A Framework to Personalise Open Learning Environments by
Adapting to Learning Styles
Heba Fasihuddin, Geoff Skinner and Rukshan Athauda
Faculty of Science and Information Technology, The University of Newcastle, Callaghan, Australia
Keywords: Adaptive Framework, Learning Styles, MOOCs, Open Learning, Personalisation.
Abstract: This paper presents an adaptive framework to personalise open learning environments. The design of the
framework has been grounded in cognitive science and learning principles. The theory of learning styles,
and more specifically the model of Felder and Silverman, has been considered and applied. The developed
framework has two main functions. First, it automatically identifies the learners’ learning styles by tracking
their behaviours and interactions with the provided learning objects. Secondly, it provides adaptive
navigational support based on the identified learning styles. Sorting learning materials based on learners’
preferences and hiding the least preferred materials are the two techniques of navigational support that have
been applied in the proposed framework. Detailed descriptions of the framework functionalities and
different components are presented in this paper. Future piloting and evaluation will test and verify this
proposed framework.
1 INTRODUCTION
Online learning evolves to take advantage of
continuous advancement of technology. Open
learning is a form of online learning that allows
learning materials to be freely available on the
Internet for anyone who is interested.
Currently, several prestigious learning
institutions, such as Harvard, MIT and Stanford,
provide learning materials in an open approach.
Coursera (Coursera, 2012), edX (edX, 2012),
Udacity (Udacity, 2012) and Udemy (Udemy, 2014)
are examples of open learning initiatives. Courses
that are provided through these open learning
environments are known as Massive Open Online
Courses (MOOCs).
As with any new model for learning, MOOCs are
still in their early stages of evolution. There are
many areas and opportunities for improvement, such
as teaching and learning methods; learning content;
assessments; identity authentication; accreditation;
and learners’ varying needs, among others. The
authors believe that considering cognitive science
and learning principles has opportunity to enhance
learning environments such as MOOCs (Fasihuddin
et al., 2013b). This view is also supported by others
(Williams, 2013).
This paper focuses on personalisation of open
learning environments based on learning styles.
Learning style refers to the way a learner receives
and processes information; therefore, every learner
has a different learning style (Felder and Silverman,
1988). Among the existing models of learning styles,
Felder and Silverman Learning Style Model
(FSLSM) was selected. This paper proposes an
adaptive framework that identifies the learners’
learning styles and consequently provides
personalised navigational support. The literature-
based approach (Graf, 2007) is used to automatically
identify the learning style. This approach has been
shown to have higher accuracy of results in
detecting learning styles (Graf, 2007). It is mainly
based on monitoring the learners’ behaviours on
determined patterns based on the FSLSM. These
patterns are determined based on learning objects
that are common in open learning environments,
such as in edX, Coursera, Udemy and Udacity.
Based on our knowledge, no previous study has
attempted to personalise the open environment using
learning styles, and this is what distinguishes this
study and the proposed framework.
The rest of this paper is organised as follows:
first, a background of the related concepts is
presented in section 2; next, section 3 presents a
review of previous work on adaptive systems based
on learning styles; after that, the proposed adaptive
framework and the development of the prototype are
296
Fasihuddin H., Skinner G. and Athauda R..
A Framework to Personalise Open Learning Environments by Adapting to Learning Styles.
DOI: 10.5220/0005443502960305
In Proceedings of the 7th International Conference on Computer Supported Education (CSEDU-2015), pages 296-305
ISBN: 978-989-758-107-6
Copyright
c
2015 SCITEPRESS (Science and Technology Publications, Lda.)
presented in sections 4 and 5 respectively; and
finally, the paper is concluded in section 6 with a
brief overview of future work.
2 BACKGROUND
2.1 Open Learning Environments
As mentioned, the evolution of technology leads to
continual change and development in online learning
approaches and, recently, open learning has emerged
as a new form of online learning. In open learning,
courses are freely available on the Internet to be
accessed by anyone who is interested. These courses
are provided by different learning providers who
could be academics representing learning institutions
or individuals who have appropriate knowledge and
expertise. Recently many courses have been offered
in this form by different prestigious institutions,
such as Stanford and MIT. These courses have
gradually refined into what are known as Massive
Open Online Courses (MOOCs).
MOOCs offer free university-level courses
online and have two key features – open access and
scalability (Yuan and Powell, 2013). These two
features allow MOOCs to be taken online by anyone
and enable the courses to be designed to support an
indefinite or even infinite number of participants.
