Computers in Human Behavior 55 (2016) 1185–1193

Contents lists available at ScienceDirect

Computers in Human Behavior

journal homepage: www.elsevier.com/locate/comphumbeh

Integrating learning styles and adaptive e-learning system: Current

developments, problems and opportunities
Huong May Truong
Corvinno Technology Transfer Centre, Eduworks ITN, Kozraktar 12/a, H-1093 Budapest, Hungary

a r t i c l e i n f o a b s t r a c t

Article history: Learning styles which refer to students’ preferred ways to learn can play an important role in adaptive
Received 22 December 2014 e-learning systems. With the knowledge of different styles, the system can offer valuable advice and in-
Revised 6 February 2015
structions to students and teachers to optimise students’ learning process. Moreover, e-leaning system
Accepted 8 February 2015
which allows computerised and statistical algorithms opens the opportunity to overcome drawbacks of
Available online 27 February 2015
the traditional detection method that uses mainly questionnaire. These appealing reasons have led to
Keywords: a growing number of researches looking into the integration of learning styles and adaptive learning
Learning styles system. This paper, by reviewing 51 studies, delves deeply into different parts of the integration pro-
Adaptive learning system cess. It captures a variety of aspects from learning styles theories selection in e-learning environment,
Literature review online learning styles predictors, automatic learning styles classification to numerous learning styles ap-
E-learning plications. The results offer insights into different developments, achievements and open problems in the
IT in education
field. Based on these findings, the paper also provides discussion, recommendations and guidelines for
future researches.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction group the opportunity to cooperate and discuss the topic. Further-
more, there is evidence in previous researches (such as in Kolb,
Different students have different preferred ways to learn. Some Boyatzis, & Mainemelis, 2001 and Plovnick, 1975) which shows the
may understand quickly through images, others may prefer texts connection between learning styles and career choices. Based on
and readings. Some may deal well with theories, others may learn this, recommendations and guidance to support the career path
through experiments and examples. By gaining insights into dif- planning can be developed. With this strong appeal, learning styles
ferent learning styles, it offers means to design and provide inter- have been gaining significant interest from researchers and edu-
ventions that tailored to individual needs. Moreover, several valu- cators. Coffield, Moseley, Hall, and Ecclestone (2004), in their re-
able advice can be provided to a wide range stakeholders. For view, reported over 70 theories that were developed over the past
example, for learners, insights into their own styles will enable 30 years.
them to be more confident in learning and optimize their learn- In a more comprehensive way, learning styles, which according
ing paths (Herod, 2004). For teachers, it will be able to offer valu- to Keefe (1979) can be defined as: “The composite of characteris-
able feedback on how to match suitable instructions and learning tic cognitive, affective, and physiological factors that serve as rela-
materials to different groups of students at the appropriate stage tively stable indicators of how a learner perceives, interacts with,
of the learning process (Stash, 2007). For instance, under Felder– and responds to the learning environment”. The crowed theories
Silverman’s theory (Felder & Silverman, 1988), learning styles can area can be divided into five groups depending on their assump-
be differentiated between the way students process information: tions on how flexible the learning styles can change over the life-
active experimentation or reflective observations. For “active” stu- time (Coffield et al., 2004). On one extreme, there are theories such
dents, they do not perform very well in a standard classroom sit- as Dunn and Dunn’s models and instruments of learning styles
uation. Conversely, they learn effectively through interaction with (cited in Dunn & Griggs, 2003), which suggests that learning styles
other students. Thus, it is advisable for teachers to provide such are constitutionally based, that is fixed. On the other extreme, re-
searchers consider tactics rather than learning styles. Learning tac-
tics assume that learning behaviours can change depending on

E-mail address: [email protected]

http://dx.doi.org/10.1016/j.chb.2015.02.014
0747-5632/© 2015 Elsevier Ltd. All rights reserved.
1186 H.M. Truong / Computers in Human Behavior 55 (2016) 1185–1193

the situation. This branch includes theories such as Entwistle’s ap- ered. The search process terminated in November 2014. Compre-
proaches and study skills inventory for students (Entwistle, 1997), hensive searches were carried out using a variety of search terms
and Vermunt’s inventory of learning styles (Vermunt, 1998). Other and their combinations including: “learning styles | (or) style”,
prominent theories include Kolb’s (Kolb, Osland, & Rubin, 1995) “measurement | classification | prediction | evaluation | modelling
and Felder–Silverman’s (Felder & Silverman, 1988) which consider | detection | recognition”, “adaptive | personalised | individualised
learning styles as rather stable indicators but may change over the | personalisation”, “integration | application | using”, “automatic”,
lifetime. “learning system | learning management system”, “intelligent tu-
Traditionally, learning styles are mainly measured using surveys toring system”, “student | user modelling”, “online |e- learning”,
and questionnaires, asking students to self-evaluate their own be- “computer-assisted learning”, “adaptive instructions”, “adaptive hy-
haviours. This development is suitable with the traditional class- permedia”, “artificial intelligent”, “education data mining”.
room where it is difficult to observe and analyse students’ pref- The scope of this research surrounds the current application
erences over the whole learning process. However, as for every and integration of learning styles theories in adaptive learning sys-
qualitative survey, this type of measurement suffers many draw- tem. Hence, the following inclusion criteria was applied: learn-
backs. Firstly, it can be biased as it depends on students’ judgment. ing styles theory/theories had to be included as part of the de-
Secondly, it is done only at a point in time while the learning sign/structure/development/modelling of the e-learning system;
styles, according to several theories, can change over time. Some there were evidences of implementation (e.g. there was descrip-
of these surveys can reach over 40-question long (such as Ver- tion or demonstration of the actual implementation, or there were
munt’s (Vermunt, 1998) and Felder–Silverman’s (Felder & Silver- evidences of models evaluation/testing). Considering such inclusion
man, 1988)) and hence, they are not easy to update. criteria, the searches resulted in 51 papers among which 39 were
These drawbacks have encouraged a growing number of re- journal papers and 12 were conferences papers. These articles were
searchers to integrate the framework of learning styles into e- then analysed, synthesized, and grouped using similar themes. The
learning system. On one hand, e-learning system, which allows re- results are shown in the next section.
searchers to observe students’ behaviours throughout the learn-
ing process and with the use of data mining and computerised 3. Results: the current developments
algorithms, to quickly identify and analyse trends in big dataset,
opens opportunities to develop new framework to observe and 3.1. Classification of results: learning styles integration process
measure learning styles through online behaviours. On the other
hand, learning styles are also useful sources to develop an adaptive By reviewing previous literature, it was identified that all the
e-learning system that effectively personalises learning resources articles followed a very similar integration and development pro-
to individuals’ learning needs. With strong appeal, the integration cess and thus, the results of the analysis are also presented ac-
of the IT-related and psychology and pedagogy-related area have cording to this course which is shown in Fig. 1. Through this result
gained significant interest over the past years. A recent paper sur- classification approach, it can be helpful for future researchers and
veying e-learning system developers by Thalmann (2014) even sug- e-learning system analyst to quickly gain insights into different re-
gested that learning styles models were the most useful frame- quired parts of the development process.
works for adaptive system development among other sources such The process of integrating learning styles into adaptive learn-
as previous knowledge and student background. ing system can be divided into two main areas: learning styles
The constructs and applications of learning styles into adaptive prediction using online data (or the online learning styles clas-
e-learning system have observed several positive results in both sification model) and the application of this model into adaptive
learning styles detection (such as in García, Amandi, Schiaffino, learning system. The development starts with choosing the learn-
& Campo, 2007; Graf, Kinshuk, & Liu, 2009; Scott, Rodríguez, So- ing styles framework. This is followed by the determinant of data
ria, & Campo, 2014; Özpolat & Akar, 2009) and applications (such sources and learning styles attributes and classification algorithm
as personalising learning materials and learning contents as in selections. After evaluating, the suitable classification models and
Kurilovas, Kubilinskiene, & Dagiene, 2014 and developing educa-
tional games as in Lin, Yeh, Hung, & Chang, 2013). This paper pro-
vides an update and a systematic review on this integration of
learning styles into adaptive e-learning system. Through the liter-
ature review, it offers insights into different methodologies, con- Data sources
structs, developments and applications that have been studied in
the research field. Moreover, broadening from previous reviews
such as papers by Vandewaetere, Desmet, and Clarebout (2011)
Learning styles
and Akbulut and Cardak (2012) which either only briefly looked Learning styles
at learning styles as part of many other personalisation character- theory
attributes selection
istics or only presented parts of the whole applications, this paper, selection
focusing on learning styles, delves deeply into the integration pro-
cess. It captures several aspects from learning styles measurement
in e-learning system, to the usage of learning styles in e-learning
application. Based on the findings, it also provides new insights Classification algorithm
into current developments, issues and challenges and as the result, Using Learning development and
valuable recommendations can be offered for future studies. Styles evaluation
Questionnaire
2. Search methods

