content, communication tools, and organization of user groups.” (Paulsen, 2002)
Paulsen’s definition is frequently used and has been adopted by MIT OKI, as well as by
several other organizations. Several other definitions exist that are similar to Paulsen’s
definition. Another example is the definition by Masie:
“Software that automates the administration of training events and contains
organizational information such as role-based curricula, learner profiles training
histories, competencies, and training resources. A learner’s development plan and jobrelated training can be stored and personalized to the individual. A LMS provides an
environment for learner registration populates courses in a catalogue, records data from
learners; and supplies reports to management. A LMS is typically designed to handle
courses by multiple publishers and content providers, both internal and external to an
50
organization. Many LMSs are designed to conform to specifications for content
interoperability.” (Maise, 2003)
I will use the broader term VLE throughout this thesis, for the reasons discussed in chapter 1,
with the exception of when LMSes are explicitly addressed as a concept. It is important to
emphasize that I do not regard the term LMS to be synonymous to the term VLE. I regard
LMSes to be a subtype of a VLE.
The definitions above indicate that LMSes have a certain focus on the administration and
management of learning, courses and learning objects, rather than on pedagogical activity
(see, e.g., Ismail, 2002). The pedagogical context and pedagogical design is often provided as
a part of the learning content, determined by the way in which the content is sequenced and
packaged. This contradicts some of the asserted basic properties of learning objects (See, e.g.,
McCormick, 2003).
LMSes are criticized for being implemented as so-called information silos (see, e.g., Downes,
2002), which means that the system does not communicate and function very well with the
surrounding world. Most LMSes are implemented from a local, isolated perspective rather
than from a global learning perspective. The effect of this is that the modular concept of
learning objects, intended for sharing, reuse and flexibility, is locked into a silo. This is also
the immediate effect for all other data that is managed and used within the VLE, such as
learner information and performance data. This effect is created by vendor lock-ins and
monolithic thinking, as well as by too restricted and unsophisticated standards (see, e.g.,
(Downes, 2002; M. Nilsson, Johnston, Naeve, & Powell, 2007; M. Nilsson, 2001). Those
issues are discussed in detail in chapter 5.
51
3.8 Enabling technologies
3.8.1 Hypertext technologies
If one single technological development would be chosen that has had an unequalled impact
on TEL, hypertext would probably be a very good candidate. Hypertext had its major
breakthrough, in the early nineties, as the technology upon which the web was built.
However, the hypertext history is older than that. The first steps were taken by Bush (1945),
who presented the idea of the Memex (memory extender). The term hypertext did not turn up
until 1965 when Nelson (1965) used it when he presented a hypertext-based Evolutionary File
Structure (ELF).
In the eighties hypertext-based systems and applications started to become publicly available.
Xerox released the Lisp-based NoteCards system in 1985 (Monty & Moran, 1986; Trigg &
Irish, 1987). The NoteCards system was unique in the sense that it was window-based and
unusually user-friendly. Another early hypertext application that became one of the most used
applications for personal computers (not only because it was bundled with all new Macs) was
Apple’s HyperCard, introduced in 1987. HyperCard was implemented using an index-card
metaphor where index-cards, organized in “stacks”, contained the data and application logics
that were programmed using a built-in programming language called HyperTalk. The
information was interlinked using hyperlinks that related information, cards and stacks to
each other. Links were represented using buttons, forms and other elements for interaction.
This was a practice that was criticized by Hall (1994) for reducing the dynamic abilities of
hypertext by hard-coding links instead of dynamic generation of hyperlinks. Although
HyperCard was presented using the metaphor of an index-card system, it was used to develop
digital learning content as well as full-fledged applications.
The first major hypertext conference (Hypertext’87) was held in 1987. The major
breakthrough of hypertext came in 1991 when the web was introduced as a global, distributed
hypertext network built on top of the already existing Internet topology (See, Nielsen, 1995).
3.8.2 Standard Generalized Markup Language (SGML)
The underlying ideas of hypertext can be implemented in several ways, and one of the more
important inventions for the wellbeing of hypertext was the SGML, introduced in the
seventies. SGML is a meta-language that can be used to define new markup languages. A
markup language that is defined using SGML does not have to be a hypertext markup
language, but defining hypertext markup languages has become one of the most important
uses of SGML. SGML was based on the work with the Generalized Markup Language
(GML) carried out in the late sixties (Goldfarb, 1996).
Applications of SGML are usually implemented using Document Type Definitions (DTD) or
schemas, which are basically a declaration of well-formed syntax for a specific markup
language. One of the major ideas of SGML (and later XML) is the concept of separating
structural markup and syntax from formatting and presentation markup. However, it is still
possible to define markup languages that do not follow this rule. The original markup
language of the web (html) is an example of this malpractice. This is something that has
created problems ever since the birth of the web, as the data part of html is not machinereadable in a way that makes it possible to process data separated from presentation and
structure. The main reason for this is that tags for defining structure and formatting
instructions are mixed with data. This means that there is no good way of knowing what is
what, and there is no well-formed syntax or semantics that can be easily extracted from the
data parts of an html-file.
52
While html is likely to be the most well known application of SGML (Nielsen, 1995), there
are also other SGML applications. One example is the Text Encoding Initiative (TEI), which
is targeting the markup of literary and linguistic texts (see, e.g., Ide & Sperberg-McQueen,
1995). Other examples of markup languages implemented as applications of SGML are
DocBook28 for technical documentation, and HyTime for multimedia (e.g., Goldfarb, 1991).
Since its birth, SGML has been used as the basis for defining complex markup languages,
often in highly specialized contexts, such as for technical documentation. Such markup often
includes both structural markup and semantic markup. There are, however, several important
issues that need to be dealt with when mixing semantics with structural markup, especially in
a distributed, eco-system-like environment such as the web (see, e.g., M Nilsson, Palmér, &
Naeve, 2002). A good practice is therefore to separate semantics from syntax and from
presentation, which is one important driving force of the Semantic Web (Berners-Lee,
Hendler, & Lassila, 2001) (see also chapter 3.9.2).
SGML is also the origin of its better-known relative: the Extensible Markup Language (XML),
which, like SGML, is a general-purpose markup language. SGML has been criticized for its
complexity, and this was one of the main reasons for the development of XML, which can be
regarded to be a simplification of SGML (e.g., J. Clark, 1997; Bray, Paoli, & SperbergMcQueen, 1998). XML is mainly intended to be used for describing the structure and syntax
of data and is commonly used for purposes such as encoding of documents, for data exchange
and for serializing data. XML consists of a family of related technologies used for different
purposes, such as XPointers, xPath, which are used for addressing in XML documents, and
the Extensible Stylesheet Language Transformation (XSLT) that is used for transformation of
XML into other XML based documents or other document formats (see, e.g., Bray, Paoli, &
Sperberg-McQueen, 1998; J. Clark & DeRose, 1999; Nussbaumer & Mistelbacher, 2002).
During its short lifespan, and because of its relative simplicity, XML has become a
multipurpose defacto standard for exchange of data and for encoding documents.
3.8.3 The Internet and the World Wide Web
The breakthrough of hypertext had a great impact on how learning processes and learning
content were organized and presented (see, e.g., Niederhauser, Reynolds, Salmen, &
Skolmoski, 2000). Chou (1999) emphasizes the relationship and similarities between
hypertext and instructional theories. He claims that knowledge about how to design
instructions is important in order to make use of the possibilities offered by hypertext.
Jonassen (1991b) takes it even further and suggests that hypertext and instructional design are
a perfect match, and that they are built on similar underlying theories that deal with how to
organize information and knowledge. This is likely to be one of the success factors for the use
of hypertext within TEL and it is also (together with its relative simplicity) most probably one
of the reasons why the web was so quickly adopted for learning and education.
Hypertext provides a clear link between learning and technology and between instructional
design and its representation. While instructions were programmatically represented by hardcoding in specialized educational software, hypertext systems, and especially the web, made it
possible to represent instructions in a much simpler way that is partly separated from, and
independent of, the components that are the building blocks of the content, i.e., instructional
units. When html is combined with scripting languages, such as JavaScript, it becomes
possible to combine hypertext with simple application logics. However, the representation of
pedagogical instruction is still bound to the context of a specific web page, web application or
other kinds of applications (such as Flash or Java Applets) delivered on the web. This
28
http://www.docbook.org/
53
problem has later been addressed through specifications such as the IMS Simple Sequencing,
and SCORM Sequencing and Navigation to allow a better separation of learning content and
pedagogical instruction.
The conditions for delivering TEL and digital learning content changed completely when the
web emerged in 1991, even if most people could not anticipate the proportions of the impact
of the web at the time. Suddenly it became possible to deliver TEL to anyone at anytime and
anywhere in the world – as long as there was an Internet connection available. The markup
language of the Internet, html, enhanced the use of hypertext and it became reasonably simple
for teachers to design their own learning content and courses on the web, mainly using static
html pages at first. User-friendly authoring tools started to emerge in the second half of the
nineties, something that boosted the educational use of the web.
In the beginning, the web was static and most of the content was delivered as static html. The
level of user interaction was limited to clicking on hyperlinks, a principle that was criticized
by Hall (1994) in another hypertext context (see chapter 3.8.1), and this critique is still of
relevance. The static nature of the web became a limiting factor and different ways of making
the web more dynamic were invented. One of the first approaches was programming using the
Common Gateway Interface (CGI). CGI-programs made it possible to execute programs
(often coded in C or Perl) at the server and send the results to the client. CGI made simple
interactions and interactivity possible by using CGI as a gateway between the web and the
application processing the data and vice versa (see, e.g., Boutell, 1996). However, CGI also
added a number of bottlenecks and it suffers from performance problems, which led to the
development of more sophisticated technologies. To some extent, the performance problems
of CGI seem to be theoretical. Studies, such as the one by Varsha, Hansen, & Reeser (2002),
shows that the performance of different web programming technologies depends on the type
of web application, which means that CGI actually performs very well in some cases.
However, one of the more severe problems of CGI is related the http-protocol. The httpprotocol is a stateless protocol, which means that it handles every request to the server as a
new request and every transaction is independent of previous transactions. This means that the
connection to (and state off) the server is lost as soon as a request is taken care of. This
becomes a problem, as sessions29 usually need to consist of a number of transactions that are
dependent on each other, which makes it harder for programmers to manage complex sessions
with multiple interactions over a period of time. At first, this was dealt with directly by
programmers using different solutions such as hidden fields and cookies. However, such
solutions were not satisfactory because web-based applications were becoming critical, and at
the same time were growing rapidly while becoming dependent on other systems, both within
and outside a single organization. This increased the need to find solutions that could provide
consistency, scalability and stability.
The problems described above have mainly been addressed using two approaches. The first,
and most widespread approach has been through the construction of generic application
development frameworks, such as Microsoft .Net and Enterprise Java, compared by Sessions
(2001), and their corresponding application servers. A second approach (usually combined
with the first approach) has been the use of design patterns, such as the Session Object
Pattern, described by Yoder and Barcalow (1997), for managing sessions states in advanced
web applications. Many design patterns have become best practice and are even built into the
application frameworks. New constructions to overcome problems caused by the limitations
of web-based technology are developed all the time.
29
I have chosen to use the term session instead of the term user session as sessions containing multiple
transactions can be held between two or more systems as well as between a user and a system.
54
The development described above mainly targets the server-side of web applications, but
much development has been done on the client-side as well. The main objective of most of
the technologies on the client-side is to find ways to construct rich and responsive user
interfaces, which are needed in order to give web applications a similar look-and-feel that
users are used to from desktop applications.
While issues on the server, mainly concerning the interaction between server and client, have
been addressed with application frameworks, clients are more heterogeneous and harder to
handle, because the provider of a service is not in control of the client and problems caused by
different web browsers, different software versions, different operating systems and different
types of clients, has to be dealt with. Efforts have been made to develop rich interfaces using
plug-in based technologies such as Java Applets and Flash, but there is always a problem of
being dependent on a third-party plug-in. Plug-ins do not run in all browsers, there are
problems caused by different installed versions and so on.
Dynamic web pages and dynamic html (dhtml) were introduced in the late nineties as an
attempt to tackle some of the limitations of the web. Dhtml made it possible to create dynamic
behavior on the client and this made it possible to create slightly richer user interfaces that can
be executed in the web browser. Dhtml is a combination of html, the Document Object Model
(DOM) and JavaScript, which was introduced in 1996 by Netscape, according to Wikipedia
(2007). Fairly dynamic web applications could be created using dhtml in combination with
server-side technologies such as CGI, Java, PHP and likewise, but combining these
technologies with dhtml added to the performance issues discussed above (see, e.g., Iyengar,
MacNair, & Nguyen, 1997). JavaScript was later standardized under the name ECMAScript30
(ECMA, 1999), but there were still incompatibility problems and different approaches were
needed for different browsers and for different browser versions as their implementation of
both html and ECMAScript differed – clearly illustrated by the Browser Compatibility Chart
from 1999 (Smith, 1999).
Recently there have been consolidations in the development of more structured approaches
for rich client functionality and one of the most well-known approaches is Asynchronous
JavaScript and XML (AJAX), which can be seen as the successor of dhtml (Paulson, 2005).
Ajax is not a single technology, but rather a combination of existing technologies, such as
JavaScript, html, XML, CSS and DOM. The basic principle of Ajax is that requests, using
XMLHttpRequest, to the server and the resulting exchange of data are handled in the
background (see, e.g., Garrett, 2005). The effect is that the application feels more responsive
and dynamic, as the user does not need to wait for the server to respond upon request. The
application logic to manage this is handled using JavaScript and data is transferred as XML.
There are however alternative approaches, such as JavaScript Object Notation (JSON)31 that is
an object-based lightweight data interchange format that can be used to serialize data in Ajax
applications. The availability of free and open libraries and frameworks for Ajax has
increased parallel with the increased interest for Ajax. This, together with emerging best
practice and patterns, indicates that Ajax is an interesting option to explore for rich
functionality in web browsers. However, it remains to see how the web browser compatibility
issue can be handled compared to dhtml.
The possibilities and limitations of web applications, as well as different approaches to deal
with those limitations, are of relevance to this thesis as they influence the way in which webbased VLEs are implemented, and hence, also the ways in which the research questions
presented in chapter 1.3 can be addressed.
30
31
ECMA-262
http://www.json.org/
55
3.9 Metadata and meta-languages
Metadata are essential for addressing all three of my research questions and can be seen as the
“glue” that keeps the infrastructure together. The importance of metadata was not primarily a
choice that was made from start; it is rather a consequence of the modularization of learning
technology. The need for sophisticated metadata boiled down to being an important common
denominator for modularity (cf., Erl, 2007; Nagarajan, 2006a; Y W Sim, Wang, Gilbert, &
Wills, 2005). This chapter illuminates the underlying theories and concepts for metadata in
order to establish a framework for understanding how and why metadata are important for my
research and for modularity in general.
3.9.1 The role and use of metadata
Metadata are one of the most prioritized areas for learning technology standardization (see
chapter 3.6.1). The reasons for this are manifold; metadata are needed in order to describe
digital resources (including learning objects), for cataloguing purposes, for use in repositories
from where resources can be located and retrieved, for describing specific properties and
dependencies, such as technical or pedagogical properties and dependencies, and so on (see,
e.g., Haynes, 2004; Duval, 2004). Metadata can be added and used by different actors – both
humans and systems. The metadata usage described above is rather frequent and well-known
application fields for metadata, and represent a traditional view of metadata as data about
data. Metadata are regarded as more or less objective descriptions of resources, supplied and
maintained by a known and approved metadata provider.
The metadata usages described above are obviously of importance, but other important
application fields for metadata are emerging and require more sophisticated approaches. The
amount of digital resources are increasing at a rate that makes it important to describe
resources and data in ways that are semantically precise to eliminate redundancy as well as
for enhancing the abilities for automatization of data processing through techniques such as
reasoning. The line between what are metadata and what are data is often ambiguous and
there is a need to be able to represent data and metadata in the same way. Even though I use
the term metadata when addressing Semantic Web technology it is important to emphasize
that Semantic Web technology is applicable to other data as well. This means that machinereadable, and machine-processable semantics are becoming increasingly essential to fulfill
those needs, and the Semantic Web addresses such needs (see, e.g., Tim Berners-Lee,
Hendler, & Lassila, 2001; Warren, 2006).
Machine-processable semantics are useful for traditional metadata applications as well. Even
simple semantics, such as the ability to distinguish between different meanings of common
terms, is useful in many situations. The more resources that are available, the more useful
machine-processable semantics become. It soon becomes overwhelming to search large
amounts of resources without a good management of metadata and an explicit meaning of
common terms. There is a risk that users drown in the amount of information that becomes
the result of badly designed queries in combination with poor metadata.
As an example: when searching the Library of Congress book catalogue using the word
“plane” a result-set containing 5709 catalogue entries is returned. Those entries contain
resources spanning over a large number of subjects, all with their own meaning of the word
“plane”, such as plane as: plane surface, airplane, level, sailing, plane down, smooth, etc. If
metadata (and query) semantics would be expressed in a machine-processable way, the query
could be limited to only comprise one meaning of the word “plane” and by that limiting the
number of returned records. This is just one simple example of how machine-processable
metadata semantics are useful. Semantics are related to the context in which it is used and to
56
the ontology that is associated with that specific context. A lightweight ontology approach to
overcome some of the drawbacks of traditional directory-based approaches, by adding formal
semantics, is suggested by Zaihrayen et al. (2007).
In a learning situation, metadata often need to be provided and maintained by different users
with different roles. An example of this is given in paper 6, where teachers and students are
encouraged to contribute to the metadata that are related to curricula. This transforms
metadata from the traditional view as being authoritarian and regarded as objective, to also
being democratic, context dependent, and subjective, evolving in a living and constantly
changing eco-system. Nilsson, Palmér, & Naeve, (2002) refers to this as a metadata ecology.
The technology that underlies the Semantic Web appears as suitable for an ecological
approach to metadata (see, e.g., M Nilsson, Palmér, & Naeve, 2002; M. Nilsson, 2001; M
Nilsson, Naeve, Duval, Johnston, & Massart, 2008; Tim Berners-Lee, Hendler, & Lassila,
2001; Palmér, Naeve, & Paulsson, 2004; Shadbolt, Hall, & Berners-Lee, 2006).
Shadbolt, Hall, & Berners-Lee (2006) emphasize that machine-processable semantics
becomes increasingly essential when there is a need to connect different pieces of data and
metadata based on their semantics. Machine-processable semantics facilitate information
integration from different and independent sources using a common standard for representing
data, i.e., the Resource Description Framework (RDF). Information integration can be
accomplished using different techniques for matching and filtering, ranging from simple
matching of terms, to the use of advanced logics and reasoning based on formal ontologies
(see, e.g., Fensel, Van Harmelen & Davies, 2003).
Beside the use of RDF for data integration, the ability to describe components and services
using machine-processable semantics is essential in modular environments (see, e.g.,
Nagarajan, 2006a; Nagarajan, 2006b; Lama, Arroyo, Sicillia, & Lópes-Cobo, 2006; Yu,
2007). Components, in terms of services and functionality, need to be described in different
ways depending on the context in which they are used. Different descriptions are needed for
different purposes and in different contexts. By adding machine-processable semantics to
service descriptions it becomes possible to dynamically match services (i.e., components)
with other services for coupling into composite applications based on the semantically
expressed qualities and characteristics (see, e.g., Nagarajan, 2006b). This is the objective of
Semantic Web Services (see chapter 3.11.4).
3.9.2 Representing machine-processable semantics: the Semantic Web
Semantic Web technology is the most common way to represent machine-processable
semantics on the web. The backbone of the Semantic Web is RDF, a specification that is
maintained by the World Wide Web Consortium (W3C) as a part of the Semantic Web
Initiative32. In simple terms, RDF is a formal specification of a standardized way of
representing data and data models. By using a standard33 and common design principles other
RDF-aware applications can interpret and process the data model without prior awareness of
that specific model (see, e.g., Ekengren, 1998; Antoniou & Van Harmelen, 2004).
The idea of a Semantic Web was first expressed by Tim Berners-Lee & Fischetti (1999) and
originates from the idea of adding a machine-processable semantic layer on top of the current
human-processable web ( see, e.g., Antoniou & Van Harmelen, 2004, pp. 16-17). The reason
for such an approach is that the web makes sense only to humans but not to machines. Hence,
the information on the web can be read and interpreted by humans but computers cannot
make any sense of the information because there is no way of separating the data and data
32
33
http://www.w3.org/2001/sw/
Technically RDF is a W3C recommendation
57
semantics from its presentation and structure. Even in cases when it is possible to separate the
data, for example using XML, the semantics of the data is still only expressed in human
language and is highly context dependent. Hence, the Semantic Web adds a formal model for
expressing semantics that can be interpreted and used by computers to extract meaning from
the web. This allows information on the web to be automatically collected and processed by
computers. Semantic Web technology makes it possible to integrate data from different,
otherwise independent, data sources without converting between data formats (see, e.g.,
Warren, 2006).
The use of structured data formats on the web has increased since the introduction of XML in
1998 (see, e.g., Bray, Paoli, & Sperberg-McQueen, 1998). XML provides mechanisms for
representing machine-readable data on the web in a way that separates data and presentation
of data using style sheets. This means that the mechanisms for separating data, syntax, and
presentation are already available. However, it should be re-emphasized that XML was
designed to represent syntax and not to represent semantics (see, e.g., Tim Berners-Lee, 1998;
Decker et al., 2000). In that respect, RDF and XML complement each other, and it is common
practice to represent RDF using XML (RDF/XML) (see, e.g., Beckett, 2004) in order to take
advantage the machine-readable capabilities of XML and related technologies.
Figure 3-5. The layers of the Semantic Web and their mutual relationship (Hazael-Mas, 2003)
By adding a semantic layer, which actually consists of several semantic layers (see Figure 35), to the existing web, it becomes possible for computers to read, interpret and process
information on the web. This creates possibilities for new kinds of services that do not only
rely on traditional search engines (such as Google34), but that use logics and reasoning in
order to analyze and optimize the use of information, combining and integrating information
and data from multiple sources, as well as utilizing personal preferences for adaptation
purposes. The foremost added value of the Semantic Web is the possibility to deliver new and
advanced services on the web based on information and service integration combining
multiple sources on the fly (see, e.g., Tim Berners-Lee, Hendler, & Lassila, 2001; Shadbolt,
Hall, & Berners-Lee, 2006; Warren, 2006; Nagarajan, 2006b).
34
http://www.google.com/
58
3.9.3 The Resource Description Framework
RDF is the core technology of the Semantic Web, and the only serious competitor to RDF is
Topic Maps (TM) (Pepper, 2000). Even though TM and RDF are quite comparable, RDF is
the technology that is most widely adopted on the web. However, there are ongoing initiatives
to harmonize the use of RDF and TM (see, e.g., Garshol, 2003; Pepper, Vitali, Garshol,
Gessa, & Presutti, 2005)). The choice of using RDF was made based on its wide adoption and
on the availability of suitable tools and code libraries for RDF.
Subject
http://www.frepa.org
Resource
Predicate
dc:creator
Property Type
Object
Fredrik Paulsson
Property Value
Figure 3-6. A simple RDF-graph expressing that: Fredrik Paulsson is the creator of the
resource http://www.frepa.org.
RDF has a fairly simple model for expressing machine-processable semantics. The RDFmodel is based on object-property-value triplets and a triplet can be represented by a directed
graph, illustrated in Figure 3-6. The graph view of the RDF representation is usable for
illustrating the basic principles of RDF.
Figure 3-6 illustrates how a statement is made about a resource, in this case my website at
http://www.frepa.org/. The website (the subject) is uniquely identified using a Universal
Resource Identifier (URI). All properties and resources must be uniquely identified using an
URI. The subject has a property “created by” (using Dublin Core) that is valued by the object.