They are learner-centred courses so learners are able
to work and learn at their own pace. The massive
number of learners in MOOCs leads to significant
variation in these learners’ needs, preferences and
even cognitive abilities. Therefore, personalisation
of MOOCs is essential.
The development of open learning environments
is a critical field due to the implications for learners,
instructors and the learning process. Therefore,
scientific principles for learning should be
considered in the development of these
environments in order to achieve the desired
learning goals. It is stated by Williams (2013) that
tailoring general learning principles and working
with cognitive scientists is one approach that needs
to be considered to enhance MOOCs and provide the
best outcomes for learners. Based on this, the
authors have considered the theory of learning styles
(Felder and Silverman, 1988) to introduce an
approach for personalising open learning
environments. This should increase learners’
satisfaction and lead to better learning outcomes.
Following is an overview of this theory and its
implications.
2.2 The Theory of Learning Styles
Learning style refers to the way a learner receives
and processes information. Therefore, different
learners will have different learning styles (Felder
and Silverman, 1988). Considering learning styles in
courseware design has been found to be effective
and beneficial in learning. It has been shown that
providing learners with learning materials and
activities that suit their preferences and learning
styles makes learning easier for them (Graf and Tzu-
Chien, 2009). This has been shown by many studies
that found that students can achieve better learning
outcomes and higher scores (Bajraktarevic et al.
2003), and can also master the learning materials in
less time (Graf and Kinshuk, 2007).
In literature, several models for learning styles
were defined and found to be valid and reliable
(Coffield et al., 2004). Felder and Silverman
Learning Style Model (FSLSM) was selected as the
most appropriate by the authors to be applied to
personalise an open learning environment. A number
of reasons led to this selection. The mechanism of
the FSLSM Index of Learning Style (ILS)
questionnaire (Soloman and Felder, nd) that
identifies learning styles can be easily applied to
adaptive systems. Furthermore, it has been shown
that the FSLSM is the most appropriate and feasible
model to be implemented in adaptive courseware
(García et al., 2008; Carver et al., 1999). Moreover,
a study that was conducted to compare the suitability
of different learning style models to be applied to
online learning also concluded that the FSLSM was
the most appropriate model (Kuljis and Liu, 2005).
The FSLSM classifies learning styles into four
dimensions and identifies two types of learners for
each dimension. The dimensions are perception,
input, processing and understanding. Firstly, the
perception dimension defines the type of
information that learners prefer to receive and learn
by: intuitive learners prefer meaning and theories
while sensory learners prefer learning by examples
and practice. The second dimension is input which
defines the approach the learners prefer to learn
with: visual learners prefer pictures, diagrams and
flowcharts while verbal learners prefer written or
spoken explanations. The processing dimension
indicates how learners prefer to process and practice
their learning: active learners prefer working with
others while reflective learners prefer thinking and
working alone. Finally, the understanding dimension
indicates how learners progress toward
understanding: sequential learners learn in continual
small steps while global learners learn holistically in
AFrameworktoPersonaliseOpenLearningEnvironmentsbyAdaptingtoLearningStyles
297
large jumps. Table 1 represents these styles and their
associated types.
Table 1: Felder and Silverman learning styles.
Dimension Preferred Learning Styles
Perception Sensory Intuitive
Input Visual Verbal
Processing Active Reflective
Understanding Sequential Global
2.3 Adaptive Systems with Respect to
Learning Styles
Adaptive systems have been described as systems
that are able to provide personalised learning support
to the learner throughout their interaction based on
the goals, preferences and knowledge of each
individual learner (Brusilovsky, 2001). It has been
found that adaptive learning systems lead to better
learning outcomes, reduce time and effort required,
and increase learners’ satisfaction (Graf and
Kinshuk, 2014). Adaptive systems can adapt to user
data, usage data and environment data (Brusilovsky,
2001). User data refers to various characteristics of
the user, such as learning styles and cognitive traits.
Usage data refers to the user’s interaction with the
system. Environment data refers to the adaptation to
the user context, including location or platform.
Providing adaptability based on these considered
factors has been classified into two different areas –
adaptive presentation and adaptive navigation
support (Brusilovsky, 2001). Adaptive presentation
comprises text and multimedia adaptation
technologies while adaptive navigation support
comprises link sorting and hiding, and providing
direct guidance.
Systems that are adaptive to learning styles need
to identify the learner’s learning style first and then
adapt to the learner’s preferences. Adaptation
methods of adaptive systems have been classified
into two different approaches – collaborative and
automatic (Brusilovsky, 1996). In the collaborative
approach, learners are asked to provide their
preferences explicitly by taking a test or filling out a
questionnaire, such as the ILS questionnaire
(Soloman and Felder, nd), in order to build the
adaptable models while in the automatic approach,
the learners’ adaptable model is built automatically
by the adaptive system through intelligent and
machine learning techniques that exploit learners’
interactions and behaviours while they are using the
system for learning.