Different studies on the integration of learning styles and adap- Applications

tive learning system reviewed in this paper were collected through
three search systems: Google Scholar, Scopus and Science Direct.
Articles for the last 10 years (from to 2004 to 2014) were consid- Fig. 1. Learning styles integration process.
 H.M. Truong / Computers in Human Behavior 55 (2016) 1185–1193 1187

their results are applied for different aspects of the adaptive e- others in online environment. The explanation of a certain the-
learning system. ory choice is not always clear. Only a small number of reviewed
Not all of the papers manage to present the whole process. papers includes some motivation and intuition behind their appli-
In this literature survey, while some articles focus on the on- cations. For examples, Germanakos, Tsianos, Lekkas, Mourlas, and
line learning styles classification model development using differ- Samaras (2008) suggested that theory such as Kolb’s was com-
ent computerised and statistical algorithms, others focus on the plex and strongly correlated with personality theories and thus it
application of learning styles and hence, traditional measurement was not adequate nor easily quantified. Consequently, they pro-
method which is questionnaire was used. This second group of pa- posed to apply Felder–Silverman’s as the theory comprised of dis-
pers is portrayed by the dashed line showed in Fig. 1. tinctive scale corresponding to different aspects of the learning
Table 1 presents the results distribution of the articles reviewed process. Feldman, Monteserin, and Amandi (2014) justified their
in this work. choice of focusing on particularly perception style as it showed
Among 51 papers, majority of the articles, 49%, emphasize the strong connection with other important factors such as career pref-
application of learning styles themselves in the development of erences, aptitudes, and management styles. Dorça, Lima, Fernan-
adaptive learning system. 23.5% of the papers provide insights des, and Lopes (2013b), in addition, argued that Felder–Silverman’s
into detailed online learning styles classification models and 27.5% stood out because it combined different main learning styles
manage to present both aspects and how the two parts of the de- models.
velopment process were interconnected.
In the next sections, each stage of the integration process:
learning styles theories selection, learning style attributes selec- 3.3. Learning styles predictors
tions, learning styles classification algorithms, applications in adap-
tive learning system will be explored and discussed which will While literature is one of the places that researchers turn to
provide insights into the current practice as well as different open first for hints and guidelines for potential measurements and vari-
problems and challenges that require further studies. ables, for the case of online learning styles, several sources of use-
ful data have been identified which can create a challenge in terms
3.2. Learning styles theories of attributes selection. The potential sources of data and the corre-
sponding attributes can be summarized as below:
The first step of the integration process is to select the right
learning styles framework which can be a challenge for researchers • Log files: this source tracks users’ actions and interaction with
and developers. This is the case because firstly, the learning styles the system’s interface. Whilst all articles use some forms of
theories landscape is crowded. In the last 30 years, over 70 theo- log file, the useful attributes used for classification vary widely.
ries were developed (Coffield et al., 2004). Some of them can over- The focuses are mainly on number of visits, time spent, perfor-
lap. For example, Felder–Silverman’s shares some very similar di- mances, characteristics and types of objects chosen, sequences
mensions with Kolb’s and Riding’s models. Secondly, according to of actions and selected search terms. There are also several
Coffield et al. (2004), most of learning styles theories suffer some activities have been tracked including searching, taking online
issues in terms of validity and reliability. Consequently, there is no exam, quiz, puzzles or self-assessment test, playing games, us-
theory that outperforms others. ing forum, mail and discussion board and reading and down-
In adaptive learning system application, a variety of theories loading materials.
have been used. Results of the content analysis according to which • Users’ history and background data: this includes static infor-
learning styles have been applied are presented in Table 2. Refer- mation such as gender, education majors and ethnicity and cul-
ences of articles are also included. ture. They are rarely incorporated in automatic classification,
Felder–Silverman’s model (Felder & Silverman, 1988), which although previous researches have shown that these factors
differentiates learning styles through four dimensions: perception can play an essential role in determining learning styles (Kolb,
(Sensory/Intuitive), information input (Image/Verbal), information 1981; Reid, 1987; Vita, 2001).
process (Active/Reflective) and understanding (Sequential/Global), • Besides those directly associated with Learning styles as above,
is by far the most widely used theory in adaptive learning sys- it is also important to note that there are other personaliza-
tem (accounted for 70.6% of all the papers in the survey). Other tion sources which include background knowledge, intelligent
notable theories include VARK which stands for Visual, Auditory, capability, cognitive traits (working memory capacity, process-
Read, Kinesthetic and Kolb’s Learning styles inventory (Kolb et al., ing speed, learning stills, reasoning ability), study goals, lan-
1995) and Honey and Mumford’s Learning styles (Honey & Mum- guage and motivation level which in some case, were consid-
ford, 1986) which both divide styles based on their proposed learn- ered alongside with learning styles (such as Germanakos et al.,
ing cycles. 2008).
It is interesting to note that all of the above learnings styles
belong to the same group of theories which proposes that learn- The identification of these sources show a dynamic picture of
ing styles are rather stable indicators but may change over time potential attributes and behaviours that can be taken into account.
(according to Coffield et al., 2004). Or in other words, this group With similar learning styles framework, variables used in previ-
which assumes that learning styles is neither fixed nor changing ous studies can be varied. For instance, with both studies used
very rapidly for every situation is appeared to be favoured over Felder–Silverman’s, García et al. (2007) employed variables related

Table 1
Overview of results classifications.