In this case the value is the creator: “Fredrik Paulsson”. All together this is called a RDF
statement, (Hayes, 2002).
The property value in the example could also be represented using a domain specific
machine-processable ontology for representing humans and human relationships in social
networks, such as the Friend Of a Friend (FOAF)35 (see, e.g., Brickley & Miller, 2007),
which its use is discussed by Li, Lina, Finin, and Joshi (2005). FOAF is defined using RDF
Schema (RDFS) (Brickley & Brickley, 2004), and the Web Ontology Language (OWL)
(McGuinness & van Harmelen, 2004). While RDF is used to represent the data models there
is also a need for ways of representing predefined domain models as ontologies. FOAF is an
example of an ontology for the domain of social networking. The concept of domain
ontologies, which originates from the Artificial Intelligence (AI) community, was later
described by Gruber (1993) as a social contract. On the Semantic Web, ontology can be
defined as a machine-processable collection of concepts and their relationships expressed
using a common language in order to create an abstract model of a specific domain. Defining
classes, properties and their mutual relationships is needed to accomplish this. RDF Schema
has limited expressive power as an ontology description language and OWL was developed to
overcome this by providing a powerful ontology language for the Semantic Web (Horrocks,
Patel-Schneider, & van Harmelen, 2003). Ontologies are often discussed in terms of formal
ontologies, but ontologies can also be of a less formal character, as illustrated in Figure 3-7.
35
http://www.foaf-project.org/
59
Figure 3-7. The relationship between different ontologies.
The expressive power of the statement in Figure 3-6 becomes much more powerful if the
creator is represented using the FOAF ontology instead of a string. FOAF contains much
more information that can be accessed and extracted; as a part of the RDF-statement in Figure
3-6 or as a part of the Semantic Web metadata ecology for use in new contexts. Grimnes,
Edwards, & Preece, (2004) point out how such ontology-based RDF statements, using the
FOAF ontology, can be used for creating adaptive and dynamic social networks.
Even though the RDF-model is simple, it is capable of expressing sophisticated semantics that
can be used to extract a fair amount of machine-processable meaning. This is exemplified by
the use of FOAF in social networks (see, e.g., Grimnes, Edwards, & Preece, 2004; Li et al.,
2005). It is possible to express complex semantics about any resource by combining several
simple RDF statements (see Figure 3-8).
http://www.frepa.org
Gustaf von Dewall
dc:creator
dc:creator
Fredrik Paulsson
http://www.vondewall.com
foaf:phone
site-contributor
uri://uri.to.my/phone.html
Kale Kula
Figure 3-8. Several RDF statements can be combined to express sophisticated machineprocessable semantics.
The Semantic Web is a fairly new phenomenon that rests on a technology base which, in
many respects, is still unchartered territory. There are still many issues to be addressed and
problems to be solved.
3.9.4 RDF and XML
As emphasized in previous chapters, the purpose of RDF is to represent semantics while the
purpose of XML is to represent syntax. This is an important distinction to make in order to
understand the relationship between RDF and XML (see Figure 3-5). A common misbelief is
that RDF is a competitor to XML and that they are contradictive because of that. This is
however not true - RDF and XML complement each other, and one of the strengths of RDF is
its ability to be represented using XML. As such, XML is the carrier of syntax, i.e., says
nothing about information semantics only about the structure, while RDF is the carrier of
60
information semantics. Listing 3-1 gives an example of how RDF is represented using
RDF/XML. The XML representation of RDF is called RDF/XML, and it is the standard way
to represent RDF data in RDF aware systems (see, e.g., Beckett, 2004). One important reason
for the use of RDF/XML is that it makes it possible to utilize several powerful XML
technologies together with RDF. Such an example was given when addressing research
question 3 (see papers 5 and 6), where XPointers and XML DOM were used to determine
identifiers for RDF annotations.
<?xml version=1.0” encoding=”UTF-16”?>
<rdf:RDF
xmlns:rdf=”http://www.w3c.org/1999/02/22-rdf-syntax-ns#”
xmlns:frepadomain=”http://www.frepa.org/rdf-ns#”>
<rdf:Description rdf:about=”http://www.frepa.org”>
<frepadomain:site-creator>
Fredrik Paulsson
</frepadomain:site-creator>
</rdf:Description>
</rdf:RDF>
Listing 3-1. An XML-representation of the RDF-graph in Figure 3-6 (without the ontology
references).
3.10 Meta-languages for learning design
When TEL evolved, and the focus changed from specialized training software and CD-ROM
to the Web, a need arose to separate the representation of pedagogical design and process
from the VLE and the learning content.
A common approach to accomplish this separation has been through the use of special
purpose meta-languages, also referred to as Educational Modelling Languages (EML)36 by
Rawlings, Koper, van Rosmalen, Rodrígues-Artacho, and Lefrere (2002). Generic metalanguages for representing pedagogical instruction and process are a rather new phenomenon,
and such representation has traditionally been hard-coded into digital learning content and
educational software, which can be argued to be a type of digital learning content. This is, for
example, the case in most CD-ROM productions.
EMLs have emerged during the last 10 years and there are several different EMLs available.
The purpose of an EML is to provide the means for machine-readable representations of
pedagogical design and pedagogical process. In this thesis, the term EML is used as a general
and collective term for all special purpose meta-languages for TEL, as defined by Rawlings et
al., 2002) (see below).
Some EMLs, such as the IMS Learning Design (IMS, 2003a), are developed from a generalpurpose perspective, while others originate from a certain discipline and are developed for a
specific purpose, such as the eLesson Markup Language (eLML) (Fisler & Bleisch, 2006).
The eLML was developed by the Geographic Information Technology Training Alliance
36
Not to be confused with the Educational Modelling Language that was developed by the Open University of
Netherlands (OUNL-EML), and refers to their specific implementation of an educational meta-language.
61
project (GITTA) and addresses learning in, and about, Geographic Information Systems (GIS).
There are a number of proprietary EMLs as well, but in this thesis I have chosen to regard
EMLs as something that must be open, non-proprietary, and preferably standardized by an
accredited organization. This is motivated by that much of the idea of an EML is lost if it is
made proprietary.
An extensive survey of six EMLs was provided by Rawlings, et al. (2002). The survey gives a
good idea about the concept of EMLs that is defined as:
“…a semantic information model and binding, describing the content and process within
a ‘unit of learning’ from a pedagogical perspective in order to support reuse and
interoperability.”
The definition is general enough to include most of the properties that are expected from an
EML.
It can be argued that specifications such as the SCORM Navigation and Sequencing (Capone,
et al., 2004) and IMS Simple Sequencing (IMS, 2003b) do not qualify as EMLs, even though
they implement similar functionality and are partly used for the same purposes. The main
argument is that they do not include the pedagogical semantics that are needed for
representing pedagogical instruction, and they are mainly being used for simple aggregation
and sequencing (see, e.g., Rawlings et al., 2002; Koper & Manderveld, 2004).
In their study, Rawlings et al. (2002) point out that even though there are similarities between
the six surveyed EMLs, there are also differences in how they comply with the definition
above. Their conclusion was that there was only one EML that fully complied with the
definition and that was the OUNL-EML. For this reason the OUNL-EML was used as the
basis for the development of the IMS LD that is discussed below.
3.10.1 IMS Learning Design
Even though the EML survey by Rawlings et al. (2002) provides a good overview of EMLs,
the fast pace of development during the last couple of years has made it a bit outdated. EMLs
are consolidated and one reason for this is that the OUNL-EML, which was found to be the
most versatile (as well as extensive) in the survey, was adopted by IMS as the basis for the
development of the IMS LD (Koper, 2005). This is likely to give IMS LD a unique position
as the EML most likely to be widely adopted by LMS vendors, at least in the foreseeable
future.
The objective of IMS LD is to be general enough to describe and represent any pedagogical
approach. Koper and Manderveld (2004) refer to this level of abstraction as a pedagogical
meta-model, i.e., an abstract model for representing pedagogical design, which they refer to as
Learning Design. IMS LD provides a data-model, a conceptual model, and the meta-language
needed for expressing pedagogical models and their associated processes in a machinereadable format. This is done using a formal notation together with an XML-binding.
However, the XML-binding results in the same kind of problems as with XML-based
metadata that were discussed in chapter 3.9. The problems are similar and related to the
different natures of graph-based versus document-based metadata, which are unable to
represent machine-processable semantics in a standard way. Hence, there is a risk of
preserving some of the problems associated with representing pedagogical design and
pedagogical process, discussed in chapter 3.4.
An important characteristic of IMS LD is that it does not have an explicit single learner focus
or an explicit group learning focus. This means that IMS LD supports group activities and
collaborative activities, as well as single learner activities. One effect of this, together with the
62
ambition to support a multitude of pedagogical approaches, is that it suggests a shift of focus,
from pre-packaged, self-phased learning, to learning that takes place in a social context, and
learning activities that involve groups of learners as well as individual activities and processes
(e.g., Koper & Tattersall, 2005). This shift in focus is also intimated by a shift in terms.
Instead of using the term “instructional design” the term “learning design” is consequently
used when refereeing to pedagogical design in relation to IMS LD. Even though IMS LD and
other EMLs originates from instructional design theory, IMS LD indicates a shift toward
approaches that focus more on learning and less on training and performance.
The metaphor of a play or a film script can be used to describe how IMS LD works (see, e.g.,
Koper, 2005; Olivier & Tattersall, 2005). As such a metaphor implies; a learning design is
constructed around a number of roles held by different actors in the system. Based on their
respective roles, actors go through a number of activities within the setting of the script.
These activities are either sequenced or based on the simple logic that is allowed by IMS LD.
The script metaphor works well for illustrating the main characteristics of IMS LD, but it also
illustrates one of the main weaknesses, which can be argued to be a legacy from instructional
design and the representation of instructional design in VLEs. The fact that the pedagogical
instruction has to be predetermined and explicitly expressed in advance, which is also the case
with a film script, does not leave much room for improvisation or too extensive discrepancies.
There is a clear purpose and an intrinsic value in explicitly outlining the whole chain of events
in advance when working with films or plays, but for learning this is limiting in a way that
may restrain the pedagogical flexibility. However, in the case of IMS LD, it should be
emphasized that the XML-binding is an important contributor to this problem, and the fact
that IMS LD is based on an abstract model creates the means to deal with this issue. Koper
(2004) suggests an alternative approach, based on Semantic Web technology, in order to solve
some of the issues (see chapter 5 for a detailed discussion).
The structures and interrelations of the activities and processes encountered in learning
situations are often complex and thereby hard to represent in a flexible enough way, using
existing EMLs, and the representation of learning design, pedagogical processes and structure
are still expected to be predetermined. This is something that is often hard to deal with –
considering the nature of constructivist learning theories (see chapter 3.2).
3.11 Component technology and modularity
This chapter reviews some of the developments and theories within Component-Oriented
Software Technology. I will also make comparisons and draw parallels to the modular
developments within TEL, and especially related to learning objects.
Computer science has strived toward reuse and interoperability through modularity for many
years. The most well-known and used approach is likely to be Object-Oriented programming,
which has become the de facto standard in the software industry. While OOP facilitates reuse
and interoperability at a micro level (that is, within software projects), at a programming
language level and at an organizational level, there is also a need for reusability and
interoperability at a macro level, i.e., across organizations, across programming languages and
across different software and hardware platforms.
The development of OOP and the endeavor toward modular approaches eventually led to the
evolvement Component-Oriented Software Engineering. The success of Component-Oriented
Software Engineering depends, like OOP, on the decomposition of software systems.
Successful decomposition of software systems results in independent but interacting software
components that together make up the system by communicating and interacting through
63
well-defined message interfaces. Szyperski and Pfister (1997) define a software component
as:
“A software component is a unit of composition with contractually specified interfaces
and explicit context dependencies only. A software component can be deployed
independently and is subject to composition by third parties.”
Nierstrasz and Dami (1995) defines a component in a similar way, but much shorter, as a
“…static abstraction with plugs”. While Szyperski and Pfister (1997) address both technical
properties and business properties in their definition, Nierstrasz and Dami only address
technical properties. A component as a unit is characterized by that it is deployed
independently, it can be used for composition by a third-party, and it has no externally
observable state (Szyperski, 2002c). Those characteristics influence the ways in which
components can be developed and it makes components differ from object in important ways.
Szyperski (2002c) points out that objects need a predetermined “construction plan” (i.e., a
class) to be initiated, describing the initial state and behavior of the created object, while
components do not. Szyperski (2002a) suggests a categorization of components into blackbox,
greybox, respectively whitebox components. Szyperski bases his categorization on the
visibility of the implementation details of components. Blackbox components completely hide
implementation details, exposing only interfaces, while whitebox components’
implementation details are fully visible, and greybox components are somewhere in between.
Nierstrasz and Dami (1995) distinguish between objects and components by emphasizing that
a component can encapsulate any useful software abstraction, while an object only
encapsulates data and behaviors. In fact, components’ internals do notn’t need to be
implemented using an OO-approach. Nierstrasz and Dami (1995) distinguish between
methodological aspects of components and technical aspects of components. This distinction
is interesting for two reasons: It separates conceptual aspects of components from technical
aspects and technical quality issues, and it provides an interesting point of view for comparing
characteristics of software components with the characteristics of learning objects – a
discussion that is continued in chapter 5.
Software components are intended to be used as building blocks for the composition of
composite applications, and as such they are not necessarily useful in isolation. Software
components are contributing to an added value when they are used as parts in the wider
context of a larger ”machinery”, i.e., a composite application (see, e.g., Nierstrasz & Dami;
1995; Szyperski, 2002b; Weinreich & Sametinger, 2001).
While traditional software is mainly static, a modular (composite) application can change at
run-time. Components in a composite application can be regarded as self-standing,
independent parts that communicate through messages. This becomes even more pronounced
in Service-Oriented Computing (se chapter 3.11.2). Nierstrasz and Dami (1995) make a
distinction between composition-time and run-time, but emphasize that composite
applications may be recomposed at run-time based on input from other parts of the system or
on external input. These are characteristics that are relevant for the design and implementation
of modular VLEs, which was demonstrated in papers 3 and 4, and for the interrelation of the
components of a modular VLE (also demonstrated and discussed in papers 3 and 4), as well
as for the representation of pedagogical design and pedagogical process, which was
demonstrated in papers 5 and 6 by relating and connecting learning objects and the curriculum
through semantic annotation (see discussion in chapter 5).
Even though the idea of software components may seem simple enough, the reality of
composite software is complicated. Components need to work together independently of
vendor and preferably also independently of software or hardware environment, but this has
64
turned out to be far from simple to accomplish. A basic condition for this to happen is the
existence of standards (see, e.g., Szyperski, 2002b). Standards and interoperability are often
associated with trade-offs and added complexity. Much of the adoption of component
frameworks has occurred in non-vendor neutral frameworks, such as the frameworks from
Microsoft37 and SUN38, and not in frameworks based on formal standards, such as the work
by the Object Management Group (OMG)39.
As pointed out in chapter 3.6, the original source of inspiration for learning objects was
Object-Oriented software development. However, when regarding the properties ascribed to
learning objects compared to the properties and definitions of software objects, respectively
software components, it boils down to the conclusion that learning objects actually have more
in common with software components than with software objects (cf. Nierstrasz & Dami,
1995). Learning objects would probably be better off compared to components than objects:
i.e., learning components seems to be a better term to describe the intended characteristics of
learning objects.
3.11.1 Component frameworks
There are several existing component frameworks, such as Microsoft’s platform, with
technologies like COM, COM+, Common Language Infrastructure (CLI) and ActiveX, the
Java platform, with technologies like as JavaBeans, Enterprise Java, and Remote Method
Invocation (RMI), and OMGs Object Management Architecture (OMA) and the Common
Object Request Broker Architecture (CORBA) that are platform neutral. The Microsoft
framework and the Java framework seem to be the two most used and referred to frameworks,
while OMA and CORBA appear to be having a hard time. Even though CORBA can be used
for bridging between the other two platforms, it is often criticized for being too complex,
which makes it expensive and complicated to implement (see, e.g., Maffeis & Schmidt, 1997;
Henning, 2006). These are problems that are likely to occur when the trade-offs are not
compensated by other advantages, which is the case with Web-Services where trade-offs in
performance and efficiency are counterbalanced by advantages, such as global
interoperability and business advantages (e.g., Henning, 2006).
The Java framework is platform neutral as well, but Java is not vendor neutral and it is not a
standard in the same sense that CORBA is. Since version 3.0, CORBA is adopting
technologies from EJB and there is a harmonization going on between the two (Szyperski,
2002b).
While these component frameworks address software components on the client, as well as
software components at the server, there is another modular approach that has boomed during
the last decade, and that is Service-Oriented Computing, and especially implemented as
Service-Oriented Architectures (SOA) and Web-Services.
3.11.2 Service-Oriented Computing
The breakthrough of the Internet and the web has changed the nature of computing (see
chapter 3.8.3). This is an ongoing change where modularity and distribution of the
infrastructure play a central role. The perception of computer systems and their demarcations
changed as the Internet started a globalization of systems that has eventually led to an
evolutionary change in how systems are designed and implemented. SOA and Web Services
play an important role in this change. One of the most notable changes in system design
37
http://msdn2.microsoft.com/en-us/library/aa139693.aspx
http://java.sun.com/
39
http://www.omg.org/
38
65
practice is that software logic is decomposed and packaged into manageable units for
distribution as services. This change affects new system development as well as old legacy
systems that are being modernized and equipped with service interfaces that at least create an
illusion of modularity to the surrounding world. The effect expresses itself in several ways.
One of the tendencies that can be seen is that it forces the design away from SOA in terms of
Silo-Oriented Architectures, in favor of SOA in terms of Service-Oriented Architectures that
provide the means for horizontal communication and interaction that take place between
systems and loosely coupled services (see, e.g., Patrick, 2005; Arsanjani & Holley, 2006).
Erl (2007, pp. 29-30) argues that Service-Orientation is a design paradigm, in the same sense
as Object-Orientation is a design paradigm, with its own design characteristics, architectural
models, as well as design standards and conventions. Erl also emphasizes that ServiceOriented Computing is a comprehensive topic, and one of the reasons for this is that it builds
on previous work on distributed computing, adding new layers and ideas that are better
adapted to present needs. He also makes a distinction between Service-Oriented Computing
and SOA: Service-Oriented Computing consists of a number of elements that are the parts
that constitute a Service-Oriented Computing platform, and SOA is only one of those
elements. Other elements are Services, Service Composition, Service Inventory, etc. ServiceOrientation is the paradigm and SOA is the technology architecture model that implements
Service-Oriented Computing. Figure 3-9 shows a conceptual view of how the different
elements of Service-Oriented Computing interrelate.
Figure 3-9. A conceptual view of how the different elements of Service-Oriented Computing
interrelate (based on the figure by Erl (2007, p. 41).
Even though Earl’s description of Service-Oriented computing is clear and semantically
relatively unambiguous it has become common to use the term SOA in a wider sense. Sim,
Wang, Gilbert, and Wills (2005) define SOA as a “…loosely couple network of
communicating services”, while Leymann, in the foreword of Cardoso & Sheth (2006a),
defines SOA as “fundamentally it describes how service requesters and service providers can
be decoupled via discovery mechanisms resulting in loosely coupled systems”. Even though it
66
may contribute to the confusion and inconsistency that is sometimes associated with the term
SOA, it is quite a common practice. SOA is, concurrently with its growing popularity,
sometimes even used synonymously with the term “Web Services”. It is however important to
emphasize that, in this thesis, Web Services are regarded as a technology platform that can be
used for the implementation of SOA, which in turn is the architecture model upon which Web
Services are implemented (Erl, 2007). SOA can be implemented using other technology
platforms as well. For example using one of the widespread component frameworks
associated with Enterprise Java, as demonstrated in paper 3, or Microsoft .Net (without
adhering to the Web Service technology platform) or using alternative architectural styles,
such as Representational State Transfer (REST) architectures discussed by Fielding (2000, pp.
76-106). However, proprietary solutions appear to be decreasing concurrently with the
penetration of Web Services and the increasing requirements for interoperability caused by
the need of global interaction and distribution of computing that was ultimately brought on by
the adoption of Internet and web-based technologies.
There are several similarities between Service-Orientation and Component-Orientation. One
of the most notable similarities is the emphasis on software composition, rather than on
programming. Services are the subject of aggregation through a process of composition in a
similar manner as with software components. A service is provided as an encapsulated and
loosely coupled software component. Loose coupling is one of the most important
characteristics of SOA, because it is the property that separates data from the processing of
data, and with that the type of service from the software providing the service (see, e.g., He,
2003). This includes hiding the implementation details of a service from the consumer of a
service. Loose coupling is also important for modularity, as services can be “pooled” together
through composition in order to form new functionality in new contexts – such as in the
context of a VLE, involving various levels of complexity. Hence, aggregating and combining
multiple services accomplishes new services, providing functionality with an added value,
and it is the composability of services that makes them reusable. Loose coupling strives to
reduce dependencies between the service, how it is implemented, and the consumers of the
service. The level of coupling is related to the abstraction level of a service, which in turn is
affected by how efficiently the internals and details of a service are hidden when the service is
provided to consumers. This also influences the level of reusability and composability of a
service – the characteristics that create flexibility (see, e.g., He, 2003).
3.11.3 Web Services
Web Services is the technology platform that has become the most widespread for
implementing SOA on the Internet, and has quickly gained the status of being an industry
standard. The Web Service technology platform seems to have evolved as an extension of the
developments on the web that were discussed in chapter 3.8.3, and can be argued to represent
a more structured and standardized approach to web application development. The initial
hype has evolved into real-life production solutions concurrently with that the technology
platform has matured and stabilized.
Web Services are implemented and exposed as services by a Service Provider. A Web Service
can implement newly developed functionality or functionality that resides in legacy systems
and that is provided with a Web Service interface to allow the exposure of functionality to
other systems. Services are located via a Web Service directory, usually based on the
Universal Description Discovery and Integration (UDDI), in order to be used by a Service
Consumer. The discovery is dependent on metadata descriptions of the available services.
The Web Service cycle is illustrated by Figure 3-10.
67
Figure 3-10. The Web Service cycle.
The technical foundation for implementing Web Services is a stack of standards that are
needed throughout the service lifecycle (Szyperski, 2002b). These standards and their
respective use are described in Table 3-4.
Usage
Specification/protocol
Comment
Discovery
UDDI
Itself a Web Service, UDDI serves as a
directory for Web Services.
Description
WSDL, WSFL/XLANG,
others to come
Given one or more Web Services, describe
properties at a meta-level.
Access
SOAP, SOAP with
attachments
Given a Web Service instance, access it via
messages.
Transfer
HTTP, SMTP, others
Transfer SOAP-messages, incl. attachments.
Transport
TCP/IP, UDP, and others
Protocols for data transport.
Table 3-4. The core Web Service protocol stack, based on the table by Szyperski (2002b, p.
229).
3.11.4 Semantic Web Services
The importance of metadata is emphasized in most literature on Service-Oriented Computing,
(see, e.g., Cardoso & Sheth, 2006a; Szyperski, 2002a; Erl, 2007; Yu, 2007). Services need to
be thoroughly described for discovery purposes, as well as for composition purposes. The
functionality of services and service interfaces needs to be described in enough detail to avoid
any ambiguousness at discovery-time, composition-time, or at run-time. This means that
standards such as the Web Service Description Language (WSDL), in combination with UDDI
are crucial. However, the more heterogeneous the environments become, the more
sophisticated metadata are needed and it is not enough to have access to syntactic
heterogeneous metadata, there is also a need for semantic heterogeneity (Cardoso & Sheth,
2006a).