In literature, two different methods for
identifying learning styles based on the FSLSM
were used – the data-driven method and the
literature-based method (Graf, 2007). Both methods
rely on some identified patterns to detect the
learning style of the learner. These patterns are
based on monitoring the provided learning objects in
such a way that they adhere to the FSLSM. The
data-driven method aims to build a model that
imitates the ILS questionnaire and uses sample data
to construct a model. Some of the techniques that
have been used to apply this method are neural
networks, decision tree, Hidden Markov Model,
fuzzy clustering and Bayesian networks. The
literature-based method uses the behaviour of
students and their actions with the systems while
they are learning in order to identify their learning
style preferences. Patterns are identified based on
findings of learners’ preferences and behaviours for
each specific learning style. This method uses
simple rules to calculate learning styles. A study
conducted to compare the efficiency of these two
methods in detecting learning styles found that the
literature-based method gives more accurate results
than the data-driven method (Graf, 2007). Although
the literature-based method has been found to be
efficient, it has been claimed by Ahmad et al. (2013)
that this method’s point of weakness is embodied in
the possibility of not considering all the potential
patterns that could affect the detection of learning
styles. Many studies have been conducted to
automatically identify learning styles and the
following section provides an overview.
3 RELATED WORK
Building adaptive systems that adapt to learners’
learning styles has been a point of interest in
research. Different studies have been conducted to
provide adaptive learning based on learning styles.
Some of these studies were based on the
collaborative adaptive approach where students were
asked to provide their preferences through answers
to the ILS questionnaire while others were based on
the automatic approach where their learning styles
were detected automatically through their
behaviours and interactions with the systems. In
literature, a variety of methods and techniques were
used. These methods differ based on the attributes
that were used for detecting learning styles
(personality factors, behaviour factors), the
underlying technique (literature-based, data-driven)
and the underlying infrastructure (Learning
Management Systems, special user interface).
CSEDU2015-7thInternationalConferenceonComputerSupportedEducation
298
Various studies have considered the variety of
learning styles and the importance of incorporating
them into learning environments. Some of these
studies were concerned with introducing models and
approaches to incorporate FSLSM into adaptive
systems based on the collaborative approach (Carver
et al., 1999; Hong and Kinshuk, 2004; Gilbert and
Han, 1999). Recently, studies have been more
concerned with automatically detecting the learners’
learning styles rather than using the collaborative
approach. As stated, two main approaches are found
in literature for learning style identification – the
data-driven approach and the literature-based
approach. In the data-driven approach, some data
mining and machine learning algorithms were used
to automatically identify the learners’ learning
styles. Some examples are: Bayesian networks
(Carmona et al., 2008; García et al., 2007), neural
networks (Cabada et al., 2009; Latham et al., 2013),
decision tree, the Hidden Markov Model (Cha et al.,
2006), NBTree (Özpolat and Akar, 2009), k-nearest
neighbour algorithm along with genetic algorithm
(Chang et al., 2009), and the AprioriAll mining
algorithm (Klašnja-Milićević et al., 2011). Graf was
the first to use the literature-based approach to
automatically identify learning styles (Graf, 2007;
Graf et al., 2008). She determined different patterns
of learner behaviours and actions based on common
learning objects in LMSs to identify learning styles.
Other studies have also used this approach to
identify some dimensions of the FSLSM (Ahmad et
al., 2013; Şimşek et al., 2010; Atman et al., 2009).
4 THE PROPOSED
FRAMEWORK FOR ADAPTIVE
OPEN LEARNING
The objective of the framework is to perform two
main functions: 1) identify the learners’ learning
style, and 2) recommend suitable learning materials
and organise them in a way that the learner prefers.
The recommendation and organisation of learning
materials has been designed through providing
navigational guidance and support based on the
preferred learning style.
The adaptive engine consists of two main agents
to perform the desired functionalities: 1) a learning
style identification agent; and 2) a recommender
agent. The identification agent is responsible for
identifying the learning styles and storing them in
the learners’ profiles which is used by the
recommender agent to provide the desired
adaptability and navigational support to learners. An
illustration of the proposed adaptive framework is
provided in Figure 1.
Figure 1: An illustration of the adaptive framework.
4.1 Learning Style Identification Agent
The design of the identification agent has been based
on the literature-based method (Graf, 2007), which
requires some determined patterns of learner
interactions with the provided learning objects to be
monitored in order to identify the learning styles.