Topic Number of articles %

Online learning styles classification models development 12 23.5

Application of learning styles in adaptive learning system development 25 49.0
Both 14 27.5
1188 H.M. Truong / Computers in Human Behavior 55 (2016) 1185–1193

Table 2
Learning styles theories applied in adaptive learning system.

Theories used Number of % References

articles

Felder–Silverman 36 70.6 Akbulut and Cardak (2012), Alkhuraiji, Cheetham, and Bamasak (2011), Baldiris et al. (2008), Cabada et al. (2009), Cabada,
Barrón Estrada, and Reyes García (2011), Carmona, Castillo, and Millán (2008), Cha et al. (2006), Dorça et al. (2013a),Dorça
et al. (2013b), Dwivedi and Bharadwaj (2013), Essalmi, Ayed, Jemni, Kinshuk, and Graf (2010), Feldman et al. (2014),
Franzoni, Assar, Defude, and Rojas (2008), García, Schiaffino, & Amandi (2008), García et al. (2007), Germanakos et al.
(2008), Graf, Kinshuk, & Liu (2008), Graf, KinshuN, Maguire, & Shtern (2010), Graf et al. (2009), Hong and Kinshuk (2004),
Jovanovic, Gašević, and Devedžić (2009), Kelly and Tangney (2005), Klašnja-Milićević, Vesin, Ivanović, and Budimac (2011),
Latham, Crockett, McLean, & Edmonds (2012), Latham et al. (2014), Limongelli et al. (2011), Lin et al. (2013), Sancho,
Martínez, and Fernández-Manjón (2005), Sanders and Bergasa-Suso (2010), Sangineto, Capuano, Gaeta, and Micarelli
(2008), Schiaffino, Garcia, and Amandi (2008), Scott et al. (2014), Sevarac, Devedzic, and Jovanovic (2012), Villaverde,
Godoy, and Amandi (2006), Wen et al. (2007), Yang, Hwang, and Yang (2013), Özpolat and Akar (2009)
Honey & Mumford 2 3.9 Del Corso, Ovcin, and Morrone (2005), Kurilovas et al. (2014)
Kolb 2 3.9 Botsios, Georgiou, and Safouris (2008), Moura, Franco, Melo, and Fernandes (2013)
VARK 5 9.8 Peter, Bacon, and Dastbaz (2010), Wang, Wang, and Huang (2008), Yasir and Sami (2011), Özyurt, Özyurt, & Baki (2013a),
Özyurt, Özyurt, Baki, & Güven (2013b)
Others 6 11.8 Essaid El Bachari and El Adnani (2011), Popescu (2010), Siadaty and Taghiyareh (2007), Sun, Joy, and Griffiths (2007),
Tseng, Chu, Hwang, and Tsai (2008), Zakrzewska (2012)

to forums, chats, exam revision etc., while Wen, Graf, Lan, An- The second popular method is Bayesian network which is based
derson, and Kinshuk (2007) detected styles using attributes re- on Bayes theorem:
late to assessment such as time spent for certain type of ques- P (B/A )P (A )
tions, performance on the test, and time taken to check the P (A/B ) =
P (B )
question.
Nevertheless, even though there are several predictors that have The theorem can be interpreted as the probability of A given B
been taken into account, none of the papers found in this review is equal to the probability to B given A times the probability of A,
manages to compare the power of different attributes in predicting divided by the probability of B. When apply to learning styles clas-
learning styles. The finding of such comparisons can play an impor- sification, the researchers use the equation to estimate the proba-
tant role in improving the performance and efficiency of different bility a user belongs to a certain style given their actions. Over the
prediction and classification models. review, 7 studies applied this method, while 3 other studies used
Naïve Bayes which is also a special case of Bayesian Network. This
3.4. Learning styles classification algorithms algorithm has also showed promising results from many studies.
One example can be study by García et al. (2007) which showed
Table 3 provides an overview of different classification algo- the accuracy levels compared to questionnaire ranging from 58%
rithms that have been applied in adaptive learning system. In a to 77% for three dimensions of Felder–Silverman’s learning styles
number of articles, more than 1 method was used. theory.
One of the most popular methods is rules-based (applied by 8 Among 51 studies, almost all of the reviewed papers used single
papers) in which researchers “translated” different styles accord- algorithms. While advanced algorithms combining different sin-
ing to the theories into different statistical rules. For example, Graf gle algorithms (or hybrid or ensemble classification) have showed
et al. (2009) used the description of learning styles from Felder good results in a variety of applications such as medical and fi-
and Silverman’s to get “hints” such as if a student visited exercise nance (Bhattacharyya, Jha, Tharakunnel, & Westland, 2011; Wang,
more often, he/she is more likely to prefer active learning style. Hao, Ma, & Jiang, 2011), only 2 studies in the review explore this
Then, they set threshold for each attributes to divide behaviours option. Cha et al. (2006) and Özpolat and Akar (2009) both com-
into “strong indication”, “average” or “in disagreement” with a cer- bined the application of Naïve Bayes and Decision Tree (or so-
tain learning style. The information from every attribute of the called NBTree) to automatically predict learning styles. In addi-
user then was averaged and normalized to the scale from 0 to 1, tion, among the papers reviewed, only 2 papers provide compar-
which in turn, was used to identify learning styles using pre-set ison between the performances of different methods. Besides ap-
threshold. By testing on 127 students, the proposed methods did plying NBTree, Cha et al. (2006) also employed Hidden Markov
show reasonable precision compared to results from the question- Chain on the sequences of students’ actions. The study showed that
naire. while NBTree was better in classifying Visual and Auditory, Hidden

Table 3
Learning styles classification algorithms.

Classification method used Number of References

articles

Bayesian network 7 Alkhuraiji et al. (2011), Botsios et al. (2008), Carmona et al. (2008), Essaid El Bachari and El Adnani (2011), García,
Schiaffino et al. (2008), García et al. (2007), Schiaffino et al. (2008)
Naïve Bayes 3 Feldman et al. (2014), Kelly and Tangney (2005), Zakrzewska (2012)
Rules 8 Dorça et al. (2013b), Graf, Kinshuk, & Liu (2008), Graf, KinshuN, Maguire, & Shtern (2010), Graf et al. (2009), Latham,
Crockett, McLean, and Edmonds (2012), Sangineto et al. (2008), Scott et al. (2014), Wen et al. (2007)
Neural network 3 Cabada, Barrón Estrada, & Reyes García (2011), Cabada et al. (2009), Villaverde et al. (2006)
Decision tree 2 Cha et al. (2006), Özpolat and Akar (2009)
Others 5 Cha et al. (2006), Dorça et al. (2013a), Hong and Kinshuk (2004), Sanders and Bergasa-Suso (2010), Wang et al. (2008),
Özpolat and Akar (2009)
 H.M. Truong / Computers in Human Behavior 55 (2016) 1185–1193 1189

Table 4
Overview of classification methods combination and comparisons.