In order to address issues of semantic metadata heterogeneity for Web Services there is
ongoing work to develop what is referred to as Semantic Web Services. In simple terms,
Semantic Web Services are a combination of Web Service technology and Semantic Web
68
technology in order to create Web Services with semantic capabilities for use on the Semantic
Web. While Web Services act as components in self-describing and self-contained
applications that are described using WSDL, WSDL provides only syntactic operational
information of the Web Service and no semantic information (see, e.g., Cardoso & Sheth,
2006a); Yu, 2007, pp. 207-217). By combining Web Service technology and Semantic Web
Technology, for example, through the application of semantic annotations to Web Services,
and connecting them to ontologies that provide explicit semantic definitions, the descriptions
of Web Services become significantly more powerful (see, e.g., Nagarajan, 2006a). There are
several initiatives suggesting specifications for use with Semantic Web Services. Examples
are the OWL-based OWL-S40, the Web Service Modeling Ontology (WSMO)41, the Semantic
Web Services Framework (SWSF) (Battle et al., 2005), and the WSDL based WSDL-S42
(Cardoso & Sheth, 2006b). Semantic annotations of Web Services affect both the descriptions
of Web Services, the use of Web Services, and the discovery of Web Services. The latter
meaning that machine-processable semantics facilitate the discovery of Web Services by
helping the matching of processes and requirements to Web Services, and by that facilitating
the automatization of Web Service composition (see, e.g., Klein & Bernstein, 2001;
Rodriguez & Egenhofer, 2003). An effect of these added characteristics is that the adaptable
nature of composite applications based on Web Services extends to the potential of being
adaptive and not only adaptable. This is a characteristic that I argue to be essential in order to
achieve modular VLEs that are pedagogically neutral and that can respond to semantic
representations of pedagogical design (see discussion in chapter 5).
3.12 Emerging frame of reference
The objective of this chapter was to elucidate major theories, concepts and technologies that
influence TEL and the implementation of learning technology. The purpose has not been to
take a stand or value learning theories or other pedagogical theory. The objective was to show
that there are several available learning theories and emphasize their similarities and
differences in order to establish a relationship to the implementation of learning technology
and the different requirements that need to be met in order for learning technology to be
reasonably neutral to pedagogical theories. In order to understand those relationships there is
a need to be aware of the variety of components that influence TEL.
The focus of this thesis is on system and software architecture and the role of modularity in
VLEs and digital learning content. However, due to the nature of the research presented in
this thesis, the theoretical frame of reference will be used to illustrate and illuminate the
interdependence between pedagogical theory, in terms of how it affects pedagogical design
and pedagogical process, and how it also affects the system representation of pedagogical
process, which leads to effects for the design and implementation of learning technology and
digital learning content. This will be discussed in detail in chapter 5.
Figure 3-11 summarizes the theoretical frame of reference that has been described and
discussed in this chapter.
40
http://www.w3.org/Submission/OWL-S/
http://www.wsmo.org/
42
http://www.w3.org/Submission/WSDL-S/
41
69
Figure 3-12. Summary of the theoretical framework for the research presented in this thesis.
70
4 Summary of research results
This chapter summarizes the research results presented in the six appended papers. The
results are further discussed and analyzed in chapter 5.
The research questions formulated in chapter 1 were:
1. How can a common abstract model for modularity, which covers both the VLE and
digital learning content, be accomplished without creating incompatibilities with
existing models?
2. How can a technical infrastructure be designed and implemented according to a
common abstract model, in order to provide the necessary supporting infrastructure,
and without disqualifying legacy?
3. How can pedagogical designs and processes be represented in order to utilize
modularity to increase pedagogical flexibility and adaptability?
These research questions were addressed in the appended papers, and the research results
presented in papers 1-6 are summarized below and discussed in chapter 5.
4.1 Results of paper 1
“Establishing technical quality criteria for Learning Objects”
Paper 1 examines definitions, designs and implementations of learning objects. The study
consisted of two parts: a comparative study of learning object implementations, and a metastudy comparing learning object definitions. The results were analyzed from the point of view
of technical quality criteria. Parts of the contents (in total over 200 learning objects) from
three different learning object repositories (LOR) were studied43, focusing on technical
implementation and software architecture issues. The studied learning objects were estimated
to be representative for how learning objects are commonly designed, implemented and
distributed – even though there are of course exceptions.
The study addressed four main properties of learning objects: architecture (in terms of
separation of data, logics, presentation and implementation of interaction interfaces),
pedagogical contextualization, the use of standards, and the extent to which they are
decomposable/composable.
As discussed in chapter 3.6.2, learning objects are frequently referred to as being reasonably
context independent, freestanding, and platform neutral chunks of content that can be
contextualized and assembled into larger units (learning modules) for use within specific
pedagogical contexts. There is a conception of learning objects as a highly flexible and
modular model for learning content that has the power to change the way in which teachers
and producers approach digital learning content. The study presented in paper 1 shows that
this conception is not entirely right and that there are many unsolved issues that need to be
dealt with in order for learning objects to comply with the used definitions. Current practice
of implementing and using learning objects, compared to commonly referred to definitions
and taxonomies of learning objects, lead to the discovery that learning objects are often
ascribed a number of properties and qualities that they do not automatically possess. There
appears to be quite a discrepancy between how the learning object concept is presented and
defined, and how it is practiced. This problem usually seems to be caused by the lack of
43
The study was anonymized in order not to single out learning object producers as producing learning objects
of poor technical quality.
71
criteria stating how learning objects properties should be implemented. The learning object
concept appears to be too fuzzy and imprecise and in need of explicit, compulsory rules and
best practice guidelines.
An issue that became clear in the study was that most learning object definitions and learning
object taxonomies are too forgiving in that they allow too much to be included. An example
of this is one of the most used definitions, the IEEE Learning Object definition (see chapter
3.6.2, page 24) that includes just about anything that is digital and can be used for learning.
Another issue is that most of the studied learning object definitions stated a number of quite
advanced properties and characteristics, without saying anything about how those can be
accomplished in a conformed and standardized way.
The above issues appear to be associated with a frequent disregard of basic software
architectural considerations, directly related to the lack of technical quality criteria and best
practice. The major result of the study was the suggestion of six action areas for establishing
technical quality criteria for learning objects:
(1) There is a need for a common (more narrow) definition of what is, and what is not, a
learning object. Excluding rather than including, which also means accepting trade-offs in
order to gain a functioning concept by defining “must-have” properties and attributes. This
includes the separation of pedagogical and technical issues as stated above. Experience from
object-orientation and component technology should be considered.
(2) In connection to narrowing down definitions, there is a need for a taxonomy that complies
with the definition, and where granularities as well as special properties are regarded.
(3) Standards used for learning objects should be extended to go beyond descriptive
information, such as metadata, sequencing, and packaging to also embrace standards for
interfaces, machine-readable descriptions of technical properties and interaction interfaces.
(4) There is a need to establish standards and recommendations that address the internal use
of data formats and data structure. This suggestion is related to area 3 and is important for the
exchange of data as well as for separating data from logics and presentation.
(5) To enhance the level of decomposability and context independence, it should be
prescribed for the architecture of learning objects to be layered as a part of best practice, in
order to separate data, presentation and application logics.
(6) Pedagogy should preferably (when it is possible) be kept outside learning objects in order
to facilitate pedagogical context independence. The contextualization (the pedagogical
design) should be done when learning objects are composed into learning modules.
The results presented in paper 1 provide the basis for further addressing research question 1
and 2 by the suggestion of areas that need to be addressed in terms of technical quality. Such
work is needed in order to accomplish the abstract model and technical infrastructure that are
needed to prepare a common (and working) modular model and framework for learning
objects and VLEs.
4.2 Results of paper 2
“Virtual Workspace Environment (VWE): A taxonomy and Service Oriented Architecture
(SOA) framework for modularized Virtual Learning Environments (VLE) – Applying the
Learning Object Concept to the VLE”
The research presented in paper 2 builds directly upon the research and results that were
presented in paper 1. Paper 2 addresses research question 1 by suggesting an abstract model
for modularity that includes both digital learning content – i.e., learning objects, and the VLE.
72
Learning objects were discussed in detail in chapter 3.6.1 and VLEs and LMSes were
discussed in chapter 3.7.1.
The main research result presented in paper 2 was an abstract model for modularity, departing
from the six areas for quality criteria presented in section 4.1. Figure 4-1 illustrates the model,
which was named the “VWE Learning Object Taxonomy” – after the VWE project, within
which the taxonomy was developed. The VWE project was described in chapter 2.1.2. The
objective of the research was to address the issues that were raised and discussed in paper 1,
while still maintaining the compatibility with existing learning object taxonomies, such as the
atomic learning object taxonomy (see chapter 3.6.1).
The VWE Learning Object Taxonomy consists of two main object44 types with different
purposes, at different levels of aggregation (see Figure 4-1).
The two main object types are: simple learning objects and resource objects. Simple learning
objects usually consist of a data object, or fundamental data object (FDO), which are
equivalent to Wiley and Nelsons fundamental learning objects (FLO) (see chapter 3.6.1).
FDOs are the most basic building blocks used for the aggregation of learning objects. FDOs
usually consist of a picture, XML-data, an animation, data, etc.
Figure 4-1. A concept map outlining the VWE Learning Object Taxonomy.
Resource objects are of two types: helper resource objects and creator resource objects.
Resource objects add functionality in terms of application logic and by that they facilitate the
separation of data and application logics. Helper resource objects are used to interpret,
manipulate render and visualize data that (in the taxonomy) are simple data objects. The
relation between helper resource objects and simple learning objects are similar to the relation
between browser plug-ins, such as a viewer plug-in, and the data file to be viewed. Hence, the
purpose of helper resource objects is to “help” other objects by providing suitable application
logics. The purpose of creator resource objects is to add other application functionality and
services to the VLE or to learning objects, other than functionality in the sense that helper
resource objects provide. Creator resource objects, unlike helper resource objects, can also be
used as stand-alone. They can implement any functionality, which means that they can be
used as building blocks for modular VLEs in ways that are similar to software components.
44
Note that the term “object” is not used in the same sense as an object in OOP.
73
However, creator resource objects can also be used in aggregations with helper resource
objects and FDOs to form advanced, aggregated learning objects that are “packaged” together
with VLE functionality.
As implied above, learning objects according to the VWE learning object taxonomy can be
either singular or grouped/aggregated. Aggregations are called grouped learning objects, and
grouped learning objects are the result of an aggregation of two or more objects. As such, they
can be aggregated using any kind of objects from the taxonomy, including other grouped
learning objects. An aggregation of grouped learning objects that is prepared for a specific
learning situation is referred to as a learning module. It is at the learning module level that the
pedagogically contextualization occurs. It should be emphasized that learning modules could
constitute content as well as the VLE or parts of the VLE. Hence, according to the VWE
Learning Object Taxonomy the VLE can be assembled and aggregated in the same fashion as
learning objects are intended to be used, according to most definitions, when addressing
digital learning content. Figure 4-2 and 4-3 are partly made out of screen dumps from the
VWE VLE (see chapter 2.1.2), and illustrate how this works in practice. Figure 4-2 illustrates
how a Simple Learning Object – in this case a file describing a chemical molecule, using the
XML-based Chemical Markup Language (CML) – works in an aggregation with a Helper
Resource Object, i.e., a CML viewer. Together they form a Grouped Learning Object that is
pedagogically relatively context independent and that can be used to explore a chemical
molecule. The chemical molecule is actually retrieved from an eternal database and loaded
into the viewer. This means that the same viewer can be reused in a different pedagogical
context involving chemical molecules, i.e., no hard coding of application logic.
Figure 4-2. Illustrating the relation of a Simple Learning Object – in this case a CML/XMLfile, and a helper resource object – in this case a CML Viewer. These are kept together using
an IMS CP structure to form a unit that is a Grouped Learning Object.
The second example (Figure 4-3) shows how the same Grouped Learning Object is used in a
new, and completely different, context. A knowledge test is put together and the molecule
viewer and the CML molecule is aggregated together with a test question in the context
(aggregation) of a knowledge test that (in the example) contains three questions. In this case,
the test complies with the IMS Questions and Test Interoperability (IMS QTI) specification.
74
Figure 4-3. The relation between several VWE Objects; in this case test questions that are
grouped learning objects assembled using a fundamental learning object and a helper
resource object. The second question uses another grouped learning object to visualize
chemical molecules. The result of the assembled and sequenced questions is a learning
module – a knowledge test.
Except for the Grouped Learning Objects from Figure 4-2, the aggregations in Figure 4-3 also
consist of two Resource Objects. These are Helper Resource Objects for dealing with IMS
QTI test questions and IMS QTI tests and a Creator Resource Object that act as IMS QTI
editor and that was used to create the test. The IMS QTI editor consists of a client and a
server, exposed as a service.
4.3 Results of paper 3 and paper 4
“Suggesting a SOA-framework for modular virtual learning environments: comparing two
implementation approaches” and “A Service Oriented Architecture-framework for
modularized Virtual Learning Environments”
For clarity reasons, the results presented in papers 3 and 4 are summarized together. Each
paper has a different focus, but the results are dependent on each other and the picture
becomes complete only when the results are considered in relation to each other. The research
presented in papers 3 and 4 is a consequence of requirements that arose from the results
presented in papers 1 and 2. The papers address research question 2 by suggesting a modular,
SOA framework that facilitate the implementation of modular VLEs and learning content in
accordance with the VWE learning object taxonomy.
Paper 4 summarizes the final results of the system design and implementation of a modular
architecture-framework, called the VWE architecture-framework, which complies with the
VWE Learning Object Taxonomy, while paper 3 compares two experimental
implementations, using different technology platforms for implementing the VWE
architecture-framework. The implementations lead to prototypes of a Java-RMI based
implementation, respectively a Web Service based implementation. The purpose of
implementing the architecture framework twice, using different technology frameworks, was
to provide a proof of concept, showing that the VWE architecture framework is general
enough to be implemented using different technology platforms, and by that also general
enough to provide the basis for suggesting it as a best practice framework approach.
75
The implementations resulted in two prototypes of web-based modular frameworks (called
VWE I and VWE II) for the aggregation of VLEs and digital learning content, emanating
from the concepts and ideas expressed by the VWE Learning Objects Taxonomy. A VWE
VLE (including the content to be used) is assembled through the aggregation of components
from the VWE Learning Object Taxonomy (see Figures 4-1, 4-2 and 4-3). The metaphors of
tools and workspaces are used for addressing the components and the resulting VLE,
respectively. Java applets were used in both prototypes for the implementation of application
logics on the client, but other client-based technologies, such as flash, ActiveX or Ajax could
also have been used, even though these were not tested.
The resulting VWE SOA framework is described in Figure 4-4. The SOA framework consists
of five basic services; User Service, Message Service, Workspace Service, Tool Service, and
File Service. Tools and other services in the VWE service registry can use these basic VWE
services, in order to aggregate new services and VWE Tools.
Figure 4-4. The VWE SOA architecture-framework, showing a set of common services that
can be utilized by tools and other services.
Each service is implemented in accordance with a common pattern and design principle that
consist of a Tool-System/proxy-Service. The same design principle is used in both prototypes.
The Web Service approach, which was found to be the most versatile and suitable for this
kind of modular architecture-framework, uses UDDI (see chapter 3.11.3) for registry and
discovery of services and tools. While UDDI was used as registry from the Web Service point
of view, the SCAM RDF-store was used to manage educational metadata and some basic
annotations of VWE-tools and their associated services. The SCAM RDF-store (the SCAM
ePortfolio) was also used to manage user generated content and files together with user
created metadata.
While the structure of a VWE workspaces was expressed using an IMS CP structure, it was
implemented using a RDF-binding of IMS CP. The use of IMS CP allows workspaces to be
represented and packaged in a standard way. However, IMS CP was not optimal for this
purpose, which (together with a striving for consistency) was the reason for using an RDFbinding of IMS CP in order to create the potential to utilize the power of RDF and machineprocessable semantics.
76
4.4 Results of paper 5
“Treating metadata as annotations: separating the content mark-up from the content”
Paper 5 addresses the development of an infrastructure and prototype (the Annofolio) for
adding externally stored RDF annotations to XML documents. The resulting prototype is
based on a methodology where metadata are regarded as being layered (illustrated in Figure
4-6) with layers of metadata markup that spans structural XML markup, semantically
machine-processable metadata, and traditional bibliographic metadata (see chapter 3.9). By
suggesting a methodology and infrastructure that can be used to enhance the representation of
pedagogical instructions and processes through the addition of RDF-annotation and the
possibility to use ontologies, paper 5 addresses research question 3.
The Annofolio methodology, which separates content, structural markup and semantic
markup, and the corresponding prototype were the major research results presented in the
paper. The Annofolio methodology facilitates machine processing and atomization of the
processing of content (i.e., data), structure and metadata. The methodology makes it possible
to apply machine-readable and machine-processable statements, not only to a resource as a
whole, but also to a part of the internal structure of an XML resource, without altering the
resource. This was made possible by storing metadata as RDF-annotations, separated from the
resource using the Annotea W3C recommendation.
Figure 4-5 outlines a schematic view of the Annofolio methodology, as it was implemented as
a prototype in the experiments that were described in papers 5 and 6.
Figure 4-5. The principle of querying the system for an XML-document to browse and/or
annotate.
XPointers are used to “point-in” annotations to a specific location in an XML file. The
XPointers are specified based on the location of the annotation relative to a specific location
of the documents XML Document Object Model (DOM) – that is, the number of characters
relative to the nearest tag in the DOM tree. The Annotea compliant annotations are stored and
managed in a SCAM RDF store that is working as an Annotea server. The XML resources are
stored in an Apache Xindice XML database.
The use of Annotea and RDF makes it possible to express any kind of metadata and
semantics, according to any application profile or metadata model. The Annofolio method
makes it possible to attach any number of metadata records to a resource, or to different parts
of the internals of a resource. These metadata records are regarded as different layers of
metadata and they are managed completely independently of each other and different users
can manage different layers. The annotations can be stored in any Annotea compliant RDFstore, which means that several independent metadata stores can be set up and maintained by
different independent actors in a metadata ecology (see chapter 3.9), for example, by utilizing
an existing Annotea infrastructure together with the Annofolio. The Annofolio metadata
ecology is illustrated in Figure 4-6.
77
4.5 Results of paper 6
“Marking the National Curriculum - a new model for semantic mark-up”
The research presented in paper 6 builds upon the results presented in paper 5. The study
presented in paper 6 examines how the Annofolio prototype can be used to add metadata
annotations to the Swedish national curriculum. The use of the Annofolio prototype was
compared to a reference implementation using a traditional approach to content markup,
based on TEI, where the semantic markup is tagged inside the XML resource using XML tags
from the TEI schema. This means that the resource has to be altered in order to add metadata
and that structure metadata and metadata semantics are mixed – a practice that was discussed
in chapter 3.9.4. Research question 3 was addressed in paper 6 by showing how RDFannotations can be used to dynamically connect different components of TEL, by utilizing
metadata semantics.
The Annofolio prototype was tested against five criteria provided by the Swedish National
Agency for Education45 for the use of the Swedish national steering documents,46 together
with the needs provided by the objective of adding metadata markup to the national curricula.
These requirements were: (1) The content, semantics and meaning of the original document
must not be altered; (2) The format for data distribution must be structured, open and
application neutral; (3) Markup of a whole resource, as well as of a part of a resource, must be
supported; (4) Several application profiles and metadata models must be supported for the
same digital resource, as well as for parts of a resource; (5) It must be possible to add markup
without interfering with the original document or existing metadata.
The Annofolio and the associated methodology managed to fulfil all five criteria while TEI
prohibits metadata markup by a third party, which means that metadata may only be added by
the controller of the resource. TEI also forces alteration of the original resource in order to
add metadata markup. Approaches like TEI, where metadata and semantics are incorporated
in the XML file together with syntax and structure, also prohibit the use of more than one
application profile or metadata model at a time, hence prohibit the principle of separating
syntax and semantics.
In theory, the Annofolio methodology allows anyone to add metadata without altering the
resource or interfering with existing or future metadata. The experiment also shows that it is
possible to use several different metadata models for the same resource and at the same time.
This contributes to the creation of a metadata ecology – according to the principle that is
illustrated in Figure 4-6.
45
Skolverket (http://www.skolverket.se/), which is the Swedish agency responsible for the national curriculum.
The Swedish steering documents are a collection of documents for Swedish schools that contain the
curriculum, course plans, laws and regulation etc. Skolverket is the agency responsible for the administration of
the Swedish national steering documents.
46
78
Figure 4-6. The principle of an Annofolio metadata ecology where metadata is “pointed in”
as annotations to the XML resource, or to a specific location within the DOM-structure of the
XML resource.
A small and well-defined experiment showed that the markup method used by the Annofolio
could also be used to map the annotated curricula to digital learning resources that were
annotated with corresponding metadata. Mapping between curriculum and digital learning
resources was tested using annotations in the curriculum as input for the search and filtering
of digital learning resource metadata in a LOR. The results of the experiment showed that the
Annofolio could be used to create dynamic relations between the curriculum and learning
resources based on RDF-annotations. Such an approach makes it possible to locate digital
learning resources that are relevant in a specific pedagogical context, which would not have
been found using a traditional search methodology. As an example, digital learning resources
that were relevant for one subject (history) were, for example, found in a repository
containing digital learning resources intended for a completely different subject (chemistry).
In practice, this phenomenon occurred when annotations to the curriculum in history were
used as input to form a query to the LOR, and the returned result set suggested learning
resources intended for chemistry about how soap was made in the Middle Ages.
The prototype was also subjected to another experiment where it was tested for the
compilation of “digital compendiums”. Adding annotations to the curricula documents
marked off the scopes of the text excerpts to be included in a compendium. Compendia were
generated using XSL transformation to transform the XML excerpts from the original
documents into a new XML file. The new XML file was used to generate a number of
different output formats. The experiment indicated that the use of annotated curriculum could
become versatile and a useful component for TEL, and as a part of the metadata ecology.
The results that have been presented in this chapter will be discussed in detail in the next
chapter, where I will also discuss how the presented research relates to each other and also
suggest some ideas and strategies for further exploration.
79
80
5 Discussion and conclusions
This chapter contains a supplementary discussion and analysis of the results and conclusions
of the research presented in the appended papers, as well as of their relationship to the
theoretical framework that was presented in chapter 3. The results are discussed in
chronological order as they appear in the papers, ending with a summarizing and concluding
discussion followed by suggestions for future research directions. The reader is referred to
the appended papers for a more detailed description and discussion of the research results.
5.1 Introduction
This thesis attempts to contribute to the understanding of the relationship between pedagogy
and technology, especially contributing to the understanding of the role of modularity in
relation to pedagogical flexibility and the ability of technologies to adapt to different
pedagogical approaches and design of pedagogical activities. This ability (discussed in
chapter 3.2) has become increasingly important as the expected learning objectives and
learning outcomes are continuously changing and often consist of tasks requiring complex
cognitive processing. It is indicated by the theoretical framework (chapter 3) that different
types of knowledge leads to the formulation of different learning objectives that require
different levels of cognitive processing, leading to a need for different pedagogical
approaches that are associated to different learning theories. Hence, flexibility from both a
pedagogical and technological perspective is becoming an increasingly important
characteristic of learning technology and digital learning content.
The theoretical framework for this thesis, as presented in chapters 3.4, 3.6, and 3.11, indicates
that modularity plays an essential role in contributing to flexibility, especially in terms of
adaptability and adaptivity. This has also been proven by modular approaches within other
areas, such as within computer science (see, e.g., Nierstrasz & Dami, 1995; Szyperski, 2002b;
Weinreich & Sametinger, 2001). See also chapter 3.11.
The purpose of the studies presented in this thesis has been to examine how, and to which
extent, modularity can improve pedagogical flexibility, and to show how experience from
computer science can be applied for this purpose (see chapter 1.4).