Graf was the first to use the literature-based method
to automatically identify learning styles using the
simple rule-based technique (Graf, 2007; Graf et al.,
2008). She determined different patterns of learners’
behaviours and actions based on common learning
objects in LMSs that are used in blended learning.
This approach has been shown to have higher
accuracy of results in detecting learning styles (Graf,
2007).
As our study is looking at open learning
environments, determining patterns for identifying
learning styles should be based on the learning
objects in these environments. For that, the authors
have observed learning objects provided in well-
known MOOCs, such as edX (edX, 2012), Coursera
(Coursera, 2012), Udemy (Udemy, 2014) and
Udacity (Udacity, 2012). The identified learning
objects include course overviews, outlines, video
lectures, a number of learning objects that vary
between textual-based and visual-based, discussion
forums, examples, exercises, quizzes with
immediate feedback and additional reading
materials.
The authors determined patterns to identify
learning styles in open learning environments based
on Felder and Silverman (Felder and Silverman,
1988) and others (Ahmad et al., 2013; Cha et al.,
2006; Graf et al., 2008; Atman et al., 2009; Graf and
Viola, 2009). These patterns consider the previously
listed learning objects. In addition, knowledge maps
have been considered as a learning object for
AFrameworktoPersonaliseOpenLearningEnvironmentsbyAdaptingtoLearningStyles
299
organising learning concepts to support learners in
open learning environments (Fasihuddin et al.,
2013b; Fasihuddin et al., 2013a). Descriptions of the
determined patterns of behaviours for each
dimension of FSLSM are given below.
First, in terms of learner perceptions, sensory
learners prefer facts, data and experimentation (i.e.
concrete materials) while intuitive learners prefer
principles and theories (i.e. abstract materials), so
annotating the learning objects to specify their types
(i.e. concrete or abstract) and examining the
learners’ access to these objects and the time spent
on them can be used as a pattern. In addition,
sensory learners like to solve problems by standard
methods and do not like surprises, while intuitive
learners like to invent new ways to solve problems.
Based on this, sensory learners are expected to
access more examples and spend more time on them,
while intuitive learners spend more time on the
learning materials. These can be considered among
other patterns to distinguish between sensory and
intuitive learners. Sensory learners are patient with
details, careful but slow, while intuitive learners tend
to be quick and careless; therefore sensory learners
will spend more time on quizzes while intuitive
learners spend less time. In regard to the input
dimension, visual learners remember what they see
better than what they hear while verbal learners
remember more of what they hear than what they
see. Visual learners learn better with diagrams,
flowcharts and pictures while verbal learners prefer
verbal explanations rather than visual
demonstrations. Therefore, annotating the learning
objects to distinguish whether they are visual or
verbal and examining the access and time spent on
them can reveal patterns.
In regard to the processing dimension, active
learners like to try out and learn by practice while
reflective learners prefer to think and reflect about
what they learn so they learn better by observation.
Based on this, active learners tend to access more
exercises and spend more time on them. In addition,
active learners like to work in groups while
reflective learners prefer to learn alone; therefore,
active learners tend to access the discussion forums
and post more than reflective learners.
Finally, in regard to the understanding
dimension, sequential learners like to learn in a
sequential process and prefer learning materials to
be organised and presented in a steady progression
of complexity and difficulty. Global learners do not
like the linear approach and might jump directly to
the more complex materials. Based on this, the
behaviour of accessing the learning materials can be
considered as a pattern. In addition, global learners
like to be provided with the overall picture of the
Figure 2: Pattern calculation method for identifying learning styles in open learning environments.
CSEDU2015-7thInternationalConferenceonComputerSupportedEducation
300
provided topic; therefore, they access and spend
more time on the overview and outline. Moreover,
global learners are expected to access the knowledge
maps of the learning concepts more than sequential
learners, so time spent accessing knowledge maps is
another pattern. The table in Figure 2 summarises all
the above mentioned patterns.
To identify the preferred learning style for each
dimension, the specified patterns of behaviours need
to be monitored in relation to pre-determined
threshold values (Graf, 2007). For instance, if the
expected time to spend on a certain example is 5
minutes, the time that a learner spends is recorded
and then a ratio is calculated and compared to the
pre-determined threshold values to give a
hint(h
,
) for the corresponding dimension. The
hint value is determined based on the ratio. If the
ratio shows a strong preference for the
corresponding dimension, then the hint value is 3. If
the ratio lies between the thresholds then the hint
value is 2. Finally, if the ratio shows a weak
preference, then 1 is marked for the hint value. After
that, the individual’s learning style for the
corresponding dimension is calculated by finding the
mean value of the available hints. The resulting
value, which will be between 1 and 3, indicates the
learning style for the corresponding dimension. This
calculation is computed for each of the four
dimensions of FSLSM. The calculation method to
determine learning styles is summarised in Figure 2.