Number of References
articles

Single methods 23 Alkhuraiji et al. (2011), Botsios et al. (2008), Carmona et al. (2008), Essaid El Bachari and El Adnani (2011), García et al.
(2007), García, Schiaffino, and Amandi (2008), Schiaffino et al. (2008), Feldman et al. (2014), Kelly and Tangney (2005),
Zakrzewska (2012), Graf, Kinshuk, and Liu (2008), Graf et al. (2009, Graf, KinshuN, Maguire, and Shtern (2010), Latham
et al. (2012), Sangineto et al. (2008), Scott et al. (2014), Wen et al. (2007), Cabada et al. (2009), Cabada, Barrón Estrada,
and Reyes García (2011), Villaverde et al. (2006), Hong and Kinshuk (2004), Sanders and Bergasa-Suso (2010), Wang et al.
(2008)
Combination of methods 2 Cha et al. (2006), Özpolat and Akar (2009)
Compared methods 2 Cha et al. (2006), Dorça et al. (2013a)

Markov Chain provided better performance in distinguishing be- popular as survey, a variety of statistical evaluation tests have
tween Sequential student who studies in a very steady progression been found including evaluations based on control and experiment
and Global student who prefers to jump from sections to sections groups’ pre and post-test performance, time spent on the task,
based on their interests. Dorça, Lima, Fernandes, and Lopes (2013a) level of completeness, engagement and cognitive loads level.
compared Genetic algorithm and Markov chain using simulation
data in which Genetic algorithm was shown to have a better per- 4. Discussion: open problems and opportunities
formance. Table 4 below shows an overview and related references
of methods combination and comparison. The integration of learning styles and adaptive learning system
still requires further researches and experiments. The results found
3.5. Adaptive learning system using learning styles applications through the literature review point out not only several interesting
developments but also opportunities and challenges.
The target for adaptation of learning styles is shown in Table 5. The findings of Section 3.2 reveal that theories considering
It is also worth to note that one paper may also have more than learning styles as stable indicators but may change over the life-
one target. time are preferred over other groups of theories. In addition,
Adaptive learning contents and learning resources are still the Felder–Silverman’s is by far the most popular theory applied in
targets that receive the most interest with 25 papers that have adaptive learning system. Thus, as a significant proportion of re-
touched the topic. Besides choosing which resources would fit searches has applied this theory, it will allow opportunity to com-
which individual, learning styles have also been applied to a pare and evaluate the findings. This direction of classifiers com-
wide variety of applications such as developing adaptive teach- parison is highly encouraged as through this review, there is not
ing strategies, intelligent tutoring and recommendation systems many papers of such type that can be found. Moreover, another
where classes are organised and teaching language, hints and existing problem is that adaptive systems using other learning the-
guidelines are provided in a way that fit to different individuals. ories are still underexplored, hence it is recommended for future
Latham, Crockett, and McLean (2014), for example, used learning research to provide models for quantifying and integrating other
styles to develop a personalised conversational tutorial. Limongelli, theories into adaptive learning systems. The performances of such
Sciarrone, Temperini, and Vaste (2011), on the other hand, un- systems compared to ones using Felder–Silverman’s can be further
der lecomp5 framework, provided a planner in which not only re- analysed based on their impact on students’ learning outcomes.
sources, learning sequence but also courses of action recommen- Nevertheless, whichever theories chosen by scholars, full recogni-
dation were matched to students’ styles. tion of strengths and limitation should be provided. Finally, there
Other applications found through the review contain: format is an opportunity for combining learning styles theories as they
of learning materials, media, assessment and practice, grouping of may support and add values to each other, which may promise
students and adaptive web-learning environment. Notable exam- higher adaptability and recommendation ability. This is also the
ples include papers by Wen et al. (2007) which employed learning view shared by authors such as Ocepek, Bosnić, Nančovska Šerbec,
styles to reduce the bias in peer assessment, Baldiris et al. (2008) and Rugelj (2013).
in which different media, format and semantic density of learning The identification of different potential learning styles predic-
resources were matched to different learning styles and Sanders tors as shown in Section 3.3 displays a very complex picture of
and Bergasa-Suso (2010) in which under the course to investigate the interconnection between a certain styles and their actual be-
and collaborate on the internet, suitable webpages were recom- haviours. While the theory might be the same, the measurements
mended. may vary between studies. There are three main sources of vari-
One of the new applications in recent years is the adaptive ed- ables that have been identified: log files, static information and
ucational games or gamification. While Feldman et al. (2014) pro- other personalisation sources. As none of the papers reviewed so
vided experiment evidences that learning styles can be measured far have been able to evaluate all these features, it points out an
through students’ behaviours when they were playing games, Lin open question regarding different online attributes’ power and per-
et al. (2013) used decision tree to predict learning path of students’ formances in predicting learning styles and as the result there is
with different learning styles which in turn, were used for set- demand for future studies to tackle the issue. The findings of such
ting educational games. The result of such adaptation system was research can not only serve as guideline for adaptive learning sys-
showed to improve the chance that the student had above-average tem developers but also contribute to improve the performance
creativity score. and efficiency of classification and prediction models.
While not all studies provide evaluation and testing on the Another important part of learning styles classification mod-
system, those who did show very promising initial results. The elling is the selection and evaluation of classification algorithms.
overview of their evaluation results are shown in Table 6. Previous researches have managed to examine a wide range of al-
The most popular evaluation method is satisfaction survey gorithms for automatic online learning styles classification ranging
which have been carried out by almost all of the papers as from questionnaire, rules based, Bayesian Network, decision trees
the main or complementary evaluation methods. Although not as to neural networks. Each of these methods has its own strengths
1190 H.M. Truong / Computers in Human Behavior 55 (2016) 1185–1193

Table 5
Learning styles application in developing adaptive learning system.

Adaptive target Number of References

articles

Learning contents, learning resources 25 Alkhuraiji et al. (2011), Baldiris et al. (2008), Cabada et al. (2009), Cabada et al. (2011), Del Corso et al. (2005),
Dwivedi and Bharadwaj (2013), Germanakos et al. (2008), Jovanovic et al. (2009), Kelly and Tangney (2005),
Klašnja-Milićević et al. (2011), Kurilovas et al. (2014), Limongelli et al. (2011), Moura et al. (2013), Popescu
(2010), Sancho et al. (2005), Sangineto et al. (2008), Sevarac et al. (2012), Siadaty and Taghiyareh (2007),
Özyurt, Özyurt, and Baki (2013a), Özyurt et al. (2013b), Sterbini and Temperini (2009), Sun et al. (2007), Tseng
et al. (2008), Yang et al. (2013), and Yasir and Sami (2011)
Learning resources format and media 4 Baldiris et al. (2008), Franzoni et al. (2008), Kelly and Tangney (2005), Yasir and Sami (2011)
Teaching strategies and 8 Cabada et al. (2009), Essaid El Bachari and El Adnani (2011), Franzoni et al. (2008), Latham et al. (2014), Latham
intelligent/recommendation system et al. (2012), Limongelli et al. (2011), Schiaffino et al. (2008), Wang et al. (2008)
Educational games 1 Lin et al. (2013)
Assessment and practice 3 Baldiris et al. (2008), Cabada et al. (2011), Wen et al. (2007)
Others 2 Sanders and Bergasa-Suso (2010), Zakrzewska (2010)

Table 6
Evaluation results of adaptive learning system using learning styles.