Three research questions were formulated for this purpose:
1. How can a common abstract model for modularity, which covers both the VLE and
digital learning content, be accomplished without creating incompatibilities with
existing models?
2. How can a technical infrastructure be designed and implemented according to a
common abstract model, in order to provide the necessary supporting infrastructure,
and without disqualifying legacy?
3. How can pedagogical designs and processes be represented in order to utilize
modularity to increase pedagogical flexibility and adaptability?
In the following chapters I will discuss the results presented in the appended papers in relation
to those research questions. I will also discuss some of the strengths and weaknesses of my
results in relation to the theoretical framework and to other studies within this research area.
In the final part of this chapter I will suggest some future research and theoretical approaches
that seem appropriate in order to take the research presented in this thesis to the next level.
81
5.2 Learning objects and modularity
The concept of learning objects (discussed in chapter 3.6) is likely to be one of the most
hyped concepts within TEL during the last 10 years. In spite of the high expectations on
learning objects and large economic investments, the success has so far failed to appear.
Learning objects are often regarded as a way of enhancing reuse and interoperability (see,
e.g., Polsani, 2003), and thus, lowering the production cost of digital learning resources by
submitting them to a standard approach that allows the combination of learning objects (from
different vendors) into aggregated units to use for learning in new contexts. In business terms
this is sometimes referred to as a learning object economy, which has been predicted to reach
a multi-million Euro turnover, which in turn is expected to catalyze the development of
learning objects and by that also favor the pedagogical upsides of learning objects (see, e.g.,
Hylén, 2002; Polsani, 2003; ADL, 2004). However, the learning object economy is still
waiting to happen.
The study presented in paper 1 shows that several important reasons for these problems lie at
the technical implementation level. The study shows on the one hand that there is quite a
wide gap between the definitions of and expectations on learning objects, expectations that
are often based on functional requirements, and to some extent even a spoonful of wishful
thinking, and on the other hand the way that learning objects are implemented. This can be
compared to the situation within computer science and component-orientation, where a
similar situation was experienced years ago (see chapter 3.11). The problems that arose within
component-orientation were partly addressed by regarding component properties from two
different viewpoints: technical properties and methodological properties also referred to as
functional properties (cf. Polsani, 2003) or business properties, depending on the adopted
vocabulary. By making such a distinction it is possible to separate technical and pedagogical
(functional) aspects of learning objects, while still keeping them both in view.
My study shows that most learning object definitions are far too wide to be realistic because
they include digital resources that range from anything to everything. Similar conclusions can
be found in the work by Wiley (2002a), Polsani (2003), and many others. From a content
perspective ‘from anything to everything’ might be a good quality (excluding nothing), but
from an implementation perspective it can be devastating. One of the most significant effects
indicated by the study is that most learning objects can only be packaged and sequenced
together, but they do not actually function together, and they are unable to support and utilize
each other in the way that is possible with software components in composite applications.
Neither can they be subordinated to a common structure, nor a common look-and-feel,
because application logics, data and presentation are often mixed, and thereby locked to the
context of that specific learning object.
In paper 1, the technical properties were addressed from the point of view of technical quality.
Technical quality is a necessity in order to address research question 1, and by that a necessity
in order to address research question 2 as well. Without distinct and unambiguous technical
definitions and properties there will be no common abstract model for modularity and no
technical infrastructure will be designed and implemented according to a common abstract
model, which are the core values and objectives of the research presented that addresses
research question 1 and research question 2.
However, in reality this kind of technical quality work cannot be done in isolation and
presented in one research paper. Such work has to be subjected to a process of standardization
and consensus, preferably verified by working implementations as a proof of concepts (such
as the prototypes discussed in paper 3 and paper 4), and not only through ‘standardization by
committee’. This is also the reason why six areas for further action were suggested instead of
82
explicit quality criteria. Such work needs to be carried out in a broader context. However, the
suggested areas for quality criteria are good enough to use as temporary criteria in order to
continue the work presented in paper 2, paper 3 and paper 4, which provide a proof of concept
of the chosen approach, addressing research question 1 and 2 by defining and implementing a
learning object taxonomy and a corresponding modular architecture framework.
One of the reasons that the concept of technical quality was chosen as a starting point is that
this approach matches previous research on learning object quality frameworks done by
others, such as by Van Assche & Vuorikari (2006). It is important that the work on technical
quality can be incorporated in existing and future work in order to be adopted and gain
momentum.
5.2.1 Reviewing the technical quality criteria
The six suggested areas for further work on technical quality (see chapter 4.1) are briefly
discussed below. They should be seen as suggested points of departure for further work in
terms of research, standardization and consensus building. The reader is referred to paper 1
for additional details.
The first suggested area – a common narrow definition, defining “must-have” technical
properties and attributes, while separating technical and pedagogical properties (as discussed
above), is most likely essential for the survival of learning objects as a concept. This is also
indicated from the development of component technology that was discussed in chapter 3.11.
It is also essential for the work in addressing the other five suggested areas.
The second suggested area is closely related to the first area, but should be a separate area, as
a taxonomy, like the one discussed as a result of the research in presented in paper 2 (see
chapter 4.2), can be regarded as an refinement of a working definition that can be
implemented, while still being an abstract model. Such a taxonomy model is important as it
formalizes the definition by the use of clear, unambiguous semantics that ultimately either
includes or excludes properties in a way that is comparable to the UML modeling of computer
systems. The VWE Learning Object Taxonomy in Figure 4-1 illustrates this. This type of
“forced clarity” is promoted by the research of Song and Andersson (1999), where they take
on a rather strict object view similar to the object definition within object-orientation.
However, the objects within VWE Learning Object Taxonomy actually have more in
common with software components than software objects, a discussion that I will return to
later in this chapter.
The third suggested area addresses technology standards. Technology standards are essential
in order to realize modularity (see chapter 3.6.1). The development of learning technology
standards needs to be based on general technology standards, such as the standards for SOA
and Semantic Web Services, discussed in chapter 3.11.2 to 3.11.4. This is indicated by the
research presented in papers 3 and 4. Such development is also being done in projects like the
e-framework (B Olivier, Roberts, & Blinco, 2005) and within IMS General Web Services
(IMS GWS) (IMS, 2005a) (also discussed in chapter 3.6.1), but there is still a long way to go,
and the research still needs to prove its validity through working implementations. While it is
easy to see that this rather abstract level of standardization of learning technology needs to be
a prioritized area for technical quality, it is in reality a long and quite arduous process that
contains many pitfalls and obstacles that need to be dealt with. While standards are essential
for the development of global systems such as the web, Web Services, learning objects etc.
(see chapter 3.6 and chapter 3.8), there is always an intricate balance in finding the right level
for standardization in order to make standards work in practice. These are issues that are
83
discussed by Duval (2004), where the standards process is outlined and this intricate balance
is discussed, resulting in similar conclusions.
The fourth suggested area addresses the internal use (inside learning objects) of technology
standard specifications. This is an important issue as the decomposability of learning objects
depends on the fact that the resulting parts of decomposition must be usable in new contexts,
i.e., they need to be standard-compliant. Considering the number of available competing and
complementing content standards, such as for images or data formats based on different XML
Schemas, there is a need for recommendations and best practice for use when authoring and
developing learning objects.
The fifth suggested area addresses the software architecture of learning objects. The software
architecture of learning objects needed to be considered when addressing research question 2.
The research presented in paper 1, and especially the suggestion of area four, is a first step
toward addressing the software architecture aspects of learning objects.
One matter that was evident from the study is that the software architecture aspects of
learning objects are in general not considered, and many of the learning objects in the study
could actually be considered to be black boxes without interfaces, which can be compared to
Szyperski’s (2002a) classification, which was also discussed in chapter 3.11. A black box
component without interfaces is virtually useless from a composite perspective. There is no
way that such learning objects can be aggregated in a way that make them function together
with other learning objects. They cannot pass messages, they cannot receive messages, they
cannot be subordinated to the same look-and-feel, and they cannot be decomposed. The
exceptions in the study were some of the SCORM-compliant learning objects that were able
to interact, but only within a SCORM environment. The SCORM reference framework was
briefly discussed in chapters 3.6, 3.7, and 3.10. SCORM compliance is of course better than
the first case, but it is still problematic, because SCORM does not by necessity make learning
objects decomposable. A second problem is that SCORM is only supported within the
learning community and it is merely a reference model for learning objects and LMSs that
does not only rely on standard specifications, and by that it becomes a somewhat proprietary
and narrow framework (see, e.g., Downes, 2002; Friesen, 2004b), which may even add to the
problem (chapter 3.7.1). In addition, the need for components of the VLE to communicate and
interact with systems and services outside their own system borders is constantly increasing.
This is indicated by projects such as the e-framework, together with the occurrence of related
concepts and ideas such as Managed Learning Environments (MLEs) (JISC, 2003). The
implementation of a completely modular, service-based, and composite environment is also
discussed related to the research results presented in papers 3 and 4, and that discussion
makes this standpoint even clearer. It may even be legitimate to question whether a concept
like system borders is even applicable to the needs of such environments?
The issues described above provide strong arguments for using general technology standards
as the basis for learning technology standards and the implementation of learning technology.
The sixth suggested area addresses the representation of pedagogical instruction (i.e.,
instructional design and learning design). A concrete suggestion was made in paper 1 to
separate the representation of the pedagogical design from single learning objects. One
consequence of this is that pedagogical instruction should not be hard-coded into learning
objects, which was the case with most of the learning objects in the study. Pedagogical
processes should be implemented in a separate layer, using a standardized data format and
meta-languages for such representation (see chapter 3.10). There are currently a number of
available options, such as IMS Simple Sequencing, SCORM Sequencing and Navigation and
IMS LD. However, these standards are still rather immature and are all inflicted with their
84
respective problems and limitations. The separation of pedagogical instructions and learning
object internals is important for enhancing decomposability as well as composability in terms
of pedagogical context-independence. The suggested separation is of course not intended as
an absolute rule, rather as a suggestion for best practice where it is possible – ultimately it will
be a matter of design decisions based on common sense, design patterns and experience.
Meta-languages and data formats for representing pedagogical instruction can be used in
several ways, such as internally within a learning object as an alternative to hard-coding, or as
a “glue” that binds several learning objects together into learning modules by representing
interrelations in terms of a pedagogical context and design. The latter was exemplified in
paper 3 (see Figure 4-2 and 4-3) using IMS CP that is rather unsophisticated and mainly
designed for packaging and structure, but the principle of separation is the same.
The relations between learning objects and pedagogical contexts are a debated issue, and there
are several views of how this issue should be handled. For example, Rehak & Mason (2003)
suggests that learning objects need to have a built-in pedagogical purpose by necessity,
otherwise they would not be learning objects, while McCormick (2003) argues that pedagogy
should be kept out of learning objects for the reason of pedagogical neutrality and reusability.
The hypothesis brought forward in this thesis is that pedagogical instruction must be
technically (i.e., software architecture-wise) separated, at least into its own layer, and that
pedagogical context should be created at composition time, using EMLs.
The suggested work areas for establishing technical quality criteria were addressed by the
research presented in papers 2-5, through the implementation of a joint learning object
taxonomy for digital learning content and the VLE – the VWE Learning Object Taxonomy,
which will be discussed in the following chapters.
5.2.2 Suggesting an abstract model for modularity
The main result of the research presented in paper 2 is the VWE Learning Object Taxonomy,
briefly described in chapter 4-2. The VWE Learning Object Taxonomy is one of the most
important results presented in this thesis, as it constitutes the axis around which much of the
other research presented here revolves.
The objective of the VWE Learning Object Taxonomy is twofold: to address the issues
regarding the architecture of learning objects that were discussed above, and to outline a
model for modularity that is common for both digital learning content and for the VLE. The
purpose of the latter model is to apply the modular concept of learning objects to the VLE as
well. As discussed in paper 2 and in chapter 3, there is a mismatch between the modular
concept of learning objects and the non-modular concept of the LMS. An LMS merely
becomes a container for learning objects, complemented with tools for administration of
learning and learning content, while many pedagogical aspects are set aside or excluded in
order to be possible to implement by learning objects. This creates additional issues regarding
where pedagogical instruction should reside – referring to area 6 in paper 1.
Even though pedagogical instruction can (and in some cases must) be implemented inside
learning objects or at the learning module level (implemented as grouped learning objects
according to the VWE Learning Object Taxonomy), there is still a need for applications and
pedagogical tools to be includable as modules in the structure, because this is a natural
consequence of the design of pedagogical instructions that utilizes learning objects. However,
this cannot be easily managed as long as the needed tools are locked into the proprietary silolike system that most LMSes provide (see 3.7). In addition, it is very hard to establish where
the responsibility of the LMS ends and the responsibility of learning objects begin in terms of
providing application logics. This problem does not occur in a modular environment because
85
the VLE - or at least the instance of a VLE generated for a specific context - and its content
does not actually exist until someone composes it, which is also the major idea of learning
objects. In the same sense that Nierstrasz & Dami (1995) claim that there is a distinction
between composing-time and run-time in component-based software (see 3.11), there is a
similar distinction in a modular VLE. A modular VLE can actually be regarded as a type of
composite application. However, this distinction does not mean that the composition cannot
change after composing-time. There is an initial context that may change during run-time,
which in turn may trigger changes in the composition by actors in the system. One technical
factor that promotes this behavior is the relative lack of hard bindings between different
components, which is a consequence of loose coupling. This characteristic can be compared
with object-orientation, where it does not exist because there is no loose coupling.
The immediate effect of applying the VWE Learning Object Taxonomy is that it becomes
possible to apply a common modular model for both the VLE and the digital learning content,
which also answers my research question 1. At the same time the VWE Learning Object
Taxonomy addresses several of the architectural issues discussed in paper 1.
The VWE Learning Object Taxonomy is compatible with most of the other learning object
taxonomies and definitions that were discussed in chapter 3.6.2 (as well as several of the
definitions that were not discussed), and should therefore be regarded as a supplementary
taxonomy rather than a replacement. The reason for this is that the use of the VWE Learning
Object Taxonomy together with taxonomies addressing pedagogical and complexity aspects
of learning objects, such as Wiley’s (2002a) Atomic Taxonomy (discussed in chapter 3.6.2),
makes it possible to adopt a distinction between methodological properties and technical
properties, similar to the distinctions made by Nierstrasz & Dami (1995), as well as by others,
from the point of view of Component-Orientation. Song and Andersson (1999) draw similar
conclusions, even though they refer to more traditional properties of Object-Orientation rather
than drawing parallels to Component-Orientation. However, most existing learning object
taxonomies and definitions contradict this view as they do not address technological
properties (see, e.g., Wiley, 2002a; McGreal, 2004; Polsani, 2003). The above distinction of
properties could be a good viewpoint to start addressing some of the confusion of ideas that
were discussed in chapter 3.6.2 and it may also help in bringing pedagogy and technology
closer together without disqualifying one or the other.
Beside the technical argumentation, there are strong pedagogical reasons to have a joint
modular model as well. The fact that there are different pedagogical approaches, methods and
learning theories, which are applicable for different pedagogical requirements, put flexibility
in the front row. One of the main arguments for modularity is flexibility – technical as well as
pedagogical. The emphasis on flexibility is evident from the development of modular
approaches within computer science, such as Object-Orientation, Component-Orientation, and
Service-Orientation (see chapter 3.11), as well as from the pedagogical field, such as
chunking and sequencing (see chapter 3.4 and 3.5), and learning objects (see chapter 3.6.2).
As an example, Ertmer and Newby (1993) describe learning theories as a continuum, where
different learning theories are suitable for different types of knowledge and by that also suit
different pedagogical approaches (see chapter 3.2.4 and chapter 3.3). The situation is
complicated by the fact that one learning situation may include different learning objectives,
involving different types of knowledge, leading to several methodological approaches. The
effect of this is a need for highly flexible and adaptable VLEs.
The research presented in the appended papers addressed this need through modularity, and
by the implementation of the VWE Learning Object Taxonomy based on the SOA-based
VWE Architecture Framework. The VWE Architecture Framework (see papers 3 and 4)
86
illustrates how the VWE Learning Object Taxonomy can be translated into a software
architecture framework, and the implementation of the two prototypes proves that it is
possible to implement a modular framework that includes VLEs and digital learning content
into the same conceptual model, and by that extends the learning object concept to embrace
the VLE. Altogether, this creates the potential to achieve the flexibility and adaptability
needed in order to address different pedagogical needs and approaches, i.e., addressing
research question 2.
5.2.3 From abstract model to architecture framework
While the VWE Learning Object Taxonomy appeared to be a practicable and well-founded
approach to address research question 1, as well as the software architectural problems
discussed in paper 1 (and above, in chapter 5.2.1) it needed to be validated and confirmed.
The most reliable method to validate the taxonomy appeared to be through experimental
implementation of the taxonomy, i.e., translating the VWE Learning Object Taxonomy into
an architectural framework (see Figure 4-4) that could then be used as a blueprint for
implementation.
The VWE Learning Object Taxonomy mainly addresses technical properties of learning
objects from a client perspective or from a stand-alone application perspective. However, as
the taxonomy is supposed to embrace digital learning content as well as the VLE, there is a
need for a corresponding architecture on the server. Even though the functionality that is
addressed by the core of the architecture framework is functionality that is traditionally
implemented by the VLE, it is often useful for learning objects to be able to access and utilize
those services independently of the VLE, mainly for reusing and sharing application logics
and in order to avoid redundancy.
When designing an architecture framework, such as the one described in Figure 4-4, there is
always an issue of what level to choose in terms of the scope of the offered services. Similar
considerations are also experienced with component-orientation (cf. Szyperski, 2002c), and
service-orientation (cf. Erl, 2007) . There are basically two ways to go. One way is to offer a
smaller, but useful set of common services that are always needed in VLEs, which was the
path that was originally taken by OKI for the development of the OSIDs (see, e.g., Thorne &
Kahn, 2003). The other alternative is to extend the basic set of services, offering services that
operate on a detailed pedagogical, methodological and administrative level, which is a
strategy that was adopted for the e-framework (formerly known as the e-learning framework)
(see, e.g., Olivier, Roberts, & Blinco, 2005; (Wilson, Blinco, & Rehak, 2004). E-framework
and OKI OSIDs were discussed in chapter 3.6. It should, however, be noted that these
frameworks are somewhat difficult to refer to, because they are moving targets and constantly
changing. This is common for work in progress, which both OKI OSIDs and the JISC
frameworks can be considered to be.
For the VWE architecture framework, a choice was made to focus on a few essential services
that are needed in all VLEs. This choice was made for mainly two reasons. The first reason
was very practical and concerned resources. Each additional service to be implemented is
very resource-consuming, and the VWE project (see chapter 2.1.2) had limited resources. In
terms of proving the validity of the VWE Learning Object Taxonomy, it would not have
contributed much to the research to implement a large number of additional services, even
though it could have been a pedagogical finesse in terms of communicating the results to
pedagogues to provide even clearer examples of functionality. The second reason for
choosing to implement only a few central services was that a larger number of pedagogical
services create a risk of steering the pedagogical approach by offering a pre-selected choice of
tools and services. This is, for example, the case with the JISC framework, in which it is
87
possible to discern traces of the British national curriculum from the selection and design of
services and tools. Such influences are of course the effect of cultural and political differences
between and within educational systems as well as the result of different pedagogical pointof-views. They are the outcome of different learning theories (see chapter 3.2) and different
learning objectives (see chapter 3.3), which, in turn, influence the chosen instructional
strategies (see chapter 3.4), and so on. For those reasons, keeping the architecture framework
on a strictly technical and infrastructural level of abstraction may better preserve pedagogical
neutrality.
New viewpoints occurred during the work with the two prototypes, which make it reasonable
to discuss the above reasoning a bit further. It could for example be argued that the services in
the VWE Architecture Framework should be a part of a general architecture and infrastructure
instead. This is probably true, even though most organizations have no such general
infrastructure – at least not yet. In cases where there are already existing services, such as
identity management, identity federation, user catalogue (such as LDAP), Single Sign-On
(SSO) and likewise, these existing services can be pooled and used to form the VWE
UserService, or parts of it. The services outlined in the architecture framework are schematic,
and in that sense abstract, and some of them have already been implemented as several pooled
services. This is a design choice that was made in order to increase flexibility (the pooled
services are decomposable), and to decrease the level of complexity for VWE Tool
developers that utilizes the services and their interfaces for the composition of new tools. This
makes it possible for third party developers to work against a smaller number of interfaces,
while maintaining the possibility to use parts of an existing infrastructure, while still
maintaining the VWE service interfaces.
However, the design issues discussed above do not actually affect the relevance of the
architecture framework as a blueprint for the implementation of the VWE Learning Object
Taxonomy, because the basic principle of modularity remains. On the contrary, it actually
strengthens the concept, because reuse can be transferred to the next level by using a general
infrastructure – when the general infrastructure is ready to be used.
The core idea of the architecture framework is that it should provide a basic infrastructure that
can be utilized when adding new service-tool pairs that target specific pedagogical needs, or
other needs. This model makes it possible to implement other frameworks, such as the eframework, on top of the VWE Architecture Framework. This way, the VWE Architecture
Framework can be extended to embrace both additional frameworks and arbitrary tools in
accordance with the VWE Learning Object Taxonomy.
The results of the experimental implementations of the architecture framework, resulting in
two prototypes (VWE I and VWE II), are discussed in the next chapters.
5.2.4 Prototype implementations
As discussed in papers 3 and 4, SOA is a suitable approach for implementing modular and
distributed software architectures, a conclusion that is also supported by others (see, e.g.,
Szyperski, 2002b; Sim, Wang, Gilbert, & Wills, 2005; Makola et al., 2006). This is also the
reason why the Service-Orientation design paradigm (Erl, 2007) and SOA was used as a basis
for the blueprint that is provided by the VWE Architecture Framework (see Figure 3-9).
SOA has a number of specific characteristics that make it suitable. There actually appears to
be no realistic alternatives to SOA, because the architecture framework must function in a
distributed fashion on the Internet and in a web environment. Services are implemented as
freestanding units (functionality modules) that can be distributed over a global network,
provided as encapsulated and loosely coupled software components. Loose coupling is one of
88
the important characteristics that make SOA so well-suitable, because it prevents strong
dependencies between components (see chapter 3.11.2).
The concept of loose coupling is important for achieving modularity, because it allows
services to be pooled together to form new functionality in new contexts – such as in the
context of a VLE or to support digital learning content, i.e., as the server part of a helper
resource object, (see Figure 4-1) that supports a specific pedagogical context. In the
experimental prototypes, such pooling is accomplished by combining different resource
objects to form grouped objects, which in turn are grouped into the components that are called
VWE Tools. These are grouped into one or more VWE Workspace(s), which is an aggregation
of grouped objects forming a VLE that contains personal and shared learning spaces,
containing pedagogical (and other) tools. Each component utilizes a pool consisting of one or
more services from the VWE service framework (see Figure 4-2) and, when applicable, toolspecific services that are implemented as the server part of a tool, provided by a tool-specific
or third party server, which in turn may very well utilize other services within or outside the
VWE Architecture Framework. The reader is referred to papers 3 and 4 for additional details
on the concepts of VWE Tools and Workspaces.
5.2.4.1 Considering some common design issues
The principle of hiding the implementation details of services from service consumers makes
the implementation details of the prototypes less important. The main focus of the discussion
in this chapter will therefore be on design principles rather than on implementation details.
Even though some implementation details are discussed in paper 3 and paper 4, these were
mainly used as examples in order to illustrate common principles. The main concern is that
the prototype implementations show that the architecture framework is possible to implement
using a SOA approach – irrespective of technology framework. Taken to its extreme, even a
poor implementation would have proven this – as long as it provides the needed service
interfaces that hide it. Hence, the two prototypes verified that the VWE Learning Object
Taxonomy is valid; i.e., it can be implemented, and that it also addresses the architectural
issues outlined in paper 1. As a result, a conceptual space that is common for both the VLE
and digital learning content has been created within the demarcations of the learning object
concept, which means that research question 1 and 2 have been affirmatively addressed.