In order to maintain any possible changes in
learners’ preferences, a dynamic adaptive approach
should be considered in the framework design. This
has been maintained by allowing learning styles to
be re-calculated and updated in the learner’s profile
after each completed module in a provided course.
The calculation of the updated learning style is done
by finding the mean value of the previously stored
learning styles in the learner’s profile. This is an
area of future extension of the study as more
research still needs to be conducted in order to
specify the optimal period of time or number of
previous values that need to be considered in the
calculation process.
4.2 Recommender Agent
After identifying the learning styles and storing them
in the learners’ profiles the recommender agent
provides adaptive navigational support for learners.
Every learner will be presented with learning objects
organised in a way that suits their learning style or
preferences. This organisation is based on the
recommended teaching methods of Felder and
Silverman for each learning style (Felder and
Silverman, 1988). Other recommendations that have
been provided by Graf and Kinshuk (2007) are also
considered. More details about these
recommendations and teaching methods for each
style are provided below.
As mentioned, sensory learners prefer to learn
from concrete materials, so these types of learning
objects need to be shown before abstract materials.
The opposite for intuitive learners – abstract
materials need to be shown to them first. In addition,
sensory learners prefer to learn by examples and
real-life applications, so examples need to be shown
to them before the explanation, while intuitive
learners prefer the reverse. Moreover, sensory
learners prefer more examples and exercises, so all
available examples and exercises need to be
recommended to them, while just some can be
recommended to intuitive learners. In terms of the
input dimensions, textual-based learning objects can
be recommended to verbal-based learners, while the
visual-based objects can be recommended to the
visual learners. In addition, verbal learners may like
to read over additional reading materials so these
can be recommended for them. For the processing
dimension, active learners prefer to learn by doing
thus more exercises will be provided to them. They
also like to invent their own approaches to solving
problems, therefore, fewer examples will be shown
to them. The reverse approach needs to be taken for
reflective learners and more examples will be shown
and less exercises. Also, additional reading materials
will be shown to reflective learners. In regards to the
understanding dimension, sequential learners prefer
to learn with a linear approach, so learning objects
involving examples and exercises need to be
organised with a linear increase of complexity and
the course conclusion and knowledge map are to be
shown last. In contrast, the knowledge maps need to
be presented first to global learners.
Providing learning objects in the above
described organisation is believed to enhance
learning experiences in open learning environments
and consequently to enhance the learners’
satisfaction and learning outcomes. This model will
be evaluated in future implementation with learners.
5 PROTOTYPE DESIGN AND
DEVELOPMENT
The proposed framework has been developed in a
website termed CALC using ASP.net technology.
AFrameworktoPersonaliseOpenLearningEnvironmentsbyAdaptingtoLearningStyles
301
The website simulates the conditions of open
learning, which is the focus of this study. First,
CALC has the advantage of having a self-regulated
learning approach where learners can learn at their
own pace. Learners have personal profiles to keep
their learning progress, interactions with the learning
objects and their preferences. In addition, CALC
provides self-assessment items with instant feedback
so that learners can evaluate their own progress and
knowledge gain. Furthermore, CALC has the
advantage of media-technology enhanced learning as
it provides learning objects in different formats in
order to suit different preferences and needs.
CALC has been designed to conduct a pilot
study at the University of Newcastle so that the
adaptive framework can be evaluated. A course from
the university has been selected and learning
materials for that course have been developed and
hosted on CALC to be learnt independently (as in
MOOCs). More details about the development of
learning materials are provided below.
5.1 Developing the Learning Materials
The development of learning materials has been
conducted with consideration of the requirements of
this study. Hence, various types of learning objects
have been developed for the selected course -
Systems and Network Administration. The learning
objects for each module of the course include the
module’s overview, lecture slides, recorded videos,
textual explanation documents, additional reading
materials, examples, exercises, concept maps and
quizzes. Each type of these learning objects has been
annotated in CALC in order to be recognised by the
adaptive engine and consequently patterns can be
tracked and learning styles identified. Table 2
provides descriptions of these learning objects and
their annotations in CALC.