Reference Result

Cabada et al. (2011) Satisfaction survey showed that most of the group “agree” or “strongly agree” with respect to the usability of the
system and easiness and time to create an intelligent tutoring system
Jovanovic et al. (2009) The interview with students in the course indicated that students founded the tool useful and handy
Klašnja-Milićević et al. (2011) Leaners in experiment group were revealed to complete the lesson in less time than the control group. They also
completed more lessons than those in control group. >70% of the learners found the system convenient and >60%
of them found the system adaptive, accurate and delivery at satisfied speed
Limongelli et al. (2011) The survey evaluated teachers’ opinions showed that >70% of them found it useful or very useful. Less time for
teachers to analyse learning objects in the tool were recorded
Özyurt et al. (2013a) The evaluation survey presented that most of the students “agree” or “strongly agree” with different statements
relate the usefulness and helpfulness of the system such as helped to understand subject better and facilitate
learning. The interview with 26 students supported further the result from the surveys
Sangineto et al. (2008) Through the analysis of students’ performance before and after using the system, it showed that students who
have accessed the system have made a much sharper progress
Sevarac et al. (2012) The evaluation pointed out that the system could be useful for students and teachers. 80% of the students found it
easy to use after some help introduction in the beginning and 79% of the teachers had positive opinion about the
recommendation system
Siadaty and Taghiyareh (2007) Students who experienced a match between their styles and pedagogy strategy produced higher results than those
who did not have this matched pedagogy strategy
Tseng et al. (2008) Statistical tests on pre and post-tests as well as time spent on searching for materials showed that students who
used adaptive system performed better than those who did not and it could also increase the learning efficiency by
reducing searching time
Yang et al. (2013) The system was showed to require lower mental effort (cognitive loads) for experiment group than control group.
In addition, the finding also revealed that the system was able to engage students in the learning process
Yasir and Sami (2011) Experiment results showed that students who had matched materials perform statistically better than those in the
control group. The survey also presented positive feedback from the students
Essaid El Bachari and El Adnani (2011) Students who used the adaptive system achieved higher significant post-test compared to those who did not.
Latham et al. (2012) Students who used the Oscar system were shown to have an increase in average post-test score. The qualitative
survey added that >80% of the students rated the system highly. Some even suggested that they would use it as
their main method of studying
Schiaffino et al. (2008) 70% of the student found the recommendation provided by eTeacher was useful
Wang et al. (2008) Both learners and teachers had positive feedback on the system efficiency. They suggested that the system could be
apply in the learning process, although there were still a number of concerns existed
Latham et al. (2014) Students who used adaptive system had higher average post score as well as higher test score improvement

and weaknesses. For instance, a number of systems identify users’ it requires not only the values of different probabilities compo-
styles by asking them to carry out a learning styles survey. Al- nents but also the structure of the network which are both not
though such method stays close with the theories and their pro- always easy to obtain. Previous papers overcome this issue through
posed measurements, the problem with this classification is that authors’ interpretation, assumptions and in some cases, for exam-
there is very little or no update throughout the course. Further- ple, see García et al. (2007) estimation has been done on a rel-
more, as several theories suggest that learning styles may change atively small sample of 50 users. As for rules-based method, as-
over time, the fixation in the method makes the classification inac- sumption and interpretation on the theories have to be made to
curate and as a result, reduce its adaptability ability. Moreover, as generate thresholds or to associate certain actions to certain styles.
shown in Section 3.4, several studies follow Bayesian network or Such assumptions make current methods difficult to be transferred
rules-based method. Both approaches have achieved very promis- to others context and user groups.
ing results. In addition, they are both interpretable. While the For the remaining studies, single learning algorithm where one
rules-based method allows learning styles to be quickly updated individual algorithm is built on the dataset is the method mostly
over time, Bayesian Network is able to estimate certainties for un- observed. This leaves opportunity for more advanced methods
observable attributes (Jensen, 1996). However, both methods rely which are still less explored such as hybrid and ensemble learn-
heavily on several assumptions and/or authors’ “translation” of the ing, where different algorithms are combined in different ways to
theories into the online world. For the case of Bayesian network, produce a higher accuracy. In more detail, under hybrid learning,
the method is based on Bayes’ theorem. For such equation to work, the first technique will be used to produce initial output which in
 H.M. Truong / Computers in Human Behavior 55 (2016) 1185–1193 1191