During the time-period of carrying out the research, presented in papers 3 and 4, Web
Services became more or less de-facto, and the SOA concept is sometimes even used
synonymously with Web Services. It is important though to emphasize that this is not the
understanding put forward by this thesis. The use of two different SOA approaches for the
implementation of the prototypes was a manifestation of this view and additional alternatives
are discussed later in this chapter. Hence, for the research presented in this thesis, the
distinction between Service-orientation, SOA and Web Services is important, a view that is
also argued by others (see, e.g., Erl, 2007; Sim et al., 2005).
As discussed in paper 4, the prototypes were implemented using one approach based on JavaRMI (VWE I) and another approach based on Web Service technology (VWE II). Additional
details on the respective implementations are given in papers 3 and 4, while the results of the
research are discussed below.
The blueprint for both implementations was the VWE Architecture Framework, illustrated in
Figure 4-4. Some minor adjustments were made for the Web Service implementation in order
to solve some design issues related to the change of technology framework. It was estimated
that these adjustments did not influence the results of the study in any important ways because
the basic principals remained the same.
89
While the server technology frameworks differed between the two prototypes, the clients
were implemented as Java Applets in both cases. This resulted in that the same client
instances could be used by both prototypes, with the exception of minor revisions, such as
bug fixes and small enhancements. The choice of Java Applets can be discussed, and the main
reason for the choice of Java Applets was the practicality to avoid additional technology load
brought on by the use of additional client technologies – considering that Java development
was already being done. Furthermore, there was an impending risk of increasing the
complexity, adding incompatibility problems by mixing different technologies and the
objective of the research was not solely to address such issues. However, this also meant that
there was no testing involving alternative technologies on the client, such as clients developed
using Flash or Ajax. Even though alternative clients should have worked (in theory) in the
second prototype, given the nature of Web Services, they would have been less likely to work
in the first prototype, considering the nature of Java-RMI. Another drawback of the Java
Applet approach is that it demands a browser plug-in, which may cause compatibility
problems. The use of a browser plug-in also puts an extra load on users, a problem that must
not be underestimated when working with education and learning. Such issues speak in favor
of technologies such as Ajax as the preferred client technology. Even though Ajax is still a
young and rather untested concept that has not yet been consolidated and still is afflicted with
a lot of problems, it is a promising approach to rich web clients (see chapter 3.8.3).
Papers 3 and 4 provide additional details on the implementation of the services in the service
framework. However, three of the services (the Tool Service, the Workspace Service, and the
Message Service) are less straight forward and merit some additional attention in this
discussion, mainly due to their central roles in enabling modularity by keeping things together
by acting as a “glue”.
The Tool Service is responsible for locating, loading, running, managing, and deploying tools
and content. Every component in the system has its associated metadata, which means that the
Tool Service manages extensive amounts of metadata. The mutual allocation of tasks between
the Tool Service and the Workspace Service could be discussed, referring back to the
previous discussion about the granularity of services. The responsibility for the interaction of
tools could, for example, have been allocated to the Tool Service instead.
The Workspace Service is responsible for keeping workspaces together by managing the
creation of workspaces, associating tools and workspaces, and associating workspaces with
users and groups, loading the right workspaces and tools at log-in, and so on. The Workspace
Service is also responsible for managing the interaction between tools and workspaces, while
the Message Service is responsible for passing messages between tools, or between two
instances of a tool, i.e., the sharing of data by two actors in the system. One such example is
the sharing of data between two or more users working with a collaborative whiteboard,
which was one of the example tools, bundled with both prototypes.
The Workspace Service handles the structure and packaging of workspaces. This is done
using IMS CP, or rather an IMS CP Binding to RDF. The main reason for the choice IMS CP
was that it (at the time) was the only available standard specification for representing package
structures for learning objects. IMS CP is mainly used as a packaging structure format for
export, transfer and import of learning objects. In the prototypes, a content package
containing the tools associated with a user’s workspace is downloaded to the browser when
the user logs in, and one or more workspaces are initiated and started in the web browser. The
reason that an RDF-binding was used was the need to represent more complex relations than
is allowed by IMS CP and the use of RDF allows a workspace structure representation to be
90
extended47 in order to represent complex relations and to take advantage of the machineprocessable semantics of RDF.
One major benefit of using RDF as the native format for representing workspace structures
and relations is that RDF makes it possible to dynamically connect workspace representations
to the semantic annotations of VWE Services and Tools. This basically creates the same
characteristics as Semantic Web Services (discussed in chapter 3.11.4), allowing semiautomatic discovery and matching of tools, services and content, based on the utilization of
machine-processable metadata semantics. This also facilitates the matching of content and
components in the VLE to other sources of data, such as pedagogical instructions, which was
illustrated in paper 6. The same mechanism allows the RDF representation of a workspace to
be dynamically connected to the pedagogical process of the representation – a relation that is
discussed further below.
One problem associated with representing workspace structures using an RDF binding to IMS
CP is that the IMS CP metamodel describes a packaging structure based on a recursive
container model, suitable to represent using XML, which is also the standard way to represent
IMS CP. IMS CP can be expressed using RDF, but the expressive power of RDF cannot be
utilized because of the significant differences between the two approaches (see chapter 3.9.4).
While powerful machine-processable semantics can be expressed using RDF, all the
semantics of XML lies in its hierarchical tree structure, and the intended use of XML is to
represent syntax – not semantics. Similar problems were encountered by Nilsson, Palmér, and
Brase (2003), regarding the RDF binding to LOM. A consequence of this incompatibility is
that every possible representation of a workspace can be represented using RDF, but not using
IMS CP and this is a bottleneck created by the compromise that occurred in the striving to be
standard compliant. An alternative could have been to use IMS CP in conjunction with IMS
LD, or EML as it was called at the time. This would have added expressive power, but
because the IMS LD also uses an XML binding for its standard representation, the problem
would have remained, although to a lesser extent. However, the use IMS LD is likely to have
provided a better foundation for future development, because one strength of the IMS LD is
that it is based on an abstract conceptual model that is not bound to XML, which means that it
could be expressed using RDF instead. The use of Semantic Web technology is emphasized
by Koper (2004) as a preferable path for future releases of IMS LD. Koper argues that the
increasing demands for adaptation, individualization, reuse and the increasing focus on selforganized learning networks will make semantic web technology crucial for representing
pedagogical design (referred to as Educational Modeling by Koper (2004)) and pedagogical
processes in the future.
Besides being used for representing workspace structure, RDF was also used to represent
metadata (or annotations) about resources at different levels of the VWE Learning Object
Taxonomy. In general, the more modularized a system becomes, the more sophisticated
metadata seem to be needed in order to describe artefacts and services for both human and
system purposes. VWE primarily uses metadata for three things: for annotations used to
register and locate resources, for describing properties for humans, and for describing
properties for use by the system. Even though RDF was mainly chosen for its machinereadable and machine-processable semantic abilities, it is RDF’s ability to represent very
simple, as well as very complex, data models that make the system flexible. The ability to
represent machine-processable data semantics enhances adaptability; more importantly, this
ability creates the foundation for adaptivity, because metadata can be used as input for
47
This does of course break the compliance with IMS CP.
91
automatically (or semi-automatically) modifying the VLE, based on input from users, the
system or the outcome of pedagogical activities.
Taken altogether, this creates an interesting potential in terms of connecting the VLE (or
rather the components of the composition that forms the VLE) and the digital learning
content, to the pedagogical process. This is discussed in more detail in the next chapter.
5.3 Representing the pedagogical process
The research results presented in papers 5 and 6 address research question 3 by suggesting an
approach for connecting components in the learning environment to the curriculum, which
represents a vital piece of the pedagogical instructions.
5.3.1 The Annofolio infrastructure
The research described in paper 5 resulted in a prototype (the Annofolio) that can be used to
annotate XML-documents with RDF metadata. As described in chapter 4.4, the Annofolio is
built on top of Semantic Web technology, and utilizes the Annotea W3C recommendation
(Koivunen, Swick, José, & Prud'Hommeaux, 2002) to “point in” RDF annotations using
XPointers. The suggested Annofolio methodology mainly accomplishes two important things:
first, it introduces graph-based markup methodology for annotation of XML resources; and
second, it separates the metadata markup (annotations) from the resource that is subjected to
markup. This approach is opposed to document centric XML-based approaches, like TEI (Ide
& Sperberg-McQueen, 1995). Besides adding to the general problems associated with XMLbased metadata (see, e.g., Nilsson, Palmér, & Brase, 2003; Nilsson, Naeve, Duval, Johnston,
& Massart, 2008), approaches such as TEI require write-access to the annotated resource.
This means that the access to the resource by definition becomes authoritarian, and that it is
not suitable for contributing to a metadata ecology, such as the one suggested by Nilsson,
Palmér, & Naeve (2002). Besides, even if everyone who wanted to annotate a resource could
be given write-access, the resource would pretty soon be a complete mess of XML-tags and it
would become unmanageable very quickly.
By handling semantic markup of XML resources as annotations that are separated from the
resource, they can be managed as an integrated part of the metadata ecology. The use of RDF
makes it possible to utilize the expressive and semantic capabilities of RDF as well as related
Semantic Web technologies, such as ontologies.
Similar conclusions have also been made by others (see, e.g., Handschuh, Staab, &
Ciravegna, 2002; Ver et al., 2003). Patriarche et al. (2005), who introduce a tool (NOESIS)
that is based on principles that are similar to the Annofolio. In that respect NOESIS verifies
the validity of the results presented in paper 5. There are some major differences though. The
most important difference is that the Annofolio is implemented solely using web technology
and is independent of any client – with the exception of the web browser, while NOESIS
depends on client software. The Annofolio implements an Ajax-like approach, using
ECMAScript (ECMA, 1999) for application logics on the client.
While NOESIS has a specific community focus, targeting health professionals and knowledge
management in the medical field, the Annofolio is a generic tool that has no bindings to any
specific application field or ontology. One of the objectives of the Annofolio is to support
annotations that use both formal ontologies and non-formal annotations. This aim, together
with the characteristic of being unbound by the resource that is annotated by storing
annotations separated from resources, as well as separated from the Annofolio system in an
external RDF compliant store, contributes to making the Annofolio suitable for contributing
to a global and decentralized metadata ecology similar to notion discussed by Nilsson,
92
Palmér, & Naeve (2002). This was also discussed as a part of my theoretical framework (see
chapter 3.9). A consequence of the methodology used by the Annofolio is that different
contributors may independently add annotations, according to different ontologies, for
different purposes, and in different contexts, to the metadata ecology without interfering with
each other. This is referred to as different metadata layers in the software architecture, which
is described in paper 5.
A differentiation between different types of metadata markup was also made in papers 5 and
6. The separation was made between internal/structural markup48, external markup49, and
internal markup50. A similar (but different) differentiation was made in the NOESIS project,
where a separation was made between annotation types for cataloguing, conceptual
annotations and textual annotations. While the Annofolio focuses on how and where metadata
are managed, NOESIS focuses on the type and purpose of the annotations. This is likely to be
a consequence of NOESIS focusing on a specific application area, while the objective of the
Annofolio is to be neutral by only providing the infrastructure for annotations of XML
resources.
XML-based approaches, like TEI, have as their field of application to specify encoding
methods for machine-readable texts, chiefly in the humanities, social sciences and linguistics
(Ide & Sperberg-McQueen, 1995), a field where most of the resources are texts that are
relatively stable and where the markup is done by a few trusted individuals in a controlled
environment, i.e., an authoritarian approach. However, such an approach does not scale well
in a global and distributed environment where markup and annotations are added
independently by a large number individuals, according to different ontologies, in different
contexts and where annotations may overlap each other in terms of scope within the annotated
resource. Besides being more or less impossible to manage using document centric
approaches, the result would be a veritable mess. In addition, the mere nature of an
infrastructure such as the Internet creates a need to go from machine-readable to machineprocessable – in order to be human-manageable in the larger perspective. This is also the very
essence of the Semantic Web as it was outlined by Berners-Lee, Hendler, & Lassila (2001)
and Berners-Lee & Fischetti (1999).
One very interesting possibility opened up by the Annofolio is the ability to connect
annotations of resources to social networks using the semantic capabilities of RDF. One
example is by the use of the Friend Of a Fried (FOAF) ontology (Brickley & Miller, 2007).
Such approaches have been described by Avram (2006), as well as by Macgregor &
McCulloch (2006) and could very well be utilized together with the Annofolio, even though
this has never been tested. The potential is there, and it is obviously interesting for TEL. In
the context of this thesis, it would be a possible approach to dynamically connect pedagogical
designs and pedagogical processes to components of the VLE and to digital learning
resources as well as to the social context of learning – whatever that may be in whatever
situation – using the semantics represented by RDF annotations and (potentially) associated
ontologies. Koper (2004) refers to similar ideas in terms of Self-Organized Learning
Networks. He argues that Semantic Web technology, together with representations of
pedagogical design, such as IMS LD will play a central role for their accomplishment. One
issue that is often raised, together with the reluctance to metadata tagging, is the lack of
48
XML syntax containing structural markup according to the XML Schema
Referring to markup addressing the resource as a whole, such as for cataloguing purposes or annotations
referring to the complete resource.
50
Referring to annotations, or RDF statements, addressing only a part of a resource with the purpose of replacing
the semantics expressed using XML markup in the TEI example.
49
93
integrated tools for the Semantic Web. Even though there is still a long way to go, the
Annofolio provides one piece of this puzzle.
Another problem that was discussed in paper 5 is the ability to manage resources that change.
When an infrastructure that allows annotation of resources that are out of control of the
annotator is put in place, there is no way of controlling whether resources change or not. This
obviously needs to be dealt with somehow. Some minor tests were carried out in order to test
the prototype’s ability to cope with changes made in annotated XML-files. The results
showed that changes were handled quite well, which are probably a result of the way
XPointers were implemented. The XPointers’ position is determined relative to the closest tag
in the XML DOM tree. This means that annotations become immune to changes of the data –
provided that this is done in the “right” section of the XML file, while annotations are most
likely to be affected by data change in some cases. This needs to be evaluated further and
possible problems need to be addressed in the future.
I will discuss how the Annofolio was put into practice in the following chapters. The research
presented in paper 6 suggests how semantic annotations can be utilized within TEL and how
it can be utilized to address research question 3 by suggesting a methodology for connecting
pedagogical instruction to digital learning content.
5.3.2 Annotating the pedagogical process
The research in paper 6 examines how the Annofolio methodology can used to represent
pedagogical instruction, in this case represented by the Swedish national steering documents;
of which the national curriculum and course plans are vital parts. The results that were
presented in paper 6 are based on a rather limited experiment in curriculum mapping and
more research needs to be carried out to verify the findings. In spite of that, the experiment
gives a couple of indications of the potential of the Annofolio, and above all an indication of
the potential of Semantic Web technology and the use of semantic annotations in TEL. The
research presented in paper 6 is strengthened by recently published research (see, e.g., Dietze,
Gugliotta, & Domingue, 2007; Koper, 2004; Psyche, Bordeau, Nkambou, & Mizoguchi,
2005; Devedzic, 2004; Richter, Nejdl, & Allert, 2002; Lama, Arroyo, Sicillia, & Lópes-Cobo,
2006), where the importance of the role of the Semantic Web technology for TEL is
emphasized.
The ontology that was used for the experiment51 in paper 6 was very simple and was
originally applied for the TEI encoding of the same documents. Most likely, this does not
influence the results presented in paper 6, and it may even be an advantage in creating a
controllable environment as the complexity is kept down and action and reaction can more
easily be determined when using a simple ontology. The basic principles of how annotations
of the curriculum can be used are still illustrated and details of the final ontology to be used
for such a purpose are outside the scope of this thesis.
The underlying idea of the experiment was to apply a set of basic annotations to course plans.
This basic set of annotation was comparable to a traditional authoritarian metadata approach,
using a formal ontology, which could then provide “hooks” to which other metadata could be
connected by semantic mapping of metadata from different sources.
In addition to these basic annotations, “anyone” can contribute to the metadata ecology in
order to create an information infrastructure with the same intended multi-purpose use as the
basic set of annotations – in this case for mapping the curriculum to digital learning resources
based on central concepts in the curriculum.
51
The ontology (that was really a taxonomy) was described in detail in paper 6.
94
By implementing a simple prototype that utilized the Annofolio annotation infrastructure and
the provided RDF annotations, it became possible for teachers and students to locate suitable
digital learning resources based on annotations in the curriculum. One effect of this was that it
also became possible to locate resources intended for a completely different subject. See
example in chapter 4.5. This effect is a result of the dynamic nature of machine-processable
semantics and its ability to help in (and even automate) locating and matching needs with
resources. In fact, it has a much greater potential in a wider perspective than was illustrated by
this rather rudimentary experiment, and this potential is, for example, discussed by Koper
(2004), and by Mohan & Brooks (2003), and even more explicitly by Lama, Arroyo, Sicillia,
& Lópes-Cobo (2006), where an ontology that connects learning objects to Semantic Web
Services is suggested based on the Web Service Modeling Ontology (WSMO) and the Web
Service Modeling Language (WSML) (Roman et al., 2005). This is the approach taken by the
LUISA project52, where competency gaps are matched with suitable learning objects with the
help of WSMO ontologies described using WSML. The research presented by Lama et al.
(2006) is especially relevant as it contributes to a formalized approach to connecting Semantic
Web Services to learning Objects – something that is still missing in the VWE Architecture
Framework, where a much simpler approach was used that does not allow the same level of
automatization. The main reason for this is that the notion of Semantic Web Services did not
occur until very late in the project and the focus was on other modularity issues, which were
discussed in previous chapters. In addition, the field of Semantic Web Services is still a young
and rather immature field and it is not yet possible to clearly determine in which direction
things are moving, but it seems pretty clear that the experiences made in the field will have a
great impact on and be beneficial for TEL. The potential is especially appealing in the light of
Semantic Web based representations of pedagogical design and pedagogical processes in
relation to modular VLEs and modular digital content in a global, distributed environment
where a manifold of pedagogical approaches need to coexist (see, e.g., Naeve & Sicilia, 2006;
Monceaux et al., 2007). I will return to this discussion in chapter 5.4 where I attempt to pull
things together.
One of the strengths of the methodology, used by the Annofolio, is that it allows a
combination of formal ontologies, such as the simple example given in paper 6, and less
formal social annotations that may evolve in the metadata ecology, much like the idea of
Folksonomies (Voss, 2007). Even though formal ontologies can be usable, they are only used
to a certain extent and there is a need to allow less formal approaches as a complement. One
example of how such approaches could be utilized is to allow teachers to annotate their use of
digital learning resources. Teachers could also be allowed to make annotations to the
curriculum, which could then be used by other teachers in other contexts. Teachers’
annotations could also be used as a way of connecting different parts of the pedagogical
design to various components of TEL.
The Swedish school system, which is extremely decentralized, has a continuum of levels in
the curriculum. The documents that represent the curriculum range from the national steering
documents, containing the national curriculum, national course plans etc., via local steering
documents at the municipal level (containing more detailed and locally adapted course plans
and other curriculum documents), to documents at the individual school level, containing
detailed course plans for implementation, and equivalents to instructional designs in the shape
of lesson plans and likewise. The lesson plan as a metaphor for instructional design was
discussed in chapter 3.4. While the latter documents, which represent the pedagogical designs
that are supposed to be implemented through teachers’ practice, may be represented using
52
http://www.luisa-project.eu
95
approaches such as IMS LD, all the different levels of the curriculum need to be semantically
interconnected through an infrastructure, such as the one made possible by the Annofolio.
Furthermore, there is a need to interconnect this infrastructure with a standard, efforts such as
IMS LD, in order to make it an integrated part of the system representation of pedagogical
design – or learning design, as Koper (2005) refers to it in the context of IMS LD. One
possible way to approach this issue would be to allow Semantic Web annotations of IMS LD.
This would work in a similar fashion as the annotation of course plans presented in paper 6.
Even though the preferable way forward would be to base IMS LD itself – or future
equivalents of IMS LD – on Semantic Web technology, which has also been suggested by
Koper (2004). This would help in solving many of the issues encountered in this thesis.
There are many factors that indicate that Semantic Web technology will have a significant
impact on TEL in the future. In the context of this thesis, Semantic Web technology can be
regarded as the (dynamic) glue that keeps everything together in a semantically
interconnected and dynamic web of components.
The Semantic Web approach to curriculum mapping, suggested in papers 5 and 6, is only the
beginning of the exploration of this field and a lot of research still needs to be done. Research
question 3 has been addressed, but only partially answered. The issue of representing
pedagogical designs and pedagogical processes opens up an abyss with endless potential
opportunities, and a huge number of occurring problems, some of which were discussed
above, related to other research in this area. However, the results presented in papers 5 and 6
provide indications of how to proceed and of future needs that need to be regarded from a
wider perspective. This wider perspective will be briefly discussed in the next chapter, where
I will also summarize my research results and discuss them in relation to each other.
5.4 Final discussion and conclusions
This thesis spans four aspects of TEL that are of interest in the light of my three research
questions. These aspects were outlined and illuminated in my theoretical framework as being
about: pedagogy, technology, content, and learning activities. Each one of these aspects
contains their respective components that were discussed in chapter 3 and to various extents
addressed by my research questions.
5.4.1 The context of my research contributions
The essence of the research presented in this thesis concerns the interdependence between
learning theories, the implementation and representation of pedagogical design (traditionally
instructional design), pedagogical process, the design and implementation of learning
technology and digital learning content. These are all interconnected in a complex network of
mutual dependencies, and changes in one place will most likely trigger changes and/or tradeoffs in other parts of this complex network.
The mapping between the desired pedagogical approach and the corresponding learning
activities is often weak and insufficient when using technology as a mediator. This mismatch
is often caused by trade-offs that are constantly made when designing and implementing
learning technology.
The distributed nature of the web and a modular learning infrastructure approach, together
with an increasing focus on the collaborative and social aspects of learning, emphasize the
need for a better mapping between learning technology and learning theories (see, e.g.,
Schluep, Bettino, & Guttormsen Schär, 2005; Olivier, Roberts, & Blinco, 2005; Counterman
et al., 2005). In addition, the introduction of social and collaborative software makes the
96
pedagogical process even more unpredictable, which in turn makes it harder to design and
represent pedagogical design and process.
So far, the most common strategy to deal with VLEs for Web-Based Learning has been to
implement general-purpose, web-based learning platforms, i.e., LMSes. One of the paradoxes
of LMSes is that the concept was first invented to be used together with learning objects, and
the paradox lies in the fact that LMSes provide a non-modular learning environment for
modular learning content, an issue that was discussed in chapter 5.2 and 3.7.1.
The idea of general-purpose LMSes appears to be appealing in many ways, but a problem is
that LMSes stand out as being too isolated and restricted in the wider perspective – it is like
putting the learners behind a fence so that they should not be tempted by what resides outside.
LMSes tend to be implemented in a silo-like fashion that prevents interaction with the outside
world and implementations are often based on old-fashioned conceptions of the requirements
of TEL. These old-fashioned conceptions emanate from previous practice and legacy of
concepts like Computer-Based Training (translated into Web-Based Training on the web), and
similar approaches that are often a non-fit to the requirements of a global world. Today,
education and learning are becoming a global affair with a strong focus on learner-centered
and personalized learning – very much the opposite of what is supported by many LMSes.
By necessity, the learning environment becomes much more complex when the step is taken
from the local, controllable environment of Computer-Based Training to the global,
distributed and much less controllable learning environment on the web. The learning
environment is extended from covering only what is provided by the teacher, in terms of
software and content, to encompassing whatever is available in terms of content and services
on the Internet.