5.2 Developing CALC
ASP.net technology has been used to develop CALC
with consideration of different browser
requirements. In CALC, every single learner has an
account in order to allow the framework to track
his/her interactions with the learning objects and
store the resulting hints in his/her profile to calculate
learning styles. Interactions that are tracked in
CALC are based on the listed patterns in Figure 2.
So time spent on learning objects is tracked to be
compared with the expected time that is pre-assigned
and saved in the database to calculate hints. Ajax
technology has been used to implement this
functionality. In addition, access to examples,
exercises and other learning objects that need to be
monitored are also tracked in order to find the total
accessed number of these learning objects in each
module and consequently to calculate hints that lead
to identification of learning styles.
The adaptive framework that has been implemented
in CALC is an automatic adaptive system.
Therefore, it requires learners to use the system first
to be able to collect data about their preferences and
consequently provide the adaptive support. When a
learner accesses the first module,
Table 2: Learning objects provided in CALC.
Learning Object Description Category Annotation
Module overview
Provides an indication of the module contents and the main
objective of learning it
Outline OUT
Lecture slides
Presentation slides that provide the learning content in an
abstracted form
Abstract ABS
Recorded videos Recorded videos of the lecturer's explanation to the lecture slides Visual VIS
Textual explanation
documents
Textual documents that provide extended details about the learning
content
Detailed
Verbal
DET
VER
Additional reading
Additional reading that is collected from different resources to
provide additional information about the learning topic
Reading READ
Examples
Provide more explanation of certain concepts or present some
solved problems
Examples EXP
Exercises
Multiple choice questions that allow learners to evaluate their level
of understanding. Instant feedback is provided with an explanation of
the right answer.
Exercises EXER
Concept maps
A graphical representation of the module’s different concepts that
demonstrates how the concepts are related to each other.
Outline OUT
Quizzes
Multiple choice questions with instant feedback and weighted
results that specify whether a module has been successfully completed.
Quiz QUIZ
CSEDU2015-7thInternationalConferenceonComputerSupportedEducation
302
the main page of that module will be represented in
the standard organisation without any adaptive
support. The standard presentation of a module has
all the forms of learning objects shown (i.e. text,
video and slides) as well as all the available
examples and exercises. Also, in the standard
presentation, the learning objects are organised as
follows: course overview; learning concept materials
in different forms; examples; exercises; and quizzes.
A screen shot of the standard main page is provided
in Figure 3.
For navigational support, hiding and sorting
techniques have been applied in CALC. These
techniques have been found to be efficient and
improve user performance by significantly reducing
navigation difficulty (Brusilovsky, 2003). Sorting in
CALC has been implemented by sorting the
different formats hiding of learning materials based
on learners’ preferences. Learning materials are
sorted from the most preferred to the least preferred
formats. For example, visual learners get the videos
listed before the slides or the textual documents
while verbal learners get the textual documents
listed first. In addition, the order of showing
examples and exercises are also based on the learner
learning style. For instance, sensory learners prefer
to have examples first while intuitive learners prefer
to have exercises first.
Figure 3: The standard main page in CALC.
In terms of, the least preferred format of learning
materials are hidden from the list with the possibility
to access them if required. If it is chosen to access
the hidden materials, the list will still be sorted
based on the learner’s preferences. For instance,
visual learners get the textual documents hidden or
coming last in the case of the learner choosing to
show the hidden materials. Moreover, the hiding
technique is also applied to adapting the presentation
of examples and exercises. In the case of learning
styles that prefer to have less examples or exercises,
just few are shown and the rest are hidden. Again,
the hidden examples or exercises can be shown if the
learner chooses that option. Finally, in the case of
balanced learners, the standard organisation is
shown to them. Some screenshots of CALC and how
the adaptive navigational support is provided are
shown in Figure 4.
6 DISCUSSION
This paper presents a framework for identifying
learning style and adapting navigational support to
learner’s preferred learning styles. The proposed
framework has been implemented using ASP.NET
in a website termed CALC.
The automatic identification of learning styles in
the proposed framework is mainly based on tracking
students’ behaviours and interactions with the
determined patterns in the learning environments.
The ILS questionnaire (Soloman and Felder, nd) is
the standard approach for determining learning
styles. One of the ways to benchmark the accuracy
of the automatic learning style identification process
is to benchmark the learning styles identified using
the proposed automated method with ILS responses
of the learner for a cohort of learners who use
CALC.
In addition, the learners’ satisfaction with regard
to personalisation based on organisation and
navigational support of learning materials can be
measured by surveying the learners about their
satisfaction level. Quantitative and qualitative
analysis of survey responses as well as considering
behaviour of learners (such as time spent on learning
object, etc.) and analysis of learner’s performance in
assessments can provide an accurate evaluation of
the personalisation based on learning styles.