turn, will be processed by the second technique to get the final considers classification methods comparison or methods combina-
output (Tsai, Hsu, Lin, & Lin, 2009). On the other hand, ensemble tion. The review also reveals that whilst adaptive learning content
learning uses the idea of developing and combining results of a di- and learning strategies dominates the application targets, in recent
verse set of single models to obtain the final output (Seni & Elder, years, the applications of learning styles are also expanded to other
2010). In the simplest case, each individual model has the same areas such as assessment, educational games and media choices.
equal vote. Then the output, which in this case, is the probability Some systems using learning styles have shown initial positive re-
that the student belongs to a certain style A will be the total pro- sults, however areas such as assessment or educational games are
portion of models that classifies/votes for the student to be belong still in the early stage of development which require further re-
to A. Both methods have several benefits. For example, ensemble search and exploration.
learning helps the model become more flexible in comparison to
single learning algorithm, reduces variance and overcomes the bias
5.2. Recommendations and future research opportunities
of individual models (Paleologo, Elisseeff, & Antonini, 2010) and
improve the accuracy compared to single algorithm (Oza & Rus-
Through the findings and discussion, different recommenda-
sell, 2000), while hybrid learning can improve accuracy by using
tions for future studies opportunities have also been proposed.
initial algorithms to tune and filter the data (Tsai et al., 2009).
Firstly, there is opportunity to explore the integration of different
Finally, the finding from Section 3.5 provides a very promising
learning styles other than Felder–Silverman’s or the combination of
picture of how learning styles can be used in adaptive learning sys-
different learning styles into adaptive e-learning system. Neverthe-
tem. Several applications have been found such as adaptive learn-
less, regardless of which theory is applied, it is recommended that
ing contents, teaching strategies, intelligent tutor system, adaptive
future studies should fully recognise the strengths and limitations
media, assessment and educational games. Many among these ap-
of the learning styles theories selected. Secondly, one of the future
plications for example assessment and educational games are still
research directions can focus on evaluating the power of different
at their early stages which require further exploration and devel-
online attributes in predicting learning styles. The solution for such
opment. Another issue identified is that not all of the reviewed
problem can contribute in improving the efficiency and perfor-
articles include an evaluation of the systems. Those who did as-
mance of classification models. Furthermore, there is opportunity
sert very promising initial results. There are two types of evalua-
to apply other advanced classification methods which combine dif-
tion tests that have been identified that offer insights into the per-
ferent algorithms together such as ensemble and hybrid learning.
formance of the system under different aspects. Firstly, the most
In addition, papers that compare performances of different classifi-
popular evaluation method is satisfaction survey. Although the re-
cation methods are also highly encouraged. Finally, many adaptive
sult of such type of test relies heavily on how questions in the
applications using learning styles are still at the early stage, thus
surveys have been constructed and implemented, all of the sur-
further development and studies are required. Evaluation for such
veys reviewed in this paper produce very encouraging findings,
future studies and existing systems, especially statistical evaluation
which may suggest the positive impact of adaptive systems us-
is highly recommended.
ing learning styles on students and teachers’ satisfaction. Though
not as popular as satisfaction survey, a variety of statistical ex-
periments have been identified, emphasising the positive effect of 5.3. Limitations
adaptive learning systems using learning styles on different fac-
tors such as performance, efficiency (time spent on different ac- While the review offers insights into trends, open problems and
tivities), cognitive loads and engagement. As many papers in the recommendations for future research opportunities, it is still im-
review did not include these two tests, further evaluation and es- portant to point out its limitations. As faced by many literature re-
pecially statistical evaluation, which is still behind compared sat- view researches, there is the possibility of missing out published
isfaction survey, for existing system and in future studies are high papers in the field which may due to a number of sources.
recommended. First, there is a huge amount of search strings as well as their
synonyms that can be related to the topic. For example, while
5. Conclusions there are authors associate the task of “detecting” online learning
styles as learning styles “classification”, others denote it as “predic-
By reviewing 51 studies and carefully examining different parts tion” or “modelling” or “evaluation” or “diagnosis” etc. Or in an-
of the learning styles integration process, this paper offers insights other sample, the application of learning styles can be found in
into current practices, opportunities and different issues existing “intelligent tutoring systems”, “adaptive instruction system”, “ per-
in the area. A variety of aspects from choosing learning styles the- sonalised learning system” or “adaptive system” etc. To minimize
ories, selecting learning styles predictors, and developing classifi- the chance of missing out important research papers, in this re-
cation algorithms to applying learning styles in adaptive learning view, a comprehensive list of key terms and synonyms was, first,
systems development are explored and discussed. brainstormed. This list and research papers were, then, to increase
the coverage, kept being updated during research process through
5.1. Summary of current developments and open problems “snow ball” method that uses hints from other research papers.
The final list of individual keywords is shown in Section 2. This big
The findings reveal a complex picture of the research field with number of individual keywords generated an even bigger number
promising results and widening applications, yet many open prob- of potential keywords combinations. Under a limited time frame,
lems. It is found that Felder–Silverman is by far, the most popu- maximum of 3-keyword combination was considered at a particu-
lar theory that has been applied in e-learning system. While the lar search. Thus, there is possibility of missing out research papers
theory can be similar, there are several variables used for mea- due to missing individual keywords and missing keywords combi-
surement that have been identified. Nevertheless, the results also nations.
point out an issue which is that none of the previous studies pro- In addition, with hundreds of thousands of results generated
vide information on the power of different attributes in classifying by search engine such as Google Scholar, only top articles could
learning styles. In addition, although many classification algorithms be considered and as the result, the quality of the research is
have been explored among which Bayesian Network and Rules- search engine’s efficiency dependent. There are 3 search systems
based are the most widely uses, only a small proportion of papers were taken into account: Google Scholar, Scopus and Science
1192 H.M. Truong / Computers in Human Behavior 55 (2016) 1185–1193