Teachers can no longer solely decide what students should be given access to and there is
always the possibility of choosing alternative ways to knowledge through the use of
alternative content, services and through interaction with others. This situation complicates
the every day life of teachers, but at the same time it is an unequalled pedagogical opportunity
and a reality that needs to be dealt with in various ways. The situation does not only
complicate the life of teachers, it complicates the life of learners as well. Suddenly everything
is at the tip of their fingers – endless possibilities in terms of information, learning content,
social interaction, services and so on, and this situation increases the need for durable learning
strategies and durable learning tools that are loosely coupled. Ultimately it will be a challenge
for teachers to adapt their teaching strategies so that they conform to this new situation, and
likewise it will be a challenge for learners to adapt their learning strategies accordingly. This
is also a challenge for learning technology to enabling teachers and learners to keep it all
together in order to create manageable learning situations rather than a confusing chaos.
The effect of globalization is reinforced by the increasing focus on phenomena like
personalization, flexible learning, life-long learning, on-demand learning, distance learning,
and the disappearance of educational system borders (such as between stages and levels), and
other trends in TEL (see, e.g., Howell, Williams, & Lindsay, 2003). At the same time, the
pace of the technological development is so fast that there is no way that LMS vendors can
keep up and always be in the frontline, implementing all the new functionality that is needed
for all desirable learning scenarios. Instead, LMSes become like Swiss army knives – they are
quite good at doing a lot of things quite poorly. As a comparison: it is relatively unusual to
see professional craftsmen who use Swiss army knives. They use a modular system with
special tools for special tasks, and tools are combined in different ways in different contexts,
addressing different tasks. Without a modular learning infrastructure, based on standards for
interoperability, it will more or less be left to the LMS vendors to decide which tools will be
97
available, and by that they also decide which pedagogical approaches that will be supported.
With a modular learning infrastructure on the other hand, special-purpose software can be
composed in a loosely coupled fashion to form VLEs that are much more powerful and at the
same time flexible. This was illustrated by the research presented in papers 2, 3, and 4.
The situation described above is not only the reality for TEL, it is also the reality within many
other areas where ICT is used, and it is also the main reason for the general ongoing
development toward modular infrastructures, where new and better services (added value) are
created by combining services from different service providers – the web is turning into a
gigantic “building kit” that is sometimes described using atoms, Lego, puzzles or any other of
the many popular metaphors.
While this development creates a number of interesting new possibilities, it also introduces a
number of new issues and problems that need to be solved. One of the most urgent problems
concerns how to keep everything together in the highly distributed and loosely connected
infrastructure that becomes the result. There is a need for sophisticated mechanisms to “keep
things together” in a way that makes the VLE appear as a coherent whole in an environment
where the “whole” is actually more than the sum of its parts – it is the sum of its part, plus
their mutual relations – and where different services, and data from different sources can be
seamlessly stitched together, using modular approaches and Semantic Web technology. This
is also the context within which this thesis contributes to TEL as a research area.
5.4.2 An eclectic summary
The theoretical framework (see chapter 3), together with a weighted analysis of the results
presented in the appended papers, can be used to pinpoint a chain of interrelationships that
lead to a couple of vital conclusions about the research results presented in this thesis, but also
about things that need to be further explored and discussed.
First of all, the part of my theoretical framework that originates from the pedagogical field
clearly shows that there are different theories about how we learn, and that these theories are
applicable to different pedagogical contexts, often motivated by different learning objectives
and types of skills or knowledge to be acquired. In addition, the pedagogical methods chosen
by teachers reflect their views on learning, and by that (consciously or not) the learning
theories they believe in. The pedagogical methods favored by teachers affect the way in
which pedagogical instructions and processes are expressed. This has consequences for how
pedagogical design (instructional design) and process need to be applied. In chapter 3.4, I
discussed how different learning theories put different demands on the practice of
instructional design. Consequently, this affects the ways in which instructional design must be
represented in VLEs. This is directly related to the view of the learning process in each
theory. System representations that only support sequences of activities, such as those used
with a chunking and sequencing approach, may be satisfying in a pedagogical situation where
a behavioristic approach is used. Behaviorists regard the pedagogical process as a sequence of
predictable activities, leading to a certain outcome in terms of changes in learners’ behavior –
changes that can then be tested in steps in the sequence. Such processes are rather
straightforward to represent using a hierarchal structure, suitable for representation using
XML.
The development within TEL has led to a change in technology focus, in terms software and
service, as well as to a change in pedagogical focus. With technologies that altogether better
support collaboration, social interaction, and a sophisticated organization of learning content,
it also becomes natural to emphasize pedagogical approaches that have the potential to make
better use of those developments. Social software and collaborative learning environments
98
create the means for pedagogical approaches and learning that better support the type of
learning that requires a higher level of cognitive processing targeting learning objectives and
contexts where many teachers favor constructivist approaches. This is also likely to be one of
the main reasons that many of the discussions about TEL during the last ten years have
addressed issues on how to better match TEL to the requirements of constructivism. This is
especially apparent regarding matters that concern the relations between instructional design,
constructivism and TEL (see, e.g., Mayer, 1999; Schank, Berman, & Macpherson, 1999;
Jonassen, 1999). Those issues were also one of the major driving forces behind the emergence
of EMLs, such as the IMS LD, as well as the ongoing change of focus from instructional
design to learning design – concepts that I refer to by using the general and collative term:
pedagogical design.
The necessity of the above discussion is clearly manifested by the fact that the situation
becomes much more complicated in pedagogical scenarios where constructivist approaches
are used. While the learning process can quite easily be predicted in a behaviorist scenario, it
becomes virtually unpredictable in a constructivist scenario. The unpredictability expresses
itself in different ways, and depending on the choice of methodological approach, it
influences different parts of the learning environment and the learning situation in different
ways. It may, for example, be hard to predict exactly what content students will need or use, it
may be hard to predict what tools and what kind of VLE students will need, and it will
certainly be hard to predict exactly what learning path(s) students will take to acquire their
desired knowledge. These are all things that were kept together by the instructional design
and its system representation in the behaviorist scenarios. However, that requires education
and learning to be predictable and curriculum-centric to an extent that it is not possible in a
typical constructivist scenario. Hence, the pedagogical design and processes need to be
represented in a much more dynamic and self-organizable way.
Therefore, the VLE must be adaptable in order to enable the utilization of such dynamic and
self-organizable representations of the pedagogical design and process. Experiences from
computer science, as well as from TEL show, simontaneously, that modularity promotes
flexibility as modules in a modular system can be composed (and decomposed) to form
compositions that have different properties and qualities for use in different contexts. This
was discussed in chapters 3.6 and 3.11 of my theoretical framework. One conclusion that was
drawn was that an important step toward the needed flexibility could be made by the
utilization of modular approaches for TEL. This is evident from the success of objectorientation and modularity within computer science and software engineering. In addition,
there is already a widespread and widely accepted model for modular digital learning content,
i.e., learning objects. A feasible approach would therefore be to extend the learning object
model to also embrace the VLE. As a matter of course, this hypothesis lead to the formulation
of research question 1 and 3.
The findings presented in paper 1 showed that there were some major technical issues that
needed to be addressed, which constituted that basis for the research in paper 2, where the
VWE Learning Object Taxonomy was formulated. This taxonomy demonstrated that it was
theoretically possible to place digital learning content and the VLE in a common conceptual
model. The research presented in papers 3 and 4 clearly demonstrated that the taxonomy
could be translated into a modular and distributed software architecture.
Together those research results provide the basis for a very flexible and modular
infrastructure that embraces both the VLE and digital learning content. This infrastructure
was implemented by the VWE prototypes. The flexibility lies foremost in a very high level of
adaptability. Both the VLE and the digital learning content is fully modular and can be
composed and configured to suit most requirements – provided that the needed tools exist, or
99
can be developed. The underlying technology and infrastructure provide the conditions
needed to participate in a global infrastructure, utilizing components and services provided by
others. Similar approaches are already being used related to the Web 2.0 phenomenon, where
lightweight technologies such as RSS and XML-RPC (and sometimes proprietary APIs) are
used to accomplish mash-ups, which can be considered to be equivalents of simple composite
applications.
The possibilities created by this kind of composite VLEs will most likely change the way in
which VLEs are regarded. Currently the VLE is often viewed as something that exists in
terms of a system with well-defined system borders, often in terms of an LMS product. The
antithesis to the LMS occurs when a modular approach is applied to its extreme. In such
cases, the VLE does not actually exist until it is created, and it will always be created for a
specific purpose, aiming to address the requirements of a specific learning situation,
associated with specific pedagogical and functional needs. The only thing that is pre-existing
is the modular infrastructure for composing VLEs and content – the learning infrastructure.
However, even though a highly adaptable learning infrastructure creates the possibility to
compose a VLE “on demand”; a VLE that can be continuously adapted and changed, there is
still a major issue that needs to be addressed and that is adaptivity. Even though the VLE is
adaptable and can be adapted at run-time, this still assumes a certain amount of predictability,
in terms of needed tools, learning content and structure.
There is an important difference between the notion of adaptability and adaptivity. An
adaptable VLE can be adapted by actors in the system, and in this case it means that
functionality and content can be added and removed from the VLE at any time from
composition-time and forward. This involves conscious actions by a user, or something that
has been pre-determined by a pedagogical designer. Adaptivity on the other hand occurs when
the VLE and the content has the ability to change automatically based on changes or feedback
from users in the system, triggered by processes and activities – such as changes in the
pedagogical process, that in turn may be triggered by the choice of a user, user profiles (such
as previous experience) or other kinds of feedback.
There is a need to connect to other components within the learning infrastructure, based on
the semantics expressed and matched to different components in the learning infrastructure.
Such connections were illustrated by the experiments presented in papers 5 and 6, where
Semantic Web technology was suggested as the approach to accomplish such dynamics. In
order to make the VLE and the digital learning content adaptive, the pedagogical design must
be represented in a way that facilitates adaptation to changes in the process based on feedback
from users, from the VLE, from content, or even from the pedagogical process itself. One
approach to this is the notion of attention metadata (see, e.g., Najjar, Wolpers, & Duval,
2006). Attention metadata can be used to represent the relationship between users and other
resources within TEL. Attention metadata are harvested from different sources and is
compiled into extensive data about users actions, activities, preferences etc., which means that
it is suitable as input to the pedagogical process.
Even though the experiments that were presented in papers 5 and 6 were quite small, and
restricted to the context of adding annotations to the curriculum, they indicate that Semantic
Web technology and annotations in a metadata ecology can be taken much further. There is
some recent research that supports this thesis, such as, by Nilsson et al. (2002), who
emphasize the importance of a metadata ecology for learning, by Koper (2004), who
discusses the use of semantic web technology for representing IMS LD, and by Psyche,
Bordeau, Nkambou, & Mizoguchi (2005), who discuss an ontology approach to connect
100
learning theories to learning design (see also, e.g., Nilsson et al. 2008; Nilsson, Johnston,
Naeve, & Powell, 2007; Powell, Nilsson, Naeve, Johnston, & Baker, 2004; DCMI, 2007).
By adding the ability to represent pedagogical process and pedagogical instruction that utilize
machine-processable data-semantics, it also becomes possible for different components in the
learning infrastructure to interpret the semantics and act upon it. By using the same approach
for other parts of the learning infrastructure, such as digital learning content, Semantic Web
Services, Semantic Web annotations etc, it becomes possible to connect and integrate data
from different sources – using technologies such as machine reasoning – in order to respond
to actions and trigger changes that can help in creating reasonably self-organizable learning
environments.
The above discussion indicates that while modularity creates the pre-conditions for
adaptability, Semantic Web technology creates the accompanying pre-conditions for
adaptivity. It should be emphasized that the idea is not that the system should create new
pedagogical design, but that the system, based on the input-types described above, can match
learning situations to reusable learning designs, for example, based on patterns for reoccurring learning scenarios.
5.5 A complex path for the future
A re-occurring issue, related to the research presented in this thesis, is complexity.
Complexity encompasses complex learning and complex cognitive processing. Complexity in
terms of learning technology and infrastructure is also included, in combination with
globalization and a constantly increasing focus on interdisciplinary approaches, as well as a
changing view of learning and knowledge. Finally, there is complexity in terms of complex
relations between different components of TEL. This was emphasized in my theoretical
framework and illustrated by Figure 3-13.
Several concepts and ideas that have been discussed in this thesis, such as modularity loose
coupling, self-organizability, adaptivity, dynamics, are concepts explored within Complexity
Theory:
“Complexity science focuses on the dynamic relationships and patterns among
phenomena rather than the static properties of isolated objects. Its explanatory framework
and core concepts—which include emergence, evolution and embodiment—have been
adopted by mathematicians, biologists, physical scientists and social scientists, as well as
by many thinkers in the humanities.” 53
Consequently, understanding complexity seems to hold the key to understanding the
embodiment of TEL, and to understand how to utilize modular learning infrastructures to
achieve adaptable as well as adaptive systems for the future. My conclusions in this thesis
include the views that modularity creates adaptable systems, and that the use of Semantic
Web technology may help in creating the flexibility needed for adaptivity – a point of view
that is also verified in other studies (discussed in chapter 3.9 and 5.4.2).
However, the interrelations between the different components (technical, social, pedagogical
and so on) are still unclear in many respects and a lot more research needs to be done in order
to take the next steps. In order to realize the learning environments of the future and to
develop better and more suitable learning technology, there is clearly a need to better
understand the interrelations of the different components of TEL, and the emergence of the
huge and complex system that learning and education become a part of. It will not be possible
53
Cited from the Complexity and Education web site at
http://www.complexityandeducation.ualberta.ca/glossary/g_complexsci.htm
101
to continue to uphold the distinction between pedagogy and technology that is made far too
often. Pedagogy and technology are both parts of the same complex system, which has been
illustrated in this thesis. My hypothesis is that all aspects of learning must be regarded and
analyzed as one modular and nested system, where everything is linked to everything else,
and where all parts may influence all of the other parts – either explicitly or implicitly.
A modular view of TEL, such as the view put forward in this thesis, does not mean that an
approach based on reductionism would be appropriate. Such a view would be far too
unsophisticated, and would most likely disregard some of the most important aspects of TEL.
Considering this point of view, Complexity Theory is suggested as a suitable point of
departure. Even though it is a fairly new scientific approach, it has already been applied to
several research areas, such as social science, biology, education (see, e.g., Brent & Simmt,
2003), and Knowledge Management, (see, e.g., Dann & Barclay, 2006), and it stands out as
being suitable for multidisciplinary and eclectic research approaches – whose importance was
emphasized in the introduction of this thesis.
Complexity Theory may be used as an approach to understand and explain many of the issues
at a macro-level, but there is still a need to understand relations and issues that may occur at
the micro-level. While the macro-level represents the complex system that is TEL, with all its
dependencies, the micro-level represents the environment and the components that are used in
a specific context, such as a learning situation (and related activities), that is interrelated to the
complex system at the macro-level. The relations and interactions at the micro-level are less
complex in that they are more clearly demarcated, and by that more manageable, which
makes it possible to regard them as an activity system. The micro-level is also more
predictable to some extent. However, in the same sense that the VLE does not exist in a
modular environment until someone creates it, the context of the micro-level does not exist
until it is created. This is a consequence of regarding the micro-level in the context of being a
part of a complex system. This rather complicated reasoning indicates that Complexity
Theory alone cannot solve all problems, and that other theories may be more suitable in some
cases, and especially at the micro-level. One such theory is Activity Theory (Engeström,
1999), which is (like constructivism) rooted in Vygotsky’s sociocultural research.
Activities are modeled within the framework of activity systems. To better understand the
actions and activities going on within an activity system, activity systems are commonly
modeled as a triad, whereas an activity is undertaken by a human (being the subject). This
activity is started for a purpose (that is the object), for example, to solve a problem, and the
activities carried out are usually mediated by some kind of tool or similar (that is an artifact).
Mediation can also be accomplished by other kinds of artifacts that are not explicitly regarded
as tools in all contexts, such as culturally established artifacts (such as language). An activity
system usually exists in a social context, constructed within the framework of a community
(Engeström, 1999).
Combining Complexity Theory and Activity Theory may provide a theoretical framework,
and the tools to facilitate the understanding of the properties needed in order to represent
pedagogical processes and design the VLEs of the future. Besides supplementing each other,
it might also be possible to use the two theoretical frameworks to verify the relevance of
research results by analyzing results using a combination of both frameworks.
102
References
Abdallah, R., El Hajj, A., Benzekri, A., & Moukarzel, I. (2002, 18-22 August). On the
Improvement of course interoperability in E-Learning Models. Paper presented at the
International Conference on Engineering in Education (ICEE 200).
Abernethy, K., Treu, K., Piegari, G., & Reichgelt, H. (2005). A Learning Object Repository in
Support of Introductory IT Courses. Paper presented at the Proceedings of the 6th conference
on Information technology education, Newark, NJ, USA.
ADL. (2001, June 22). The SCORM Overview. SCORM Sharable Content Object Reference
Model Retrieved 11 September, 2003, from http://www.adlnet.org
ADL. (2004). SCORM 2004 2nd Edition, Overview. Alexandria, VA: ADL.
Anderson, L. W., & Krathwohl, D. R. (2001). A taxonomy for learning, teaching and
assessing: A revision of Bloom's Taxonomy of educational objectives (Complete edition).
New York: Longman.
Anderson, L. W., & Sosniak, L. A. (1994). Bloom's taxonomy: a forty-year retrospective.
Ninety-third yearbook of the National Society for the Study of Education. Chicago: The
National Society For The Study Of Ed.
Anderson, N. (2006). Tim Berners-Lee on Web 2.0: "nobody even knows what it means".
Retrieved 30 September, 2007, from http://arstechnica.com/news.ars/post/200609017650.html
Anderson, P. (2007). What is Web 2.0? Ideas, technologies and implications for education
(No. TSW0701): JISC.
Andersson, C. (2000). Kunskapssyn och lärande - i samhälle och arbetsliv (1 ed.). Lund,
Sweden: Studentlitteratur.
Antoniou, G., & Van Harmelen, F. (2004). A semantic web primer. In Cooperative
information systems (pp. 61-104). Cambridge, Mass.: MIT Press.
Arsanjani, A., & Holley, K. (2006, September 18-22). The Service Integration Maturity
Model: Achieving Flexibility in the Transformation to SOA. Paper presented at the IEEE
International Conference on Services Computing (SCC'06), Chicago, USA.
Avram, G. (2006). At the Crossroads of Knowledge Management and Social Software
Electronic Journal of Knowledge Management 4(1).
Baldwin, C., & Clark, K. (2000). Design Rules: The Power of Modularity (Vol. 1).
Cambridge, MA: The MIT Press.
Bandura, A. (1977). Social Learning Theory: Englewood Cliffs, N.J. : Prentice Hall.
Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory.
Englewood Cliffs, NJ: Prentice Hall.
Bandura, A., & Walters, R. H. (1963). Social Learning and Personality Development. New
York Holt, Rinehart and Winston, cop.
Bannan-Ritland, B., Dabbagh, N., & Murphy, K. (2002). Learning Object Systems as
Constructivist Learning Environments: Related Assumptions, Theories, and Applications. In
D. A. Wiley (Ed.), The Instructional use of Learning Objects (1 ed., pp. 61-97). Bloominton,
IA: Agency for Instructional Technology, and Association for Educational Communications
& Technology.
103
Battle, S., Bernstein, A., Boley, H., Grosof, B., Gruningen, M., Hull, R., et al. (2005).
Semantic Web Services Framework (SWSF) Overview: W3C.
Beckett, D. (2004, 10 February 2004). RDF/XML Syntax Specification (Revised). Retrieved
17 October, 2007
Berners-Lee, T. (1998, 1998-10-14). Why RDF model is different from the XML model.
Retrieved January 15, 2004, from http://www.w3c.org/DesignIssues/RDF-XML
Berners-Lee, T., & Fischetti, M. (1999). Weaving the Web: The original design and ultimate
destiny of the World Wide Web, by its inventor (1 ed.): Collins.
Berners-Lee, T., Hendler, J., & Lassila, O. (2001). The Semantic Web - A new form of Web
content that is meaningful to computers will unleash a revolution of new possibilities
[Electronic Version]. Scientific American. Retrieved 10 october 2006 from
http://www.sciam.com/article.cfm?id=the-semantic-web.
Bitzer, D., Lyman, E., & Easley, J. (1965). The Uses of PLATO: a computer controlled
teaching system (No. R-268). Illinois: Coordinated Science Laboratory, University of Illinois.
Bjerg, J. (2000). Pedagogik, en grundbok. Stockholm: Liber AB.
Bloom, B., & Krathwohl, D. R. (1956). Taxonomy of Educational Objectives: The
Classification of Educational Goals, by a committee of college and university examiners.
Handbook I: Cognitive Domain. New York, NY: Longmans.
Boutell, T. (1996). CGI Programming in C and Perl. New York: Addison Wesley.
Bratina, T. A., Hayes, D., & Blumsack, S. L. (2002). Preparing Teachers To Use Learning
Objects. Technology Source (Nov-Dec 2002).
Bray, T., Paoli, J., & Sperberg-McQueen, C. M. (1998). Extensible Markup Language (XML)
1.0 [Electronic Version], 37. Retrieved 10 June 2007 from http://www.w3.org/TR/1998/RECxml-19980210.pdf.
Brent, D., & Simmt, E. (2003). Understanding learning systems: Mathematics education and
complexity science. Journal of Research in Mathematics Education, 34(2), 137-167.
Brickley, D., & Brickley, R. V. (2004, 10 February 2004). RDF Vocabulary Description
Language 1.0: RDF Schema. Retrieved 10 October 2007, from http://www.w3.org/TR/rdfschema/
Brickley, D., & Miller, L. (2007, 2 November 2007). FOAF Vocabulary Specification 0.91.
0.91. Retrieved 18 November, 2007, from http://xmlns.com/foaf/spec/
Bruner, J. (1961). The act of discovery. Harvard Educational Review (31), 21-32.
Bush, V. (1945/1994). As We May Think [Electronic Version]. The Atlantic Monthly, July
1945. Retrieved 2007-03-01 from http://www.press.umich.edu/jep/works/vbush/vbush.txt.
Capone, W., Christensen, C., Falls, J. M., Fletcher, D., & al., E. (2004). SCORM: Sharable
Content Object Reference Model - SCORM Sequencing and Navigation, Version 1.3.1.
Retrieved 2006-06-01, from http://www.adlnet.org/.
Cardoso, J., & Sheth, A. P. (2006a). The Semantic Web and its applications. In J. Cardoso &
A. P. Sheth (Eds.), Semantic Web and Beyond: Computing for Human Experience Semantic
Web Services, Processes and Applications (pp. 3-33). New York: Springer Science+Business
Media, LLC.
104
Cardoso, J., & Sheth, A. P. (2006b). The Semantic Web and its applications. In J. Cardoso &
A. P. Sheth (Eds.), Semantic Web and Beyond: Computing for Human Experience
Semantic Web Services, Processes and Applications (pp. 3-33). New York: Springer
Science+Business Media, LLC.
CEdMA. (1995). CeDMA Sigs/Task Forces.
http://www.cedma.net/sigs_task_forces.php
Retrieved March 7, 2005, from
Chou, C. (1999). Developing hypertext-based learning courseware for computer networks: the
macro and micro stages. Education, IEEE Transactions on, 42(1), 39-44.
Christiansen, J.-A., & Anderson, T. (2004). Feasibility of Course Development Based on
Learning Objects: Research Analysis of Three Case Studies. International Journal of
Instructional Technology & Distance Learning, 1(3).