In future, the authors intend to deploy the
proposed framework and evaluate both learning
style identification as well as personalisation
impacts on learners.
All the
different
formats are
provided to
learners and
also all the
available
examples &
exercises
AFrameworktoPersonaliseOpenLearningEnvironmentsbyAdaptingtoLearningStyles
303
Figure 4: Some examples of the adaptive navigational support in CALC.
7 CONCLUSIONS
This paper introduces a framework to personalise
open learning environments based on the theory of
learning styles and particularly the Felder and
Silverman Learning Style Model (FSLSM). A
detailed description of the framework and its
components along with the underlying
functionalities is provided. The framework provides
adaptive navigational support through sorting and
hiding the learning materials based on learners’
learning styles and the involved preferences.
A prototype that simulates an open learning
environment in terms of offering open online
courses has been developed and the proposed
framework has been incorporated. In addition,
learning materials have been developed in such a
way that they fulfil the requirements of testing and
evaluating the efficiency of the framework. Future
work of this study involves piloting the developed
prototype with a cohort of learners in order to
evaluate the precision of identifying learning styles
as well as the learners’ satisfaction about the
provided adaptability and navigational support.
REFERENCES
Ahmad, N., Tasir, Z., Kasim, J. & Sahat, H. 2013.
Automatic detection of learning styles in learning
management systems by using literature-based
method. Procedia-Social and Behavioral Sciences,
103, 181-189.
Atman, N., Inceoğlu, M. M. & Aslan, B. G. 2009.
Learning styles diagnosis based on learner behaviors
in web based learning. Computational Science and Its
Applications–ICCSA 2009. Springer.
Bajraktarevic, N., Hall, W. & Fullick, P. 2003.
Incorporating learning styles in hypermedia
environment: Empirical evaluation. Workshop on
Adaptive Hypermedia and Adaptive Web-Based
Systems. Nottingham, UK, pp. 41-52.
Brusilovsky, P. 1996. Methods and techniques of adaptive
hypermedia. User modeling and user-adapted
interaction, 6, 87-129.
Brusilovsky, P. 2001. Adaptive hypermedia. User
modeling and user-adapted interaction, 11, 87-110.
Brusilovsky, P. 2003. Adaptive navigation support in
educational hypermedia: the role of student knowledge
level and the case for metaadaptation. British
Journal of Educational Technology, 34, 487-497.
Cabada, R. Z., Estrada, M. L. B., Cabada, R. Z. & Garcia,
C. a. R. 2009. A fuzzy-neural network for classifying
Videos come first
when expanding the list
and showing all the
Abstract materials
(slides) are recommended
for intuitive learners
- Exercises listed first for
intuitive learners
- Reflective and intuitive
learner so only few
exercises are shown
All available examples are
shown for reflective learners
CSEDU2015-7thInternationalConferenceonComputerSupportedEducation
304
learning styles in a Web 2.0 and mobile learning
environment. Web Congress, 2009. LA-WEB '09.
Latin American, 9-11 Nov. 2009. 177-182.
Carmona, C., Castillo, G. & Millan, E. 2008. Designing a
dynamic Bayesian network for modeling students'
learning styles. 8th IEEE International Conference on
Advanced Learning Technologies. 2008. 346-350.
Carver, C. A., Jr., Howard, R. A. & Lane, W. D. 1999.
Enhancing student learning through hypermedia
courseware and incorporation of student learning
styles. IEEE Transactions on Education 42, 33-38.
Cha, H., Kim, Y., Park, S., Yoon, T., Jung, Y. & Lee, J.-
H. 2006. Learning styles diagnosis based on user
interface behaviors for the customization of learning
interfaces in an intelligent tutoring system. In: Ikeda,
M., Ashley, K. & Chan, T.-W. (eds.) Intelligent
Tutoring Systems. Springer Berlin Heidelberg.
Chang, Y.-C., Kao, W.-Y., Chu, C.-P. & Chiu, C.-H.
2009. A learning style classification mechanism for e-
learning. Computers & Education, 53, 273-285.
Coffield, F., Moseley, D., Hall, E. & Ecclestone, K. 2004.
Should we be using learning styles?: What research
has to say to practice. Learning & Skills Research
Centre.
Coursera. 2012. Coursera [Online]. Available:
https://www.coursera.org/ [Accessed 25-7-2012].
Edx. 2012. edX [Online]. Available:
http://www.edxonline.org/ [Accessed 26-5-2012].