Direct. Consequently, there might also be published papers avail- Germanakos, P., Tsianos, N., Lekkas, Z., Mourlas, C., & Samaras, G. (2008). Capturing
able in other databases which have not been taken into account. essential intrinsic user behaviour values for the design of comprehensive web-
based personalized environments. Computers in Human Behavior, 24(4), 1434–
1451.
Acknowledgement Graf, S., Kinshuk, K. D., & Liu, T.-C. (2008). Identifying learning styles in learning
management systems by using indications from students’ behaviour. In Eighth
IEEE international conference on advanced learning technologies, 2008. ICALT’08
This work was carried out under the framework of Eduworks (pp. 482–486). <http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=4561743>.
Initial Training Network, Marie Skłodowska-Curie actions (MSCA) Graf, S., KinshuN, Q. Z., Maguire, P., & Shtern, V. (2010). An architecture for dynamic
FP7 of the European Commissions. student modelling of learning styles in learning systems and its application for
adaptivity. In Proc. international conference on cognition and exploratory learning
in digital age (pp. 103–110), Citeseer.
References Graf, S., Kinshuk, K. D., & Liu, T.-C. (2009). Supporting teachers in identifying stu-
dents’ learning styles in learning management systems: An automatic student
Akbulut, Y., & Cardak, C. S. (2012). Adaptive educational hypermedia accommodating modelling approach. Educational Technology & Society, 12(4), 3–14.
learning styles: A content analysis of publications from 2000 to 2011. Computers Herod, L. (2004). Learning styles and strategies. <http://www.edu.gov.mb.ca/ael/all/
& Education, 58(2), 835–842. publications/learning_styles_and_strategies_aug_2004.pdf>.
Alkhuraiji, S., Cheetham, B., & Bamasak, O. (2011). Dynamic adaptive mechanism in Honey, P., & Mumford, A. (1986). Using your learning styles. Peter Honey Maiden-
learning management system based on learning styles. In 11th IEEE international head, UK. <http://library.wur.nl/WebQuery/clc/535141>.
conference on advanced learning technologies (ICALT), 2011 (pp. 215–217). <http: Hong, H., & Kinshuk, D. (2004). Adaptation to student learning styles in web based
//ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=5992305>. educational systems. In World conference on educational multimedia, hyperme-
Baldiris, S., Santos, O. C., Barrera, C., Boticario, J. G., Velez, J., & Fabregat, R. (2008). dia and telecommunications (vol. 2004, pp. 491–496). <http://www.editlib.org/
Integration of educational specifications and standards to support adaptive p/12978/>.
learning scenarios in ADAPTAPlan. <http://www.tmrfindia.org/ijcsa/v5i16.pdf>. Jensen, F. V. (1996). An introduction to Bayesian networks (vol. 210). London: UCL
Bhattacharyya, S., Jha, S., Tharakunnel, K., & Westland, J. C. (2011). Data mining for Press. <http://www.prodstyr.ihh.kvl.dk/vp/2009/slides/BayesNet-I-6.pdf>.
credit card fraud: A comparative study. Decision Support Systems, 50(3), 602–613. Jovanovic, J., Gašević, D., & Devedžić, V. (2009). TANGRAM for personalized learning
Botsios, S., Georgiou, D., & Safouris, N. (2008). Contributions to adaptive educational using the semantic web technologies. Journal of Emerging Technologies in Web
hypermedia systems via on-line learning style estimation. Educational Technol- Intelligence, 1(1), 6–21.
ogy & Society, 11(2), 322–339. Keefe, J. W. (1979). Learning style: An overview. Student Learning Styles: Diagnosing
Cabada, R. Z., Barrón Estrada, M. L., & Reyes García, C. A. (2011). EDUCA: A web 2.0 and Prescribing Programs, 1–17.
authoring tool for developing adaptive and intelligent tutoring systems using a Kelly, D., & Tangney, B. (2005). “First Aid for You”: getting to know your learn-
Kohonen network. Expert Systems with Applications, 38(8), 9522–9529. ing style using machine learning. In Fifth IEEE international conference on ad-
Cabada, R., Estrada, M., Sanchez, L., Sandoval, G., Velazquez, J., & Barrientos, J. vanced learning technologies, 2005. ICALT 2005 (pp. 1–3). <http://ieeexplore.ieee.
(2009). Modeling student’s learning styles in web 2.0 learning systems. World org/xpls/abs_all.jsp?arnumber=1508590>.
Journal on Educational Technology, 1(2), 75–88. Klašnja-Milićević, A., Vesin, B., Ivanović, M., & Budimac, Z. (2011). E-Learning per-
Carmona, C., Castillo, G., & Millán, E. (2008). Designing a dynamic bayesian network sonalization based on hybrid recommendation strategy and learning style iden-
for modeling students’ learning styles. In Eighth IEEE international conference on tification. Computers & Education, 56(3), 885–899.
advanced learning technologies, 2008. ICALT’08 (pp. 346–350). <http://ieeexplore. Kolb, D. A. (1981). Learning styles and disciplinary differences. The Modern American
ieee.org/xpls/abs_all.jsp?arnumber=4561705>. College, 232–255.
Cha, H. J., Kim, Y. S., Park, S. H., Yoon, T. B., Jung, Y. M., & Lee, J.-H. (2006). Kolb, D. A., Boyatzis, R. E., & Mainemelis, C. (2001). Experiential learning theory:
Learning styles diagnosis based on user interface behaviors for the customiza- Previous research and new directions. Perspectives on Thinking, Learning, and
tion of learning interfaces in an intelligent tutoring system. In Intelligent tutor- Cognitive Styles, 1, 227–247.
ing systems (pp. 513–524). Springer. <http://link.springer.com/chapter/10.1007/ Kolb, D. A., Osland, J. S., & Rubin, I. M. (1995). Organizational behavior: an ex-
11774303_51>. periential approach. Englewood Cliffs, N.J.: Prentice-Hall. <http://library.wur.nl/
Coffield, F., Moseley, D., Hall, E., & Ecclestone, K. (2004). Learning styles and peda- WebQuery/clc/932813>.
gogy in post-16 learning: A systematic and critical review. London: Learning and Kurilovas, E., Kubilinskiene, S., & Dagiene, V. (2014). Web 3.0–Based personalisation
Skills Research Centre. <http://www.voced.edu.au/content/ngv13692>. of learning objects in virtual learning environments. Computers in Human Be-
Del Corso, D., Ovcin, E., & Morrone, G. (2005). A teacher friendly environment to havior, 30, 654–662.
foster learner-centered customization in the development of interactive educa- Latham, A., Crockett, K., & McLean, D. (2014). An adaptation algorithm for an intel-
tional packages. IEEE Transactions on Education, 48(4), 574–579. ligent natural language tutoring system. Computers & Education, 71, 97–110.
Dorça, F. A., Lima, L. V., Fernandes, M. A., & Lopes, C. R. (2013). Automatic student Latham, A., Crockett, K., McLean, D., & Edmonds, B. (2012). A conversational intelli-
modeling in adaptive educational systems through probabilistic learning style gent tutoring system to automatically predict learning styles. Computers & Edu-
combinations: A qualitative comparison between two innovative stochastic ap- cation, 59(1), 95–109.
proaches. Journal of the Brazilian Computer Society, 19(1), 43–58. Limongelli, C., Sciarrone, F., Temperini, M., & Vaste, G. (2011). The lecomps5 frame-
Dorça, F. A., Lima, L. V., Fernandes, M. A., & Lopes, C. R. (2013). Comparing strate- work for personalized web-based learning: A teacher’s satisfaction perspective.
gies for modeling students learning styles through reinforcement learning in Computers in Human Behavior, 27(4), 1310–1320.
adaptive and intelligent educational systems: An experimental analysis. Expert Lin, C. F., Yeh, Y., Hung, Y. H., & Chang, R. I. (2013). Data mining for providing a per-
Systems with Applications, 40(6), 2092–2101. sonalized learning path in creativity: An application of decision trees. Computers
Dunn, R., & Griggs, S. (2003). Synthesis of the Dunn and Dunn learning styles model & Education, 68, 199–210.
research: who, what, when, where and so what–the Dunn and Dunn learning styles Moura, F. F. de, Franco, L. M., Melo, S. L. de, & Fernandes, M. A. (2013). Development
model and its theoretical cornerstone. New York: St John’s University. of Learning Styles and Multiple Intelligences through Particle Swarm Optimiza-
Dwivedi, P., & Bharadwaj, K. K. (2013). Effective trust-aware E-learning recom- tion. In IEEE international conference on systems, man, and cybernetics (SMC), 2013
mender system based on learning styles and knowledge levels. Educational Tech- (pp. 835–840). <http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6690461>.
nology & Society, 16(4), 201–216. Ocepek, U., Bosnić, Z., Nančovska Šerbec, I., & Rugelj, J. (2013). Exploring the relation
Entwistle, N. (1997). Contrasting perspectives on learning. The Experience of Learning, between learning style models and preferred multimedia types. Computers &
3–22. Education, 69, 343–355.
Essaid El Bachari, E. H. A., & El Adnani, M. (2011). E-Learning personalization Oza, N. C., & Russell, S. (2000). Online ensemble learning. In AAAI/IAAI (p. 1109).
based on dynamic learners’ preference. <http://citeseerx.ist.psu.edu/viewdoc/ <http://www.aaai.org/Papers/AAAI/2000/AAAI00-190.pdf>.
summary?doi=10.1.1.207.5482>. Özpolat, E., & Akar, G. B. (2009). Automatic detection of learning styles for an e-
Essalmi, F., Ayed, L. J. B., Jemni, Kinshuk, M., & Graf, S. (2010). A fully personalization learning system. Computers & Education, 53(2), 355–367.
strategy of E-learning scenarios. Computers in Human Behavior, 26(4), 581–591. Özyurt, Ö., Özyurt, H., & Baki, A. (2013). Design and development of an inno-
Felder, R. M., & Silverman, L. K. (1988). Learning and teaching styles in engineering vative individualized adaptive and intelligent e-learning system for teaching–
education. Engineering Education, 78(7), 674–681. learning of probability unit: Details of UZWEBMAT. Expert Systems with Applica-
Feldman, J., Monteserin, A., & Amandi, A. (2014). Detecting students’ perception tions, 40(8), 2914–2940.
style by using games. Computers & Education, 71, 14–22. Özyurt, Ö., Özyurt, H., Baki, A., & Güven, B. (2013). Integration into mathemat-
Franzoni, A. L., Assar, S., Defude, B., & Rojas, J. (2008). Student learning styles adap- ics classrooms of an adaptive and intelligent individualized e-learning environ-
tation method based on teaching strategies and electronic media. In Eighth IEEE ment: Implementation and evaluation of UZWEBMAT. Computers in Human Be-
international conference on advanced learning technologies, 2008. ICALT’08 (pp. havior, 29(3), 726–738.
778–782). <http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=4561832>. Paleologo, G., Elisseeff, A., & Antonini, G. (2010). Subagging for credit scoring mod-
García, P., Amandi, A., Schiaffino, S., & Campo, M. (2007). Evaluating Bayesian net- els. European Journal of Operational Research, 201(2), 490–499.
works’ precision for detecting students’ learning styles. Computers & Education, Peter, S. E., Bacon, E., & Dastbaz, M. (2010). Adaptable, personalised e-learning in-
49(3), 794–808. corporating learning styles. Campus-Wide Information Systems, 27(2), 91–100.
García, P., Schiaffino, S., & Amandi, A. (2008). An enhanced Bayesian model to de- Plovnick, M. S. (1975). Primary care career choices and medical student learning
tect students’ learning styles in Web-based courses. Journal of Computer Assisted styles. Academic Medicine, 50(9), 849–855.
Learning, 24(4), 305–315.
 H.M. Truong / Computers in Human Behavior 55 (2016) 1185–1193 1193

Popescu, E. (2010). A unified learning style model for technology-enhanced learn- Thalmann, S. (2014). Adaptation criteria for the personalised delivery of learning
ing: what, why and how? International Journal of Distance Education Technologies materials: A multi-stage empirical investigation. Australasian Journal of Edu-
(IJDET), 8(3), 65–81. cational Technology, 30(1) <http://ascilite.org.au/ajet/submission/index.php/AJET/
Reid, J. M. (1987). The learning style preferences of ESL students. TESOL Quarterly, article/view/235>.
21(1), 87–111. Tsai, C.-F., Hsu, Y.-F., Lin, C.-Y., & Lin, W.-Y. (2009). Intrusion detection by machine
Sancho, P., Martínez, I., & Fernández-Manjón, B. (2005). Semantic web technologies learning: A review. Expert Systems with Applications, 36(10), 11994–12000.
applied to e-learning personalization in< e-aula>. Journal of Universal Computer Tseng, J. C., Chu, H.-C., Hwang, G.-J., & Tsai, C.-C. (2008). Development of an adaptive
Science, 11(9), 1470–1481. learning system with two sources of personalization information. Computers &
Sanders, D. A., & Bergasa-Suso, J. (2010). Inferring learning style from the way stu- Education, 51(2), 776–786.
dents interact with a computer user interface and the WWW. IEEE Transactions Vandewaetere, M., Desmet, P., & Clarebout, G. (2011). The contribution of learner
on Education, 53(4), 613–620. characteristics in the development of computer-based adaptive learning envi-
Sangineto, E., Capuano, N., Gaeta, M., & Micarelli, A. (2008). Adaptive course gener- ronments. Computers in Human Behavior, 27(1), 118–130.
ation through learning styles representation. Universal Access in the Information Vermunt, J. D. (1998). The regulation of constructive learning processes. British Jour-
Society, 7(1–2), 1–23. nal of Educational Psychology, 68(2), 149–171.
Schiaffino, S., Garcia, P., & Amandi, A. (2008). ETeacher: providing personalized as- Villaverde, J. E., Godoy, D., & Amandi, A. (2006). Learning styles’ recognition in e-
sistance to e-learning students. Computers & Education, 51(4), 1744–1754. learning environments with feed-forward neural networks. Journal of Computer
Scott, E., Rodríguez, G., Soria, Á., & Campo, M. (2014). Are learning styles useful Assisted Learning, 22(3), 197–206.
indicators to discover how students use Scrum for the first time? Computers in Vita, G. D. (2001). Learning styles, culture and inclusive instruction in the multicul-
Human Behavior, 36, 56–64. tural classroom: A business and management perspective. Innovations in Educa-
Seni, G., & Elder, J. F. (2010). Ensemble methods in data mining: improving accuracy tion and Teaching International, 38(2), 165–174.
through combining predictions. Synthesis Lectures on Data Mining and Knowledge Wang, G., Hao, J., Ma, J., & Jiang, H. (2011). A comparative assessment of ensemble
Discovery, 2(1), 1–126. learning for credit scoring. Expert Systems with Applications, 38(1), 223–230.
Sevarac, Z., Devedzic, V., & Jovanovic, J. (2012). Adaptive neuro-fuzzy pedagogical Wang, T.-I., Wang, K.-T., & Huang, Y.-M. (2008). Using a style-based ant colony sys-
recommender. Expert Systems with Applications, 39(10), 9797–9806. tem for adaptive learning. Expert Systems with Applications, 34(4), 2449–2464.
Siadaty, M., & Taghiyareh, F. (2007). PALS2: Pedagogically adaptive learning system Wen, D., Graf, S., Lan, C. H., Anderson, T., & Kinshuk, K. D. (2007). Supporting web-
based on learning styles. In Seventh IEEE international conference on advanced based learning through adaptive assessment. FormaMente Journal, 2(1–2), 45–79.
learning technologies, 2007. ICALT 2007 (pp. 616–618). <http://ieeexplore.ieee. Yang, T.-C., Hwang, G.-J., & Yang, S. J.-H. (2013). Development of an adaptive learn-
org/xpls/abs_all.jsp?arnumber=4281107>. ing system with multiple perspectives based on students? Learning styles and
Stash, N. (2007). Incorporating cognitive/learning styles in a general-purpose adap- cognitive styles. Educational Technology & Society, 16(4), 185–200.
tive hypermedia system (vol. 68). <http://cepes.uh.cu/∼rey/udg/sii-ii/biblio/ Yasir, E. A. M., & Sami, M. S. (2011). An approach to Adaptive E-learning Hyperme-
nstash_thesis.pdf>. dia System based on Learning Styles (AEHS-LS): Implementation and evaluation.
Sterbini, A., & Temperini, M. (2009). Adaptive construction and delivery of web- International Journal of Library and Information Science, 3(1), 15–28.
based learning paths. In Frontiers in Education Conference, 2009. FIE’09. 39th Zakrzewska, D. (2010). Building group recommendations in e-learning systems.
IEEE (pp. 1–6). IEEE. Retrieved from <http://ieeexplore.ieee.org/xpls/abs_all.jsp? In Agent and multi-agent systems: technologies and applications (pp. 391–400).
arnumber=5350579>. Springer. <http://link.springer.com/chapter/10.1007/978-3-642-13480-7_41>.
Sun, S., Joy, M., & Griffiths, N. (2007). The use of learning objects and learning styles Zakrzewska, D. (2012). Building group recommendations in e-learning systems. In
in a multi-agent education system. Journal of Interactive Learning Research, 18(3), Transactions on computational collective intelligence VII (pp. 144–163), Springer.
381–398. <http://link.springer.com/chapter/10.1007/978-3-642-32066-8_7>.