Clark, J. (1997, 15 December 1997). Comparison of SGML and XML. Retrieved 20
February, 2008, from http://www-mit.w3.org/TR/NOTE-sgml-xml-971215
Clark, J., & DeRose, S. (1999). XML Path Language (XPath) Version 1.0, W3C
Recommendation 16 November 1999: W3C.
Clark, R. (2006). Building Expertise: Cognitive methods for Training and Performance
Improvement (2nd ed.). UK: Jossey Bass Wiley.
Collins, A. (1977). Processes in acquiring knowledge. In R. C. Anderson, R. J. Spiro & W. E.
Montague (Eds.), Schooling and the acquisition of knowledge (pp. 339-363). Hillsdale, NJ:
Erlbaum.
Counterman, C., Golden, G., Gollub, R., Norton, M., Severance, C., & Speelmon, L. (2005,
March 5, 2005). Sakai Java Framework. Version 1.0 RC5. Retrieved June 11, 2006, from
http://www.columbia.edu/acis/courseworks/sakai/externaldocs/Sakai-Java-Framework-05.pdf
Crowder, N. A. (1960). Automatic Tutoring by Intristic Programming. In A. A. Lumsdaine &
R. Glaser (Eds.), Teaching Machines and Programmed Learning: A source book (1 ed., Vol.
1, pp. 286-298): National Education Association.
Crowder, N. A. (1964). On the difference between linear and intrinsic programming. Phi
Delta Kappan, (44), 250-254.
Dann, Z., & Barclay, I. (2006). Complexity Theory and Knowledge Management Application.
The Electronic Journal of Knowledge Management, 4(1), 11-20.
DCMI. (2004, 2004-12-20). Dublin Core Metadata Element Set, Version 1.1: Reference
Description. Retrieved 10 February, 2008, from http://dublincore.org/documents/dces/
DCMI. (2007, 31 October 2007). Joint DCMI/IEEE LTSC Taskforce. Retrieved 10 February,
2008, from http://dublincore.org/educationwiki/DCMIIEEELTSCTaskforce
Decker, S., Melnik, S., van Harmelen, F., Fensel, D., Klein, M., Broekstra, J., et al. (2000).
The Semantic Web: the roles of XML and RDF. Internet Computing, IEEE, 4(5), 63-73.
Dempsey, J. V., & Van Eck, R. N. (2007). Distributed Learning and the Field of Instructional
Design. In R. A. Reiser & J. V. Dempsey (Eds.), Trends and Issues in Instructional Design
and Technology (2 ed., pp. 288-300). Upper Saddle River, NJ: Pearson Education.
Devedzic, V. (2004). Education and the Semantic Web. International Journal of Artificial
Intelligence in Education, (14), 39-65.
105
Dewey, J. (2004). Individ, skola och samhälle : utbildningsfilosofiska texter. Stockholm:
Natur och Kultur.
Dietze, S., Gugliotta, A., & Domingue, J. (2007). Context-Adaptive Learning Designs by
Using Semantic Web Services. Journal of Interactive Media in Education (Adaptation and
IMS Learning Design. Special Issue), (September 2007).
Dodani, M. (2002). The Dark Side of Object Learning: Learning Objects. Journal of Object
Technology, 1(5), 37-42.
Dodig-Crnkovic, G., & Crnkovic, I. (2003, January 2003). Computing Curricula: Teaching
Theory of Science to Computer Science Students. Paper presented at the Hawaii International
Conference on Education, Hawaii.
Douglas, I. (2001, 10-13 October). Instructional Design based on reusable learning objects:
Applying lessons of Object-Oriented software engineering to learning systems design. Paper
presented at the ASEE/IEEE Frontiers in Education Conference, 31st annual, Reno, NV,
USA.
Downes, S. (2002, 11-13 November). Design and Reusability of Learning Objects in an
Academic Context: A New Economy of Education. Paper presented at the eLearning: una sfida
per l’universita, Milan, Italy.
Dusek, V. (2006). The Philosophy of Technology - An Introduction (1 ed.): Blackwell
Publishing ltd.
Duval, E. (2004). Learning Technology Standardization: Making Sense of it All. Computer
Science and Information Systems, 1(1), 33-43.
ECMA. (1999). Standard ECMA-262 ECMAScript Language Specification (Standard
specification): ECMA.
Edwards, D., Rai, S., Philips, R., & Fung, L. C. C. (2007). A Framework for Interoperable
Learning Objects for E-Learning. In A. Koohang & K. Harman (Eds.), Learning Objects and
Instructional Design (pp. 437-469). Santa Rosa, CA: Informing Science Press.
Egidius, H. (2003). Lärande för 2000-talet (4 ed.). Stockholm: Natur och kultur.
Ekengren, D. (1998). Metadata on the Web using RDF. Kista: SITI.
Engeström, Y. (1999). Activity theory and individual and social transformation. In Y.
Engeström, R. Miettinen & R.-L. Punamäki (Eds.), Perspectives on Activity Theory (pp. 1938). Cambridge: Cambridge University Press.
English, H. B. (1942). How Psychology can Facilitate Military Training: A Concrete
Example. Journal of Applied Psychology, 3-7.
Erl, T. (2007). SOA Principles of Service Design (1 ed.). Boston, MA: Prentice Hall.
Ertmer, P. A., & Newby, T. J. (1993). Behaviorism, Cognitivism, Constructivism: Comparing
Critical Features from a Design Perspective. Perfomance Improvement Quarterly, 6(4), 50-72.
Fensel, D., & Bussler, C. (2002). The Web Service Modeling Framework WSMF. Electronic
Commerce: Research and Applications, 1(2), 118-130.
Fensel, D., Van Harmelen, F., & Davies, J. (2003). Towards the semantic web ontologydriven knowledge management. Chichester, England ; Hoboken, NJ: J. Wiley.
Fielding, R. T. (2000). Architectural Styles and the Design of Network-based Software
Architectures. University of California, Irvine.
106
Fisher, D. (2007, 2007). Instructional Design -- The Taxonomy Table. Retrieved February 10,
2008, from http://oregonstate.edu/instruct/coursedev/models/id/taxonomy/
Fisler, J., & Bleisch, S. (2006, April 11-13). ELML, The e-lesson markup language:
Developing Sustainable e-Learning Content Using an Open Source XML Framework. Paper
presented at the WEBIST 2006 - International Conference on Web Information Systems and
Technologies, Setubal, Portugal.
Fletcher-Flinn, C. M., & Gravatt, B. (1995). The Efficacy of Computer Assisted Instruction
(CAI): A Meta-Analysis. Journal of Educational Computing Research, 12(3), 219-241.
Forehand, M. (2005). Bloom's Taxonomy: Original and Revised. Emerging perspectives on
learning, teaching, and technology Retrieved September 12, 2006, from
http://www.coe.uga.edu/epltt/bloom.htm
Friesen, N. (2004a, 30 august - 1 September). Learning objects and standards: pedagogical
neutrality and engagement. Paper presented at the IEEE International Conference on
Advanced Learning Technologies, Joensuu, Finland.
Friesen, N. (2004b). Three Objections to Learning Objects and E-learning Standards. In R.
McGreal (Ed.), Online Education Using Learning Objects (pp. 59-70). London: Routledge.
Gagne, R. M. (1962). Psychological Principles in System Development. New York: Holt,
Rinehart and Winston, cop.
Gagné, R. M. (1965). Teaching Machines and Programmed Learning, III. Washington DC:
National Education Association.
Gagné, R. M. (1977). The conditions of learning (3. ed.). New York: Holt Rinehart &
Winston.
Garrett, J. J. (2005). Ajax: A New Approach to Web Applications [Electronic Version].
Javalobby. Retrieved 20 February 2008 from http://www.javalobby.org/articles/ajax/.
Garshol, L. M. (2003). Living with topic maps and RDF. Retrieved 18 October, 2007, from
http://www.web-services.gov/tmrdf.pdf
Gibbons, A. S., & Nelson, J. (2002). The Nature and Origin of Instructional Objects. In D. A.
Wiley (Ed.), The Instructional Use of Learning Objects (1 ed., pp. 25-59). Bloomington IA:
Association for Educational Communications & technology, and Agency for Instructional
Technology.
Goldfarb, C. F. (1991). Standards-HyTime: a standard for structured hypermedia interchange.
IEEE Computer, 24(8), 81-84.
Goldfarb, C. F. (1996). The Roots of SGML -- A Personal Recollection. Retrieved March 26,
2007, from http://www.sgmlsource.com/history/roots.htm
Greeno, J. G. (1989). A perspective on thinking. American Psychologist, 44(2), 134-141.
Grimnes, G. A. G., Edwards, P., & Preece, A. (2004). Learning Meta-descriptions of the
FOAF Network. In The Semantic Web – ISWC 2004 (Vol. 3298/2004, pp. 152-165).
Heidelberg: Springer Berlin.
Gruber, T. R. (1993). A Translation Approach to Portable Ontology Specifications.
Knowledge Acquisition, 5(2), 199-220.
Hall, W. (1994). Ending the tyranny of the button. IEEE Multimedia, 1(1), 60-68.
107
Handschuh, S., Staab, S., & Ciravegna, F. (2002, 1-4 October 2002). S-Cream, Semiautomatic creation of metadata. Paper presented at the European Conference on Knowledge
Acquisition and Management, Sigüenza, Spain.
Havice, B., & Galloway, C. A. (1998). Integrating HyperStudio Multimedia with Elementary
School Curriculum: A Third Grade Multimedia Activity. Technology and Children, 3(2), 3-7.
Haviland, V. S., & McCall, M. J. (1999). Transformation through Technology: How
HyperStudio Updated Middle School Research. The English Journal, 89(1), 63-68.
Hayes, P. (2002). RDF Model
http://www.w3c.org/TR/rdf-mt/
Theory.
Retrieved
February
14,
2003,
from
Haynes, D. (2004). Metadata for information management and retrieval. London: Facet.
Hazael-Mas, D. (2003). RDF in Real Life - Some examples of RDF applications. Retrieved
17 november, 2007, from http://www.w3.org/2003/Talks/semtour-athens-rdfapp/
He, H. (2003, 2003-09-30). What is Service-Oriented Architecture? Retrieved November 01,
2005, from http://webservices.xml.com/lpt/a/ws/2003/09/30/soa.html
Henning, M. (2006). The rise and fall of CORBA. Component technologies, 4(5), 28-34.
Hodgins, H. (2000, April 05, 2002). Everything you ever wanted to know about learning
standards but were afraid to ask. Learnativity. Retrieved 7 August, 2005, from
http://www.learnativity.com/standards.html
Hodgins, W. (2001, December 17, 2004). LALO. What's that? Some history behind current
Learning Standards Initiatives and some updates! Retrieved January 4, 2005, from
http://www.learnativity.com/lalo.html
Homme, L. E. (1960). The Rationale of Teaching by Skinner's Machines. In A. A. Lumsdaine
& R. Glaser (Eds.), Teaching Machines and Programmed Learning: a source book (Vol. 1,
pp. 133-136). Washington, D. C.: National Education Association of the United States.
Horrocks, I., Patel-Schneider, P. F., & van Harmelen, F. (2003). From SHIQ and RDF to
OWL: the making of a Web Ontology Language. Journal of Web Semantics: Science,
Services and Agents on the World Wide Web, 1(1), 7-26.
Howell, S. L., Williams, P. B., & Lindsay, N. K. (2003). Thirty-two Trends Affecting
Distance Education: An Informed Foundation for Strategic Planning. Online Journal of
Distance Learning Administration, 6(3).
Hylén, J. (2002). E-lärande som utmaning. Stockholm: Utbildningsdepartementet.
Ide, N. M., & Sperberg-McQueen, C. M. (1995). The Text Encoding Initiative: Its History,
Goals, and Future Development. Computers and the humanities, 29(1), 5-18.
IEEE. (2002, June 12). Final 1484.12.1-2002 LOM Draft
http://ltsc.ieee.org/wg12/files/LOM_1484_12_1_v1_Final_Draft.pdf
Standard.
from
IMS. (2001, 28 September). IMS Learning Resource Meta-Data Best Practice and
Implementation Guide. Version 1.2.1 Final Specification. Retrieved November 11, 2007,
from http://www.imsproject.org/
IMS. (2003a, 20 January 2003). IMS Learning Design Best Practice and Implementation
Guide: Version 1.0 Final Specification. Retrieved February 10, 2008, from
http://www.imsglobal.org/learningdesign/index.html
108
IMS. (2003, 03 March 2003). IMS Simple Sequencing Best Practice and Implementation
Guide. Version 1.0 Final Specification 1.0. Retrieved November 10, 2007, from
http://www.imsglobal.org/simplesequencing/ssv1p0/imsss_bestv1p0.html
IMS. (2004, 04 October 2004). IMS Content Packaging Best Practice and Implementation
Guide. Version 1.1.4 Final Specification. Retrieved
November 11, 2007, from
http://www.imsglobal.org/content/packaging/cpv1p1p4/imscp_bestv1p1p4.html
IMS. (2005a, July 21). IMS General Web Services Base Profile. Version 1.0 Public Draft
Specification. Retrieved August 28, 2005, from http://www.imsglobal.org/gws/index.html
IMS. (2005b, Tue, 11 Oct 2005). IMS Learner Information Package Specification. Retrieved
October 12, 2005, from http://www.imsglobal.org/profiles/index.html
Ismail, J. (2002). The design of an e-learning system: Beyond the hype. The Internet and
Higher Education, 4(3-4), 329-336.
Iyengar, A., MacNair, E., & Nguyen, T. (1997, 3-8 November). An analysis of Web server
performance. Paper presented at the Global Telecommunications Conference (GLOBECOM
apos), Phoenix, AZ, USA.
Jank, W., & Meyer, H. (1997). Didaktikens centrala frågor. In M. Uljens (Ed.), Didaktik (pp.
47-74). Lund: Studentlitteratur.
Januszewski, A., & Yeaman, A. (2001). Educational Technology: The Development of a
Concept (1 ed.): LIBRARIES UNLIMITED INC.
JISC. (2003). Managing the future with MLEs [Electronic Version]. Retrieved February 02,
2003 from http://www.jisc.ac.uk/publications/publications/pub_mlesg.aspx.
Johnson, R. T. (1985). Effects of Cooperative, Competitive, and Individualistic Goal
Structures on Computer-Assisted Instruction. Journal of Educational Psychology, 77(6), 668677.
Jonassen, D. H. (1991a). Evaluating constructivistic learning. Educational Technology, 31(9),
28-33.
Jonassen, D. H. (1991b). Hypertext as instructional design. Educational Technology Research
and Development, 39(1), 83-92.
Jonassen, D. H. (1991c). Objectivism vs constructivism: do we need a new philosophical
paradigm. Educational Technology Research and Development, 39(3), 5-14.
Jonassen, D. H. (1999). Designing Constructivist Learning Environments. In C. M. Reigeluth
(Ed.), Instructional-Design Theories and Models: A New Paradigm of Instructional Theory
(Vol. 2, pp. 215-239). Mahwah, NJ: LEA.
Klein, M., & Bernstein, A. (2001). Searching for Services on the Semantic Web Using
Process Ontologies. Paper presented at the International Semantic Web Working Symposium
(SWWS), Stanford University, California, US.
Klinger, E. (1999). Thought Flow: Properties and Mechanisms Underlying Shifts in Content.
In J. L. Singer & P. Salovey (Eds.), At Play in the Fields of Consciousness: Essays in Honor
of Jerome L. Singer (1 ed., pp. 29-50): Lawrence Erlbaum Associates.
Koivunen, M.-R., Swick, R., José, K., & Prud'Hommeaux, E. (2002). Annotea: Metadata
based annotation infrastructure. Honolulu: W3C, presented at WWW 2002 Developers Day.
109
Koper, R. (2004). Use of the Semantic Web to Solve Some Basic Problems in Education:
Increase Flexible, Distributed Lifelong Learning, Decrease Teacher's Workload. Journal of
Interactive Media in Education(6), 1-23.
Koper, R. (2005). An Introduction to Learning Design. In R. Koper & C. Tattersall (Eds.),
Learning Design; A Handbook on Modeling and Delivering Networked Education and
Training (1 ed., pp. 3-20). Berlin Heidelberg: Springer verlag.
Koper, R., & Manderveld, J. (2004). Educational modeling language: modeling reusable,
interoperable, rich and personalized units of learning. British Journal of Educational
Technology, 5(35), 537-551.
Koper, R., & Tattersall, C. (2005). Learning design: A Handbook on Modeling and
Delivering Networked Education and Training. New York: Springer Berlin Heidelberg.
Kraan, W., & Wilson, S. (2003, 2002-10-01). Dan Rehak: SCORM is not for everyone.
Retrieved January 10, 2006, from http://www.cetis.ac.uk/content/20021002000737
Lama, d. N., Arroyo, S., Sicillia, M.-A., & Lópes-Cobo, J.-M. (2006). Making Learning
Objects Useful with Semantic Web Services. Paper presented at the eChallenges 2006:
Exploiting the Knowledge Economy: Issues, Applications, Case Studies, Barcelona, Spain.
Li, D., Lina, Z., Finin, T., & Joshi, A. (2005, 03-06 January). How the Semantic Web is Being
Used: An Analysis of FOAF Documents. Paper presented at the 38th Annual Hawaii
International, Hawaii.
Lysaught, J. P., & Williams, C. A. (1963). A Guide to Programmed Instruction. New York:
Wiley.
Macgregor, G., & McCulloch, E. (2006). Collaborative tagging as a knowledge organization
and resource discovery tool. Library Review, 55(7), 291-300.
Maffeis, S., & Schmidt, D. C. (1997). Constructing reliable distributed communication
systems with CORBA. Communications Magazine, IEEE, 35(2), 56-60.
Makola, D., Sim , Y. W., Wang, C., Gilbert, L., S, G., & Wills, G. (2006, 23-26 May). A
Service-Oriented Architecture for a Collaborative Orthopedic Research Environment. Paper
presented at the WWW2006, Edinburgh, Scotland.
Marton, F., & Booth, S. (2000). Om lärande (1 ed.). Lund: Studentlitteratur AB.
Masie, E. (2003, November 2003). Making Sense of Learning Specifications & Standards: A
Decision Maker's Guide to their Adoption. 2nd. Retrieved January 15, 2005, from
http://www.masie.com/standards/
Mayer, R. E. (1999). Designing Instruction for Constructivist Learning. In C. M. Reigeluth
(Ed.), Instructional-Design Theories and Models: A New Paradigm of Instructional Theory
(Vol. 2, pp. 141-159). Mahwah, NJ: LEA.
McCormick, R. (2003, August 26-30). Keeping the Pedagogy out of Learning Objects. Paper
presented at the EARLI 2003.
McGreal, R. (2004). Learning Objects: A practical Definition. International Journal of
Instructional Technology and Distance Learning, 1(9), 21-32.
McGuinness, D. L., & van Harmelen, F. (2004, 10 February). OWL Web Ontology Language
Overview. Retrieved October 10, 2007, from http://www.w3.org/TR/owl-features/
Meece, J. L. (2002). Child and adolescent development for education (2 ed.). New York:
McGraw-Hill.
110
Merrill, D. M. (1983). Component Display Theory. In C. M. Reigeluth (Ed.), InstructionalDesign Theories and Models: an overview of their current status (pp. 279-334). Hillsdale, NJ:
Lawrence Erlbaum Associates.
Merrill, D. M. (1985). Where is the authoring in authoring systems. Journal of ComputerBased Instruction, 14(1), 7-11.
Merrill, D. M. (1999). Instructional Transaction Theory (ITT): Instructional Design Based on
Knowledge Objects. In C. M. Reigeluth (Ed.), Instructional-Design Theories and Models A
new Paradigm of Instructional Theory (1 ed., Vol. 2, pp. 397-424). Mahwah, NJ: Lawrence
Erlbaum Associates, Inc.
Merrill, D. M. (2002). First Principles of Instruction. Educational Technology Research and
Development, 50(3), 43-59.
Merrill, D. M., Li, Z., & Jones, M. K. (1991). Instructional Transaction Theory. Educational
Technology, 31(6), 7-12.
Meyer, S. R. (1960). Report on the Initial Test of a Junior High-School Vocabulary Program.
In A. A. Lumsdaine & R. Glaser (Eds.), Teaching Machines and Programmed Learning: a
source book (Vol. 1, pp. 229-246). Washington, D. C.: National Education Association of the
United States.
Mohan, P., & Brooks, C. (2003). Learning objects on the semantic Web. Paper presented at
the The 3rd IEEE International Conference on Advanced Learning Technologies.
Monceaux, A., Naeve, A., Sicilia, M.-A., Garcia-Barriocanal, E., Arroyo, S., & Guss, J.
(2007). Targeting Learning Resources in Competency-based Organizations: A Semantic Webbased Approach. In J. Cardoso, M. Hepp & M. D. Lytras (Eds.), The Semantic Web; Real
World Applications from Industry (pp. 143-167). New York: Springer Verlag.
Monty, M. L., & Moran, T. P. (1986). A LONGITUDINAL STUDY OF AUTHORING
USING NOTECARDS. ACM SIGCHI Bulletin, 18(2), 59-60.
Mylov, P. (2000). Lärande i och om praktiken. In J. Bjerg (Ed.), Pedagogik, en grundbok (pp.
127-172). Stockholm: Liber AB.
Naeve, A. (2001, October 24-26). The Concept Browser - a new form of Knowledge
Management Tool. Paper presented at the 2nd European Web-based Learning Environments
Conference (WBLE 2001), Lund, Sweden.
Naeve, A. (2005). The Human Semantic Web – Shifting from Knowledge Push to Knowledge
Pull. International Journal of Semantic Web and Information Systems (IJSWIS), 1(3), 1-30.
Naeve, A., & Sicilia, M.-A. (2006, 20-23 June). Learning processes and processing learning:
From Organizational Needs to Learning Designs. Paper presented at the Adaptive
Hypermedia conference (AH'06) Dublin.
Nagarajan, M. (2006a). Semantic Annotations in Web Services. In J. Cardoso & A. P. Sheth
(Eds.), Semantic Web Services, Processes and Applications (pp. 35-61). New York: Springer
Science+Business Media, LLC.
Nagarajan, M. (2006b). The Semantic Web and its applications. In J. Cardoso & A. P. Sheth
(Eds.), Semantic Web Services, Processes and Applications (pp. 3-33). New York: Springer
Science+Business Media, LLC.
Najjar, J., Wolpers, M., & Duval, E. (2006, 23-26 May). Attention Metadata: Collection and
Management. Paper presented at the 15th International World Wide web Conference,
Edinburgh, Scotland.
111
Nelson, T. H. (1965, August 24). A file structure for the complex, the changing and the
indeterminate. Paper presented at the ACM Annual Conference/Annual Meeting, Proceedings
of the 1965 20th national conference, Cleveland, Ohio, United States.
Niederhauser, D. S., Reynolds, R. E., Salmen, D. J., & Skolmoski, P. (2000). The Influence of
Cognitive Load on Learning from Hypertext. Journal of Educational Computing Research,
23(3), 237-255.
Nielsen, J. (1995). Hypertext on the Internet. In Multimedia and Hypertext: The Internet and
Beyond (1 ed., pp. 165-216). London: Morgan Kaufmann Publishers.
Nierstrasz, O., & Dami, L. (1995). Component-Oriented Software Technology. In O.
Nierstrasz & D. Tsichritzis (Eds.), Object-Oriented Software Composition (1 ed., pp. 3-28).
Hertfordshire: Prentice Hall International (UK) Ltd.
Nilsson, B. (2001). Socialkonstruktivism - positioner, problem och perspektiv (1 ed.). Malmö:
Liber AB.
Nilsson, M. (2001). The Semantic Web: How RDF will change learning technology standards
[Electronic
Version].