Fasihuddin, H., Skinner, G. & Athauda, R. 2013a. Insights
into the use of Knowledge Maps in Online Learning
Environments: A Pilot Study. The 1st Int. Conference
on Technical Education, 2013. Bangkok. 27-32.
Fasihuddin, H. A., Skinner, G. D. & Athauda, R. I. 2013.
Boosting the opportunities of open learning (MOOCs)
through learning theories. JoC, 3, 112-117.
Felder, R. M. & Silverman, L. K. 1988. Learning and
teaching styles in engineering education. Engineering
education, 78, 674-681.
García, P., Amandi, A., Schiaffino, S. & Campo, M. 2007.
Evaluating Bayesian networks’ precision for detecting
students’ learning styles. Computers & Education, 49,
794-808.
García, P., Schiaffino, S. & Amandi, A. 2008. An
enhanced Bayesian model to detect students’ learning
styles in Web-based courses. Journal of Computer
Assisted Learning, 24, 305-315.
Gilbert, J. E. & Han, C. Y. 1999. Adapting instruction in
search of ‘a significant difference’. Journal of Network
and Computer applications, 22, 149-160.
Graf, S. 2007. Adaptivity in learning management systems
focussing on learning styles. Ph.D. Thesis, Vienna
University of Technology.
Graf, S. & Kinshuk, K. 2007. Providing adaptive courses
in learning management systems with respect to
learning styles. In: Bastiaens, T. & Carliner, S. (eds.)
World Conference on E-Learning in Corporate,
Government, Healthcare, and Higher Education 2007.
Quebec City, Canada: AACE.
Graf, S. & Kinshuk 2014. Adaptive technologies.
Handbook of Research on Educational
Communications & Technology. Springer New York.
Graf, S., Kinshuk & Tzu-Chien, L. 2008. Identifying
learning styles in learning management systems by
using indications from students' behaviour. 8th IEEE
International Conference on Advanced Learning
Technologies. 482-486.
Graf, S. & Tzu-Chien, L. 2009. Supporting teachers in
identifying students' learning styles in learning
management systems: An automatic student modelling
approach. Journal of Educational Technology &
Society, 12, 3-14.
Graf, S. & Viola, S. 2009. Automatic student modelling
for detecting learning style preferences in learning
management systems. Int. Conference on Cognition
and Exploratory Learning in Digital Age, 172-179.
Hong, H. & Kinshuk, D. 2004. Adaptation to student
learning styles in web based educational systems.
World Conference on Educational Multimedia,
Hypermedia and Telecommunications, 2004. 491-496.
Klašnja-Milićević, A., Vesin, B., Ivanović, M. &
Budimac, Z. 2011. E-Learning personalization based
on hybrid recommendation strategy and learning style
identification. Computers & Education, 56, 885-899.
Kuljis, J. & Liu, F. 2005. A comparison of learning style
theories on the suitability for elearning. Web
Technologies, Applications, and Services, vol. 2005,
pp. 191-197.
Latham, A., Crockett, K. & Mclean, D. 2013. Profiling
student learning styles with multilayer perceptron
neural networks. IEEE International Conference on
Systems, Man, and Cybernetics, 2510-2515.
Özpolat, E. & Akar, G. B. 2009. Automatic detection of
learning styles for an e-learning system. Computers &
Education, 53, 355-367.
Şimşek, Ö., Atman, N., İnceoğlu, M. & Arikan, Y. 2010.
Diagnosis of learning styles based on active/reflective
dimension of Felder and Silverman’s learning style
model in a learning management system. In: Taniar,
D., Gervasi, O., Murgante, B., Pardede, E. &
Apduhan, B. (eds.) Computational Science and its
Applications–ICCSA2010. Springer Berlin Heidelberg.
Soloman, B. A. & Felder, R. M. nd. Index of Learning
Styles Questionnaire [Online]. Available:
http://www.engr.ncsu.edu/learningstyles/ilsweb.html
[Accessed 7/2/2014].
Udacity. 2012. Meet Udacity! [Online]. Available:
http://www.udacity.com/.
Udemy. 2014. Udemy [Online]. Available:
https://www.udemy.com/ [Accessed 22-1-2014].
Williams, J. J. 2013. Improving learning in MOOCs with
Cognitive Science. AIED 2013 Workshops
Proceedings Volume, 2013. 49.
Yuan, L. & Powell, S. 2013. MOOCs and open education:
Implications for higher education. CETIS JISC, 21,
2013.
AFrameworktoPersonaliseOpenLearningEnvironmentsbyAdaptingtoLearningStyles
305