Retrieved
April
11,
2006
from
http://zope.cetis.ac.uk/content/20010927172953. Centre for Educational Technology
Interoperability Standards (CETIS).
Nilsson, M., Johnston, P., Naeve, A., & Powell, A. (2007). The Future of Learning Object
Metadata Interoperability. In K. Harman & A. Koohang (Eds.), Learning Objects: Standards,
Metadata, Repositories, and LCMS (pp. 255-313): Informing Science press.
Nilsson, M., Naeve, A., Duval, E., Johnston, P., & Massart, D. (2008). ProLEARN
Deliverable D4.7: Harmonization of Metadata Standards. ProLEARN Deliverables. Retrieved
10 February, 2008, from http://ariadne.cs.kuleuven.be/lomi/images/5/52/D4.7-prolearn.pdf
Nilsson, M., Palmér, M., & Brase, J. (2003). The LOM RDF Binding - Principles and
Implementation. Proceedings of the Third Annual ARIADNE conference.
Nilsson, M., Palmér, M., & Naeve, A. (2002). Semantic Web Meta-data for e-Learning Some Architectural Guidelines. Paper presented at the 11th World Wide Web Conference
(WWW2002), Hawaii, USA.
Norton, M. J. (2004, April 5, 2004). A Comparison between the JISC and Sakai Frameworks.
3rd. Retrieved May 30, 2006, from http://sakaiproject.org/tech/S040405N.pdf
Nussbaumer, A., & Mistelbacher, A. (2002). XML ge-packt : XSLT, CSS und XSL-FO,
XML-Schema, XPath, Xlink und XPointer (1. Aufl. ed.). Bonn: mitp.
O'Reilly, T. (2006, June 24, 2006). What Is Web 2.0 Design Patterns and Business Models for
the Next Generation of Software. Retrieved October 27, 2007, from
http://facweb.cti.depaul.edu/jnowotarski/se425/What%20Is%20Web%202%20point%200.pdf
OECD. (2007). Giving Knowledge for Free: The emergence of open educational resources:
OECD, Centre for Educational Research and Innovation (CERI).
Olivier, B., Roberts, T., & Blinco, K. (2005, July 2005). The eFramework for Education and
Research: An Overview. R1. Retrieved June 30, 2007, from http://www.eframework.org/Portals/9/Resources/eframeworkrV1.pdf
Olivier, B., & Tattersall, C. (2005). The Learning Design Specification. In R. Koper & C.
Tattersall (Eds.), Learning Design
112
A Handbook on Modeling and Delivering Networked Education and Training (1 ed., pp. 2140). Berlin Heidelberg: Springer. verlag.
Orlich, D. C. (2000). Teaching Strategies: A Guide to Better Instruction (6 ed.). New York:
Houghton Mifflin Company.
OSI. (2007). The Open Source Definition. Retrieved February 06, 2007, from
http://www.opensource.org/docs/definition.php
Palmér, M., Naeve, A., & Paulsson, F. (2004). The SCAM Framework: Helping Semantic
Web Applications to Store and Access Metadata. Paper presented at the the European
Semantic Web Symposium 2004, Heraclion Greece.
Papert, S. (1994). Hur gör giraffen när den sover? Göteborg: Daidalos AB.
Parrish, P. E. (2004). The Trouble with Learning Objects. Educational Technology Research
& Development (ETR & D), 52(1), 49-67.
Pask, G. (1960). Adaptive Teaching With Adaptive Machines. In A. A. Lumsdaine & R.
Glaser (Eds.), Teaching Machines and Programmed Learning: a source book (Vol. 1, pp.
349-366). Washington, D. C.: National Education Association of the United States.
Patriarche, R., Diallo, G., Bernhard, D., Bassolet, C. G., Ferriol, S., & Gedzelman, S. (2005,
19-20 October). A Tool for Textual and Conceptual Annotations of Documents. Paper
presented at the eChallenges 2005, Ljubljana, Slovenia.
Patrick, P. (2005, June 13-16). Impact of SOA on enterprise information architectures. Paper
presented at the ACM SIGMOD international conference on Management of data, Baltimore,
Maryland.
Paulsen, M. F. (2002, July 2002). Online Education Systems: Discussion and Definition of
Terms
Retrieved
2007-01-20,
2007,
from
http://www.nettskolen.com/forskning/Definition%20of%20Terms.pdf
Paulson, L. D. (2005). Building Rich Web Applications with Ajax. Computer, 38(10), 14-17.
Paulsson, F. (1998). Towards a Virtual Teacher College. Retrieved December 12, 2006, from
http://www.educ.umu.se/~frepa/presentations/1997/vtc.pdf
Paulsson, F. (2002). Märkning av nationella styrdokument (pp. 1-8). Stockholm: Skolverket.
Paulsson, F. (2003). Standardized Content Archive Management – SCAM. IEEE Learning
Technology newsletter, 5(1), 40-42.
Pavlov, I. P. (1927). Conditioned Reflexes: An Investigation of the Physiological Activity of
the Cerebral Cortex (G. V. Anrep, Trans.). London: Oxford University Press.
Pedersen, J. (1998). Informationstekniken i skolan. En forskningsöversikt. (No. 98:343).
Stockholm: Skolverket.
Pepper, S. (2000). The Tao of Topic Maps: finding the way in the age of infoglut. Paper
presented at the XML Europe 2000, Paris, France.
Pepper, S., Vitali, F., Garshol, L. M., Gessa, N., & Presutti, V. (2005, 29 March, 2005). A
Survey of RDF/Topic Maps Interoperability Proposals. Retrieved April 2, 2005, from
http://www.w3.org/TR/rdftm-survey/
Piaget, J. (1968/2006). Barnets själsliga utveckling. Stockholm: Norstedts Akademiska
Förlag.
113
Piaget, J., & ebrary Inc. (1959/2002). Judgment and reasoning in the child. London:
Routledge.
Piaget, J., & Inhelder, B. (1997). Selected works.
Piccoli, G., Ahmad, R., & Ives, B. (2001). Web-Based Virtual Learning Environments: A
Research Framework and a Preliminary Assessment of Effectiveness in Basic IT Skills
Training. MIS Quarterly, 25(4), 401-426.
Polsani, P. R. (2003). Use and Abuse of Reusable Learning Objects. Journal of digital
information, 3(4).
Powell, A., Nilsson, M., Naeve, A., Johnston, P., & Baker, T. (2004, 4 June 2007). DCMI
Abstract Model. Retrieved 10 February, 2008, from http://dublincore.org/documents/abstractmodel
Pressey, S. L. (1927). A machine for automatic teaching of drill material. School and Society,
25(645), 549-552.
Psyche, V., Bordeau, J., Nkambou, R., & Mizoguchi, R. (2005). Making Learning Design
Standards Work with an Ontology of Educational Theories. Paper presented at the 12th
International Conference on Artificial Intelligence in Education, Amsterdam.
Quinn, C., & Hobbs, S. (2000). Learning Objects and Instruction Components. Educational
Technology & Society, 3(2).
Rawlings, A., Koper, R., van Rosmalen, P., Rodrígues-Artacho, M., & Lefrere, P. (2002).
Survey of Educational Modeling Languages (EMLs).
Rehak, D., & Mason, R. (2003). Keeping Learning in Learning Objects, in the Special Issue
on Reusing Online Resources: A Sustainable Approach to eLearning. Journal of Interactive
Media in Education, 2003(1), 22-27.
Reigeluth, C. M. (1983). Instructional-design theories and models: an overview of their
current status (Vol. 1). Hillsdale, NJ: Lawrence Erlbaum Associates.
Reigeluth, C. M. (1999). A new paradigm of instructional theory (1 ed. Vol. 2). Mahwah, N.J.
: Lawrence Erlbaum.
Reigeluth, C. M. (1999a). Instructional-Design Theories and Models: A New Paradigm of
Instructional Theory (Vol. II). Mahwah, NJ: LEA.
Reigeluth, C. M. (1999b). The Elaboration Theory: Guidance for Scope and Sequence
Decisions. In C. M. Reigeluth (Ed.), Instructional-Design Theories and Models: A New
Paradigm of Instructional Theory (Vol. 2, pp. 425-453). Mahwah, NJ: Lawrence Erlbaum
Associates, Inc.
Reigeluth, C. M. (1999c). What is Instructional-Design Theory and how is it changing? In C.
M. Reigeluth (Ed.), Instructional-Design Theories and Models: A New Paradigm of
Instructional Theory (Vol. 2, pp. 5-31). Mahwah, NJ: LEA.
Reigeluth, C. M. (2008). Instructional-design theories and models, Volume III (1 ed. Vol. 3).
Mahwah, N.J. : Routledge.
Richter, C., Nejdl, W., & Allert, H. (2002). Learning Objects on the Semantic Web. Explicitly
Modelling Instructional Theories and Paradigms. Paper presented at the World Conference
on E-Learning in Corporate, Government, Healthcare, and Higher Education, Chesapeake,
VA.
114
Rodriguez, M. A., & Egenhofer, M. J. (2003). Determining Semantic Similarity Among
Entity Classes from Different Ontologies. IEEE Transactions on Knowledge and Data
Engineering, 15(2), 442-456.
Roman, D., Keller , U., Lausen, H., de Bruijn, J., Lara, R., Stollberg, M., et al. (2005). Web
Service Modeling Ontology. Applied Ontology, 1(1), 77-106.
Saettler, P. (2004). The Evolution of Educational American Technology. Connecticut:
Information Age Publishing.
Salmenjoki, K., Paulsson, F., & Vainio, A. (2007, 6-8 September). Social Learning - How
Web is Changing the Way We Process Information and Knowledge. Paper presented at the 5th
International Conference on Emerging e-learning Technologies and Applications (ICETA),
Stará Lesná, Slovak Republic.
Schank, R. C., Berman, T. R., & Macpherson, K. A. (1999). Learning by Doing. In C. M.
Reigeluth (Ed.), Instructional-Design Theories and Models: A New Paradigm of Instructional
Theory (Vol. 2, pp. 161-181). Mahwa, NJ: LEA.
Schluep, S., Bettino, M., & Guttormsen Schär, S. (2005, March 3-4, 2005). Modularization
and Structured Markup for Learning Content in an Academic Environment. ProLearn-iClass
Thematic Workshop on Learning Objects in context. Retrieved August 27, 2005, from
http://www.cs.kuleuven.ac.be/~hmdb/ProlearnIClass/papers/Schluep.pdf
Schult,
L.
Bloom's
Taxonomy.
Retrieved
http://www.odu.edu/educ/llschult/blooms_taxonomy.htm
October
10,
2007,
from
Schunk, D. H. (1991). Learning Theories: An Educational Perspective (1 ed.). New York:
Macmillan.
Schunk, D. H. (2003). Learning Theories An Educational Perspective (4 ed.). New Jersey:
Pearson Higher Education.
Seels, & Richey. (1994). Instructional technology: The definition and domains of the field.
Bloomington, IN: Association for Educational Communications and Technology.
Sessions, R. (2001, March 28, 2001). Java 2 Enterprise Edition (J2EE) versus The .NET
Platform Two Visions for eBusiness. Retrieved October 01, 2007, from
http://www.objectwatch.com/FinalJ2EeandDotNet.doc
Shadbolt, N., Hall, W., & Berners-Lee, T. (2006). The Semantic Web Revisited. IEEE
Intelligent Systems, 21(3), 96-101.
Shaw, R. (2005, December 07). Web 2.0? It doesn't exist. Retrieved February 16, 2007, from
http://blogs.zdnet.com/ip-telephony/?p=805
Shuell, T. J. (1990). Phases of meaningful learning. Review of Educational Research, 60(4),
531-547.
Silveira, I. F., Fernando deAraújo Jr., C., Amaral, L. H., Alcântara de Oliviera, I. C.,
Schimiguel, J., Ledón, F.-P., Manuel, et al. (2007). Granularity and Reusability of Learning
Objects. In A. Koohang & K. Harman (Eds.), Learning Objects and Instructional Design (pp.
139-170). Santa Rosa, CA: Informing Science Press.
Sim, Y. W., Wang, C., Carr, L. A., Davis, H. C., Gilbert, L., Grange, S., et al. (2005, 19-22
September). A Web/Grid Services Approach for a Virtual Research Environment
Implementation. Paper presented at the Fourth e-Science All Hands Meeting (AHM 2005),
Nottingham, UK.
115
Sim, Y. W., Wang, C., Gilbert, L., & Wills, G. B. (2005). An Overview of Service-Oriented
Architecture (Technical Report No. ECSTR-IAM05-004). Southampton: Electronics and
Computer Science, University of Southampton.
Simpson, T. L. (2002). Dare I oppose constructivist theory? The Educational Forum(66), 347354.
Skinner, B. F. (1958). Teaching Machines. Reprinted from Science, 128 (3330), 969-977.
Skinner, B. F. (1960). Teaching Machines. In A. A. Lumsdaine & R. Glaser (Eds.), Teaching
Macines and Programmed Learning: a source book (Vol. 1, pp. 137-158). Washington, D. C.:
National Education Association of the United States.
Skinner, B. F. (1976). About Behaviourism. New York: Vintage Books.
Smith, J. (1999, October 29, 1999). Browser Compatibility Chart. Retrieved February 12,
2008, from http://labnt1.bizlab.mtsu.edu/Lab/tectrain/BrowserCompatibility.htm
Snelbecker, G. E. (1974). Learning theory, instructional theory, and psychoeducational
design. New York.
Snelbecker, G. E. (1999). Current Progress, Historical Perspective, and Some Task of the
Future of Instructional Design Theory. In C. M. Reigeluth (Ed.), Instructional-Design
Theories and Models: A New Paradigm of Instructional Theory (Vol. 2, pp. 653-674).
Mahwa, NJ: LEA.
Song, W. W., & Andersson, M. B. (1999, October 24-30). A Component-based Framework
for Description and Management of Learning Objects. Paper presented at the WebNet 99
World Conference on the WWW and Internet, Honolulu, Hawaii.
Song, W. W., Andersson, M. B., Hakkarainen, S. E., Karlberg, P., & Soltesz, T. (1999).
Metadata for the management of electronic documents in the governmental organizations and
learning objects in the learning domain (Research report No. SITI 99:03). Kista: SITI, SISU.
Sosteric, M., & Hesemeier, S. (2002). When is a Learning Object not an Object: A first step
towards a theory of Learning objects. International Review of Research in Open and Distance
Learning, 3(2).
Subramaniam, M. A. (2002, Summer 2002). Garden-Based Learning in Basic Education: A
Historical Review. Monograph (Focus). Retrieved September 13, 2007, from
http://cyd.ucdavis.edu/publications/monograph.html
Szyperski, C. (2002a). Component Software - Beyond Object-Oriented Programming (2 ed.
Vol. 1). New York: ACM Press.
Szyperski, C. (2002b). Object and component "wiring" standards. In Component Software Beyond Object-Oriented Programming (2 ed., Vol. 1, pp. 207-230). New York: ACM Press.
Szyperski, C. (2002c). What a component is and is not. In Component Software - Beyond
Object-Oriented Programming (2 ed., Vol. 1, pp. 35-48). New York: ACM Press.
Szyperski, C., & Pfister, C. (1997). Workshop on Component-Oriented programming,
Summary. Special Issues in Object-Oriented Programming - ECOOP ´96.
Söderlund, A. (2000). Det långa mötet. IT och Skolan. Om spridning och anammande av IT i
den Svenska skolan. Centrum för forskning i lärande: Luleå tekniska universitet.
Tennyson, R., Schott, F., Seel, N., & Dijkstra, S. (1997). Instructional design: International
perspectives. Mahwah, NJ: Erlbaum.
116
Thorne, S., & Kahn, J. (2003). O.K.I.™ Architectural Concepts (Technical report (draft)):
Massachusetts Institute of Technology.
Trigg, R. H., & Irish, P. M. (1987, November 13-15). Hypertext Habitats: Experiences of
Writers in NoteCards. Paper presented at the Hypertext'87, Chapel Hill, North Carolina, USA.
Walker, S. F. (1990). A brief history of connectionism and its psychological implications. AI
& Society, 4(1), 17-38.
Van Assche, F., & Vuorikari, R. (2006). A Framework for Quality of Learning Resources. In
U.-E. Ehlers & J. P. Pawlowski (Eds.), European Handbook for Quality (pp. 443-456):
Springer Berlin Heidelberg.
Warren, P. (2006). Knowledge Management and the Semantic Web: From Scenario to
Technology. IEEE Intelligent systems, 21(1), 53-59.
Varsha, A., Hansen, T., & Reeser, P. (2002). Performance comparison of dynamic web
platforms. Computer Communications, 26(8), 888-898.
Watson, J. B. (1925). Behaviorism. New York: The People's Institute Publishing Co.
Watson, J. B., & Rayner, R. (1920). CONDITIONED EMOTIONAL REACTIONS. Journal
of Experimental Psychology, 3(1), 1-14.
Weinreich, R., & Sametinger, J. (2001). Component Models and Component Services:
Concepts and Principles. In G. T. Heinemann & W. T. Council (Eds.), Component-Based
Software Engineering: putting the pieces together (pp. 33-48). Upper Saddle River, NHJ:
Adison-Wesley.
Weller, M. j., Pegler, C. A., & Mason, R. D. (2003, 9-12 February). Putting the pieces
together: What working with learning objects means for the educator. Paper presented at the
eLearn International, Edinburgh, Scotland.
Ver, M. V., E, M., Dominigue, J., Lanzoni, M., A, S., & Ciravegna, F. (2003, July 22-26).
Ontology Driven Tool for Semantic Markup. Paper presented at the ECAI 2002 Workshop on
Semantic Authoring, Annotation & Knowledge Markup, Lyon.
Wikipedia. (2007, April 05, 2007). Dynamic web page - Wikipedia, the free encyclopedia.
Retrieved April 09, 2007, from http://en.wikipedia.org/wiki/Dynamic_Web_pages
Wiley, D. A. (1999). The Post-LEGO Learning Object. Retrieved 16 March, 2002, from
http://wiley.ed.usu.edu/docs/post-lego.pdf
Wiley, D. A. (2000). Learning Object Design and Sequencing Theory. Bringham Young
University.
Wiley, D. A. (2002a). Connecting learning objects to instructional design theory: A
definition, a metaphor, and a taxonomy. In D. A. Wiley (Ed.), The Instructional Use of
Learning Objects (First ed., pp. 3-23). Bloominton: Agency for Instructional Technology and
Association for Educational Communications & Technology.
Wiley, D. A. (2002b). The Instructional Use of Learning Objects (First ed.). Bloominton:
Agency for Instructional Technology and Association for Educational Communications &
Technology.
Wiley, D. A. (2003). Learning Objects: Difficulties and opportunities. Retrieved 20 February,
2008, from http://wiley.ed.usu.edu./docs/lo_do.pdf
Wiley, D. A. (2007). On the Sustainability of Open Educational Resource Initiatives in
Higher Education: OECD, Centre for Educational Research and Innovation (CERI).
117
Williams, J. (2001). The Business Case for Components. In G. T. Heinemann & W. T.
Council (Eds.), Component-Based Software Engineering: Putting the Pieces Together (1 ed.,
Vol. 1, pp. 79-97): Addison-Wesley.
Wilson, S., Blinco, K., & Rehak, D. (2004, July, 2004). An e-Learning Framework A
Summary. Paper presented at the Alt-I-lab 2004.
Wilson, S., Olivier, B., Jeyes, S., Powell, A., & Franklin, T. (2004). A Technical Framework
to
Support
e-Learning.
Retrieved
August
28,
2005,
from
http://www.jisc.ac.uk/uploaded_documents/Technical%20Framework%20feb04.doc
Wilson, S., & Rees Jones, P. (2005). What Is...IMS Learner Information Packaging? CETIS
briefings on e-learning standards series. Retrieved October 05, 2005, from
http://www.cetis.ac.uk/lib/media/lipbrief.pdf
Voss, J. (2007, July 18-24). Tagging, Folksonomy & Co - Renaissance of Manual Indexing?
Paper presented at the 10th International Symposium of Information Science, Cologne,
Germany.
Vygotsky, L. (1978). Mind in society: the development of higher psychological process.
Cambridge: Harvard University Press.
Yoder, J., & Barcalow, J. (1997, September 3-5). Architectural Patterns for Enabling
Application Security. Paper presented at the 4th Pattern Languages of Programming
Conference, Monticello, Illinois, USA.
Yu, L. (2007). Introduction To the Semantic Web and Semantic Web Services (1 ed.). Boca
Rayton, FL: Taylor & Fancis Group.
Zaihrayen, I., Sun, L., Giunchiglia, F., Pan, W., Ju, Q., Chi, M., et al. (2007, November
2007). From Web Directories to Ontologies: Natural Language Processing Challenges. Paper
presented at the 6th International Semantic Web Conference, ISWC 2007 + ASWC 2007,
Busan, Korea.
118
Appendix A – List of acronyms
This appendix contains a list of the acronyms that are most frequently used in this thesis. The
acronyms are only briefly described in the table since most of them are explained and
discussed in detail in chapter 3 and 5.
Acronym
Concept
Short description
AJAX
Asynchronous JavaScript and XML
A collection of technologies and
principles for dynamic user interfaces
on the web
CBI
Computer-Based Instruction
Instructional concept using computers.
CBT
Computer-Based Training
Instructional concept using computers.
CGI
Common Gateway Interface
A standard for interfacing with web
servers.
DTD
Document Type Definition
A schema language for XML and
SGML.
EML
Educational Modeling Language
A collective term for metalanguages
that describe pedagogical process.
FOAF
Friend of a Friend
A schema for representing people and
social networks.
HTML
Hypertext Markup Language
The original markuplanguage of the
web.
HTTP
Hypertext Transfer Protocol
A communication protocol for the
Internet and the web
LMS
Learning Management System
Software system for managing courses
and learning content.
LOM
Learning Object Metadata
A metadata model for describing
learning objects.
LOR
Learning Object Repository
Repository for learning resources and
metadata.
OKI
Open Knowledge Initiative
The initiative behind the OSID.
OSID
Open Service Interface Definitions
OKI’s Software (service) interfaces for
SOA.
OWL
Web Ontology Language
An ontology markup language.
RDF
Resource Description Framework
RDFS
RDF Schema
A framework for representing data on
the Semantic Web.
Provides basic elements for describing
ontologies with RDF.
REST
Representational state transfer
An architecture style that utilizes the
characteristics of the web.
119
RMI
Remote method Invocation
A Java technology
invocations.
RSS
Really Simple Syndication, Rich Site Formats for syndicating content.
Summary or RDF Site Summary
SCORM
Sharable Content Object Reference A reference model for sharable content
Model
and LMSes.
SGML
Standard
Language
SOA
Service Oriented Architecture
An architectural model for loosely
coupled services.
SOAP
Simple Object Access Protocol
Protocol for exchanging XML-based
messages over the Internet.
TEI
Text Encoding Initiative
A markuplanguage for digital texts.
TEL
Technology Enhanced Learning
Learning that are enhanced using ICT.
UDDI
Universal Description,
and Integration
VLE
Virtual Learning Environment
General term for ICT-based learning
environments.
W3C
The World Wide Web Consortium
A
consortium
that
promotes
interoperable technologies for the web.
WBL
Web-Based Learning
Education and learning distributed on
the web.
WBT
Web-Based Training
Education and training distributed on
the web.
WSDL
Web Service Description Language
A description
services.
XML
Extensible Markup Language
A standard
languages.
XML-RPC
XML Remote Procedure Call
An XML-based protocol for remote
calls over http.
XSLT
Extensible Stylesheet
Transformations
Generalized
for
remote
Markup A metalanguage for defining markup
languages.
Discovery A Web Service registry.
language
for
defining
for
web
markup
Language A language for transforming XML
documents.
120
Appendix B – Appended papers
This appendix contains the six papers that are appended to this thesis
121
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement