International Journal of Industrial Ergonomics 43 (2013) 450e461

Contents lists available at SciVerse ScienceDirect

International Journal of Industrial Ergonomics

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

Personality and its effects on learning performance: Design guidelines

for an adaptive e-learning system based on a user model
Jieun Kim a, Ahreum Lee b, Hokyoung Ryu b, *
a
Innovation Design Engineering, School of Design, Royal College of Art, UK
b
Department of Industrial Engineering, College of Engineering, Hanyang University, Seoul 133-791, Republic of Korea

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

Article history: An increasingly widespread interest in developing fully adaptable e-learning systems (e.g., intelligent
Received 6 March 2011 tutoring systems) has led to the development of a wide range of adaptive processes and techniques. In
Received in revised form particular, advances in these systems are based on optimization for each user’s learning style and
25 January 2013
characteristics, to enable a personalized learning experience. Current techniques are aimed at using a
Accepted 15 March 2013
Available online 1 May 2013
learner’s personality traits and its effect on learning preferences to improve both the initial learning
experience and the information retained (e.g., top-down or bottom-up learning organization). This study
empirically tested the relationship between a learner’s personality traits, analyzed the effects of these
Keywords:
User model
traits on learning preferences, and suggested design guidelines for adaptive learning systems. Two
Design guideline controlled experiments were carried out in a computer-based learning session. Our first experiment
Learning styles showed a significant difference in the learning performance of participants who were identified as in-
Personality trait troverts vs. those who were identified as being extroverts, according to the MBTI scale. As the distinction
Learning performance between extroverted personality types vs. introverted personality types showed the strongest correlation
MBTI in terms of different learning styles, we used this criteria in our second experiment to determine whether
Depth-first and breadth-first design design guidelines for appropriate content organization could reinforce the aforementioned correlation
Forward learning
between personality type and learning experience.
Top-down learning strategy
Relevance to industry: The findings from this article provide how one can practically apply personality
traits to the design of e-learning systems. The structure and level of extraversion could be the features to
be examined in this regard.
Ó 2013 Elsevier B.V. All rights reserved.

1. Introduction algorithms have thus been proven to enhance learner satisfaction

The technological landscape of modern e-learning applications
(e.g., adaptive e-learning systems) has advanced due to the avail-
ability of new artificial intelligence (AI) algorithms that allow for
effective and efficient learning experiences (e.g., Vandewaetere
et al., 2011, Papatheocharous et al., 2012). A variety of issues, such
as the customization of learning content in computer-based
learning activities, serve as the driving forces behind the wide
range of adaptive capabilities. Many e-learning applications have
been developed to accommodate a certain level of adaptability to
an individual’s performance based on their usage data, such as how
many times they had visited for a particular learning module or
which learning process patterns were seen. Machine-learning
* Corresponding author. Tel.: þ82 2 2220 0474; fax: þ82 2 2220 4743.
E-mail addresses: jieun.kim@network.rca.ac.uk (J. Kim), ahrmlee@hanyang.ac.kr
(A. Lee), hryu@hanyang.ac.kr, hokyoung.ryu@gmail.com (H. Ryu).
(e.g., Gerjets et al. 2009), and many studies have now turned their
attention to the intrinsic natures of learners (e.g., learning goals,
interests, personality, and knowledge level) in order to achieve the
best learning experiences (e.g., Brusilovsky, 2001, Germanakos
et al., 2008; Vandewaetere, et al., 2011). Pre-emptive algorithms,
as compared to reflective machine-learning algorithms, have been
widely thought to be promising 21st-century e-learning tech-
niques, as they quickly adapt to a student’s learning activities. What
is still unknown, however, is which learner characteristics (i.e., the
learner’s user model) should be collected and how these charac-
teristics should be addressed when designing computer-based
learning systems.
Early studies (e.g., Riding and Rayner, 1999; Piombo et al., 2003)
on learners’ usage models claimed that learners have three onto-
logically distinct features: (i) Personality features, which dictate the
student’s learning attitude; (ii) Overlay features, which denote the
student’s current domain knowledge level; and (iii) Cognitive fea-
tures, which represent the student’s information processing

0169-8141/$ e see front matter Ó 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.ergon.2013.03.001
 J. Kim et al. / International Journal of Industrial Ergonomics 43 (2013) 450e461
characteristics. The last two features have been well studied in
instructional design (Graesser et al., 2007; Graf et al., 2008). It has
been postulated that the effects of personality are negligible, since it
is the weakest organized set of characteristics possessed by an in-
dividual, but primarily because it is thought to be already cemented
in his or her cognitive features. However, extensive studies on per-
sonality effects (e.g., Germanakos et al., 2008; Honey and Mumford,
1986) have indicated that personality does affect the attitudes and
behaviors that determine an individual’s preferred way of learning.
Therefore, a learner’s experience may be significantly altered if the
instruction style of an e-learning system were to match their
learning style as derived from personality features.
Personalization in online education not only facilitates learning
through different strategies to create various learning experiences,
but it also enables computer-based learning systems to include
varied teaching or instructional packages. For example, some
studies (e.g., Carver et al., 1999; Vincent and Ross, 2001; Kinshuk
and Lin, 2004) identified learner’s attitudes, learning goals, in-
terests, and knowledge levels as critical adaptive parameters in
personalizing learning content. These researchers assumed that the
aforementioned items could be used to determine each learner’s
cognitive style (Kogan, 1971; Messick, 1970, 1976). Therefore, it is
necessary to determine a systematic method of determining a
user’s cognitive style in advance using relevant attributes. At the
same time, the issue of usability has been continually investigated
in order to improve e-learning system quality. For example,
Barcellini et al. (2009) empirically demonstrated use of a user
participatory method in the design process of an e-learning system
called ‘Python’. In addition, recent articles have proposed design
criteria and objective evaluation scales dedicated to e-learning
platforms, including research by Hsu et al. (2009) and Oztekin et al.
(2010).
More comprehensively, Brusilovsky (2001) proposed seven at-
tributes for use in user models of adaptive e-learning systems, as
shown in Fig. 1: learners’ backgrounds, knowledge, goals/tasks,
previous learning experience, preferences, interests, and interac-
tion style. This model showed a significant impact on subsequent
user modeling activities for personalizing adaptive e-learning
systems.
However, Jungian-based psychologists have contended that
people’s personality preferences influence the way they may or
may not want to become more actively involved in their learning
activities, as well as whether they take responsibility for self-
direction and discipline (e.g., Felder et al., 2002; Soles and Moller,
2001). Following a similar line of thought, several researchers
(e.g., Gilbert and Han, 1999; Kwok and Jones, 1985; Papanikolaou
et al., 2002; Moallem, 2003;) tried to integrate learning style into
an adaptive application, matching personal learning style with an
appropriate instruction design in order to adapt to that person’s
strengths and preferences; however, these researchers did not
attempt to examine personality effects.
Fig. 1. The user model of an adaptive e-learning system, extended from Brusilovsky
(2001).
451
Therefore, the goal of this study was to examine the inclusion of a
learner’s personality features in a user model. The findings were
then applied to learning materials, which were empirically tested.
This paper is organized as follows. In Section 2, we reviewed the
possible relationships between a learner’s personality and the
learning styles included in the user model of adaptive learning
systems. In Sections 3 and 4, we examined the personality effect in
adaptive e-learning systems. Our first experiment explored the
relationship between different personality traits and their effects on
learning performance. The second experiment determined whether
personality differences can serve as an appropriate criterion for
designing an e-learning system that best suits a learner’s strengths
and preferences, thereby connoting the personality effect in the user
model. Finally, in Section 5, we discuss our empirical findings, as well
as several design guidelines for adaptive learning systems.
2. Personality effect in adaptive e-learning systems
2.1. Personality effects and learning styles
As briefly discussed above, many Jungian-based educational
psychologists (e.g., Bayne, 2004; Corno and Snow, 1986; Keirsey,
1998; Kwok and Jones, 1985; Soles and Moller, 2001) have claimed
that personality influences the way learners may or may not want to
become more actively involved in their learning processes. There
seem to be significant variables for determining learning perfor-
mance; however, they have not yet been fully examined. The
aforementioned researchers argued that personality is closely tied to
preferences for learning materials in that a particular format reflects
a person’s preferences for taking in information and making de-
cisions. Very few adaptive e-learning systems have considered these
features in their user models, because there is no easy way to model
personality effects. The only method thus far is AHA! (Stash et al.,
2004), which specifies the learner’s style as “Activist/Reflector”,
based on a self-rated personality type.
There are many different schemes of personality types, e.g.,
Kersey’s temperament theory (Keirsey, 1998), the Learning Style
Inventory (LSI; Kolb, 1984), the Big Five framework (Costa and
McCrae, 1992), and the MBTI (Myers, 1993). In their extensive
empirical studies, Keirsey (1998) demonstrated that there are
four personality types that are highly relevant to learning style:
the Rational type (NT e the intuitive thinking type focuses on the
strategic intellect), the Idealist type (NF e the intuitive feeling
type focuses on the diplomatic intellect), the Artisan type (SP e
the sensory perception type focuses on the tactical intellect), and
the Guardian type (SJ e the sensory judgment type focuses on the
logical intellect). Personality types would thus intrinsically
reflect the learner’s preferences for taking in information and
making decisions, which may be defined by one’s learning style.
For instance, SJ-type learners would prefer procedural organi-
zation of learning content over declarative organization. Indeed,
the LSI (Kolb et al., 2000) classified personality types according
to practical learning styles: conversing, accommodating,
diverging, and assimilating. While LSI is highly effective for
determining the learning style of each student and is of great use
in the development of appropriate lesson preparation, the four
LSI classifications have not been widely used in e-learning
design. In an empirical sense, it is not easy to determine each
individual’s personality in such a relatively exclusive and
exhaustive manner.
2.2. MBTI learning styles
By comparison, the Myers-Briggs Type Indicator (MBTI) has
been widely used and validated in the education domain (DiTiberio,
452 J. Kim et al. / International Journal of Industrial Ergonomics 43 (2013) 450e461

1998) and has long been considered an important instrument by information through what is happening and by focusing on
educational psychologists (Blaylock and Rees, 1984; Stewart, 2006). observable facts, data, and phenomena (i.e., declarative knowl-
Consequently, questionnaire-based identification processes have edge). Conversely, intuitive-type learners assess information by its
been compiled by many observations and interviews. possibilities, focusing on the big picture and searching for con-
The MBTI questionnaire examines personality traits in four nections, patterns, relationships, and insightful meaning (e.g.,
distinct spectrums: Extraverted (E)/Introverted (I), Sensing (S)/ patterns of information). Hence, in an e-learning situation, sensing
Intuitive (N), Thinking (T)/Feeling (F), and Judging (J)/Perceiving learners may need a structured framework, with vivid guidelines
(P). The MBTI type theory has two basic assumptions. First, that and directions (Soles and Moller, 2001), whereas the intuitive
equal development of all four functions is not desirable (Bayne, learner may prefer abstract content, as they prefer theory to
2004), and the dominant spectrum should be the most devel- experience. E-learning can provide the sensing learner with prac-
oped. Second, that people are not capable of developing concurrent tical work (using a structured framework with specific guidelines
interests in both traits in a spectrum (i.e. both what is happening (S) and directions). Learning styles that arise from both EeI and SeI
and in possibilities (N) at the same time). This tends to lead to one dimensions are thus key to the empirical study in this article.
preference being dominant, and subsequently used more, trusted Although the last two dimensions of the MBTI temperaments
more and being more developed, in a cumulative and spiral pro- (i.e., ThinkingeFeeling and JudgingePerceiving) are not as strong as
cess. Hence, the general picture in normal-type development is of the first two dimensions in terms of personality distinctions, some
each person gradually discovering what they are best at e their studies have indicated that the former (i.e., ThinkingeFeeling)
talents, gift and central motives e and then spending more time might prove useful in determining learning style. The Thinkinge
and energy on what they are best and necessarily less on other Feeling (TeF) dimension refers to how people make decisions.
things. Feeling-type learners tend to decide on subjective values, while
That being said, the outcomes of the MBTI can be easily con- thinking-type learners base their decisions on logic, facts, and ob-
nected to the learning styles of each individual learner in a more jectivity. In e-learning situations, thinking learners may enjoy an e-
straightforward way than outlined by other personality theories learning process more if it follows a traditional instruction pattern,
(Myers and McCaulley, 1985; Myers, 1993). The MBTI reports a i.e., case studies, planned interactive activities, and tests to measure
person’s preferences in terms of e-learning technologies in the four progress (Felder et al., 2002).
spectrums (Table 1). As discussed above, many studies have consistently concluded
The most dominant function of the MBTI types is the Extraverte that learners with different personality types tend to develop
Introvert (EeI) dimension. Extraverts tend to focus on the outer different learning styles. These differences should be addressed in e-
world of information and actions, whilst introverts focus on the learning systems. Knowing the personality type of each learner can
inner world of ideas and emotions. Because of these differences, help to identify their learning preferences and strengths, which
extraverts are more likely to express their ideas freely, and readily can be utilized in instructional designs so as to maximize the
seek feedback from others. They also have a tendency to act first learner’s potential. Identifying learners’ preferences will help the
and then reflect. Introverts, by comparison, will think things designer create customized educational materials tailored to each
through before acting, and are not as expressive of their emotions. individual. However, some researchers have argued that personality
As an example, the face-to-face interaction provided by videocon- traits are not stable (e.g., Heckmann, 2006), and that personality
ferencing technology may be preferred by extraverts, whereas the type is not a legitimate factor in instructional design. Nonetheless,
introverted learner may prefer asynchronous communication, some computational algorithms (e.g., neural networks or hidden
which enables them to take time to reflect on their ideas and think Markov Models) are capable of adapting to the dynamic nature of a
through a reply or feedback prior to communicating with others. user’s personality. Empirical evidences were reported by Moallem
The SensingeIntuitive (SeI) dimension dictates how information (2003).
is gathered, and is thus the dimension most relevant to instruc- Anonymous reviewers of an earlier draft of this article raised the
tional design in e-learning systems. Sensing-type persons collect issue of the validity of the MBTI theory and the questionnaire-based
identification process. The evidence of the validity of the MBTI
Table 1 theory is substantial, although it is related more to preference than
The four MBTI preferences and the basic definition of the preference. to type dynamics (Bayne, 2004). A key goal of the empirical study
Personality Basic definition Possible examples of described in this article was the determination of the learning style
types and the preference technology in use relevant to an individual’s personality traits and how these traits
Extraverted/ Where they prefer Videoconferencing in e-learning dictate the level of preference for which the user model should
introverted to focus their systems may be of interest for the embody the personality feature. The MBTI questionnaire itself
attention extraverted learners, which provides has also been widely researched. Some studies have discussed
face-to-face personal interaction.
the relationships between psychological type and online learners’
Sensing/ The way they prefer Sensing learners may need the
intuitive to take in structured framework of the course performance and success (Felder et al., 2002; Horikoshi, 1998;
information with specific guidelines and directions. Kilmann,1998; Meisgeier et al.,1989; Whittington and Dewar, 2000;
Thinking/ The way they prefer Thinking learners want to see precise, Whitworth, 2005), as well as how to develop different teaching
feeling to make decisions action-oriented cognitive, affective packages for different learning styles (Soles and Moller, 2001).
and psychomotor objectives, thus
they may enjoy an e-learning situation
more if it follows a more traditional 2.3. The present study
course style, i.e. with case studies
and solving logical, planned interactive Our main research question was whether or not a learner’s
activities and tests to measure progress.
personality type affects their success when using an e-learning
Judging/ How they orient Judgers expect an organized routine
perceiving themselves to the and will push for decisions to be system and how to integrate individualized learning methods into
external world made and then carry them out. e-learning systems. We devised the following research questions:
Perceivers, on the other hand, usually
need to gather more information and  Experiment I: Whether the different personality types of
will postpone decisions.
learners (especially, ExtravertedeIntroverted, Sensing e
 J. Kim et al. / International Journal of Industrial Ergonomics 43 (2013) 450e461
Intuitive) would have different effects on learning performance
in adaptive e-learning systems.
 Experiment II: Whether the learner’s personality may influence
personal preferences that can then be used for appropriate
instructional design.
Although prior research has provided a useful foundation for
analyzing the relationship between various personality traits and
selected aspects of academic motivation, such as performance
goals and learning styles, very little research has examined the
relationship between personality traits and computer-based
learning systems design, in particular, adaptive e-learning
environments.
3. Experiment I: the role of personality in e-learning systems
Experiment I was intended to determine the impact of per-
sonality type on learning performance with an adaptive e-learning
system. The four major characteristics (EeI, SeN, TeF and JeP) of
the MBTI were employed to relate with the learning performance in
the experiment.
3.1. Method
3.1.1. Participants
Eighty-five students from Massey University (New Zealand)
took part in this experiment. Upon completion, their names were
entered into three $50 drawings. The students had some degree of
homogeneity in that they were all undergraduates (third year
students) majoring in Computer Science (CS). The eight experi-
mental groups were formed by factorially combining the four
personality types, as shown in Table 2. In addition, on the grounds
of the MBTI assumptions, we only collected the dominant person-
ality type of each participant, and no auxiliary types were used in
the experiment.
3.1.2. Apparatus
We developed an adaptive e-learning system used for teaching
LISP based on ELM-ART (see Fig. 2: Brusilovsky et al., 1996; Weber
and Brusilovsky, 2001). The same adaptive logic and interfaces
found in ELM-ART were used in this experiment. ELM-ART is
considered by some to be obsolete; however, there are two reasons
we chose to use this system. First, in Experiment I, we considered
both declarative knowledge (i.e., factual information or knowl-
edge) and procedural knowledge (e.g., how to use the factual in-
formation to make a pattern of information or knowledge). LISP, an
artificial intelligence (AI) language, and its built-in adaptive logics
in ELM-ART, were suitable for this purpose. We obtained the
permission of the developers to use this platform. Furthermore,
our participants (i.e., third year Computer Science students) had
little to no knowledge of the new AI language. Therefore, we hy-
pothesized that declarative knowledge would be preferred by the
extraverts, and procedural knowledge would be preferred by the
intuitive and thinking learners. Secondly, personality features are
not included in the user model of ELM-ART, and thus differences in
learning performance are easier to interpret.
Table 2
The participants of Experiment I.
Personality Extraversion Intuitive Thinking Judging
(introversion) (sensing) (feeling) (perceiving)
No. of participants 15 (11) 10 (12) 8 (11) 11 (7)
453
3.1.3. Experiment design
The experiment was designed as a one-way between-subjects
study. The independent variable was the dominant personality
function, as based on the four MBTI types. The dependent variables
functioned to assess learning performance and learning experience
with the modified ELM-ART for teaching LISP. These involved: the
time taken (to complete all of the lessons), the number of correct
answers (out of 40 test questions), the number of navigations (all
navigational movements) and the number of repetitions (how
many times learners returned to previous pages visited). This
unique range of the dependent variables was used to determine
which learning style was adopted by the participants. For example,
sensing-type learners might revisit more of ten if ELM-ART does
not provide structured guidance and directions, as they would
prefer less abstract content, learning by seeing connections, and
dealing with theory.
3.1.4. Procedure
All of the participants were provided with instructions
regarding the experiment. These instructions gave information
pertaining to the experimental procedure, the purpose of the study,
and the data protection policy. All of the participants then
completed an MBTI test administered by a licensed psychologist.
They were all seated in a laboratory where the modified ELM-ART
was installed on each computer. They were given sufficient time
to learn all of the material and were not allowed any discussions or
interactions with other participants or teachers. However, to avoid
any random clicks by the participants, we added an exercise section
between learning topics. Each learner needed to correctly answer a
minimum of three answers in a row to continue to the next learning
topic. At the end of the learning session, they were told to answer
40 questions about the learning content (20 declarative knowledge
and 20 procedural knowledge questions).
3.1.5. Data analysis
It should be noted that we originally planned to use a between-
subjects analysis of variance. However, an inspection of the
empirical data showed that the standard deviation was closely
correlated with the mean and Levene’s tests for heterogeneity of
variance were found to be significant, suggesting that the data were
not suitable for analysis of variance. It would have been possible to
transform the data to remove this correlation, but this would make
the mean difficult to interpret and still would not allow an analysis
of some of the data from the experiment. For these reasons, the
simpler solution of using Mann-Whitney U tests was adopted. This
allowed us to compare the two pairing personality traits for each
learning topic.
3.2. Results (Experiment I)
3.2.1. The effect of personality traits
Table 3 summarizes the learning performance comparisons
(time taken, correct answers, the number of navigations and rep-
etitions) by MBTI personal traits. It can be seen that the first
dimension (i.e., EeI) of the MBTI dominated the salient difference
in using ELM-ART. In terms of the four measures used in this
experiment, the extraverted learners all outperformed their coun-
terparts. However, the other personality spectrums (i.e., Sensing e
Intuitive, Thinking e Feeling, and Judging e Perceiving) did not
provide consistent results. These results seemed to confirm the
previous findings that introverts to some extent prefer to concen-
trate on a single activity at a time and prefer to thoroughly un-
Total
derstand a topic before they moving onto the next learning topic
(number of repetitions: 15.68 vs. 20.36). In contrast, the extraverts
85
readily sought out new information and knowledge (time taken:
454 J. Kim et al. / International Journal of Industrial Ergonomics 43 (2013) 450e461

Fig. 2. The web-based e-learning system for teaching LISP used in Experiment I (redesigned based on ELM-ART).

18.05 min vs. 22.28 min). Of course, this does not mean that the
introverts were slow to take in or understand new information or
knowledge, and care is needed to interpret the outcomes shown in
the second column (correct answers: 82.49% vs. 75.77%). This will
be discussed later in this paper.
3.2.2. Personality traits against content to be learnt
The second analysis compared the personality traits against the
content used in the experiment, i.e., declarative knowledge vs.
procedural knowledge. Table 4 represents the critical data sepa-
rated from the total scores. Note that the total scores are the same
means given in Table 3.
Comparing the extraverts (mean ¼ 45.38) and the introverts
(mean ¼ 46.98) with the basic LISP primitives (i.e., declarative
knowledge) reveals an almost identical rate of learning. However,
for the procedural LISP knowledge, the extraverted participants
showed a significantly higher mean (37.11) than the introverted
participants (mean ¼ 28.79). Table 3 shows a significant difference
between extraverted and introverted participants with regard to
the correct answers given. This might arise from differences
learning the procedural LISP knowledge, making this effect even
more striking.
Markedly different patterns were also observed for the other
personality types. As seen in Table 3, there were no differences in
the total scores for the other personality traits. However, the data
given in Table 4 reveal that there were certain distinctions in the
continuum of the two personality preferences (i.e.,
sensing ¼ 27.88% vs. intuitive ¼ 37.51%, and thinking ¼ 24.45% vs.
feeling ¼ 15.80%) in learning the procedural knowledge. In terms of
the procedural LISP knowledge, both the intuitive and thinking

Table 3
Task performance by personality types (mean/s.d.).

Time taken (unit: min) Correct answer (unit: %) No. of navigations No. of repetitions
Extraverted (n ¼ 15) 18.05 (2.44) p < .01 82.49 (5.47) p < .01 53.32 (12.42) p < .01 15.68 (8.95) p < .01
Introverted (n ¼ 11) 22.28 (4.59) 75.77 (8.07) 66.45 (9.62) 20.36 (8.52)
Sensing (n ¼ 10) 19.69 (3.97) n.s. 79.65 (6.40) n.s. 59.48 (13.56) p < .01 19.38 (8.67) p < .05
Intuitive (n ¼ 12) 17.95 (2.35) 83.34 (7.95) 49.90 (8.29) 10.10 (5.95)
Thinking (n ¼ 8) 18.40 (2.26) n.s. 81.51 (4.78) n.s. 56.08 (10.91) n.s. 17.15 (8.50) n.s.
Feeling (n ¼ 11) 20.92 (5.25) 78.78 (9.91) 58.92 (14.91) 16.69 (10.23)
Judging (n ¼ 11) 18.20 (3.27) p < .01 80.83 (5.57) n.s. 54.10 (10.91) p < .05 15.55 (5.55) p < .05
Perceiving (n ¼ 7) 22.27 (3.17) 79.92 (10.23) 65.50 (15.47) 21.20 (4.78)
 J. Kim et al. / International Journal of Industrial Ergonomics 43 (2013) 450e461 455

Table 4 their learning experience in computer-based learning activities.

Correct answers by personality types (mean/s.d). However, neither of these possibilities has been demonstrated
Declarative knowledge Procedural knowledge Total empirically.
(20 questions) (20 questions) score Furthermore, the outcomes of this experiment suggested that
Extraverted 45.38 (3.71) U ¼ 62.00 37.11 (3.04) U ¼ 12.00 82.49 our participants appeared to differ in terms of learning topics that
(n ¼ 15) (n.s.) (p < .01) best fit their learning style. These experimental findings resemble
Introverted 46.98 (2.01) 28.79 (1.24) 75.77
those of several publications on personality theory (e.g., Felder
(n ¼ 11)
Sensing 51.77 (9.09) U ¼ 52.00 27.88 (4.90) U ¼ 8.00 79.65 et al., 2002; Rao, 2002; Shuck and Phillips, 1999; Sloan and Jens,
(n ¼ 10) (n.s.) (p < .01) 1982). Learning about ‘lists’ in LISP requires an understanding of
Intuitive 45.83 (2.06) 37.51 (1.69) 83.34 complex symbolic expressions in order to work with data and
(n ¼ 12) procedures. This suggests that intuitive personality may achieve
Thinking 57.06 (3.56) U ¼ 19.50 24.45 (1.42) U ¼ 4.50 81.51
(n ¼ 8) (p < .05) (p < .01)
better learning outcomes (Soles and Moller, 2001). This pattern is
Feeling 62.98 (7.39) 15.80 (1.85) 78.78 also advantageous for the thinking personality trait. Our results
(n ¼ 11) showed that participants with this trait were better at learning
Judging 53.37 (6.04) U ¼ 37.00 27.46 (3.12) U ¼ 28.00 80.83 procedural knowledge than feeling learners, which has also been
(n ¼ 11) (n.s.) (n.s.)
found in previous research, as the thinking personality trait values
Perceiving 54.35 (7.98) 25.57 (5.14) 79.92
(n ¼ 7) logical processes (Myers et al., 1993; Vincent and Ross, 2001).
However, this goes against our expectation that perceiving learners
would enjoy flexibility and dislike rigid timelines and tightly
personalities showed significant effects. These results are in line controlled interaction steps (Lawrence, 1984, 1997).
with previous findings. Many studies (e.g., Soles and Moller, 2001; It is, of course, not unreasonable to believe that a user may
Felder et al., 2002) have suggested that intuitive-type learners tend employ a learning style that differs from their assessed personality
to assess information by its possibilities, focusing on the big picture trait in real learning situations. Humans are complex and unique,
and searching for connections, patterns, relationships, and and participants may have a mixture of both orientations. For
insightful meaning. Hence, in our experimental context, this might example, extraversion varies along a continuum. A person who acts
enable an intuitive person to do well when learning procedural LISP introverted in one learning situation may thus act extraverted in
knowledge rather than declarative knowledge. In a similar vein, the another. Neither of the behavior differences by personality
thinking-type participants more readily learned and then retained observed in this study should be considered to be invariant.
the procedural knowledge than the declarative knowledge, which However, the data can be taken to suggest that, at the very least,
was not the case for the feeling-type participants. care is needed when designing e-learning systems, and that per-
sonality trait awareness and learning content need to be translated
3.2.3. The effect of previous experience for e-learning practitioners. The following section addresses this
One might question whether the participants’ previous experi- issue further, developing design guidelines for adapting e-learning
ence with other learning systems may change the findings. For systems to learning styles.
instance, someone who regularly used an e-learning system that
behaved like one of our experimental conditions might be expected
4. Experiment II: applying personality to e-learning design
to find such an e-learning system easier to learn or be able to use it
more effectively. The results may therefore reflect user experience
Experiment I demonstrated the need to establish guidelines for
rather than the basic personality traits considered in this study.
incorporating learners’ personality traits into e-learning systems.
We therefore conducted a survey before the first experiment
However, in order to rigorously validate the effects of personality,
which included questions concerning which e-learning systems the
further experiments must be conducted. Experiment II was
participants had used, and how often. The survey examined the
designed to test whether or not a learner’s personality dictates
level of expertise (or experience) that participants had had with
their preferences for a particular style of instructional design. To
any computer-based learning systems. Very few participants were
empirically demonstrate this claim, one would ideally want to
able to report what or how many computer-based learning systems
study personality use in real design situations. There are relatively
they had used; therefore, only their incomplete recall was collected.
few articles in the personality and e-learning literature in which
Of the 85 participants, around 90% of the participants (76) were
real designers take personality traits into account, so little data is
using (or had used) WebCTÔ and/or BlackboardÔ, and most of
available for comparison.
these participants had downloaded teaching slides or learning
A small number of undergraduate computer science students
materials. Only 2% had used game-based learning systems. None
were recruited for Experiment II, and these students were taught by
had previously used adaptive tutoring systems. We cannot say
either an extraversion-friendly or an introversion-friendly e-
what influence this had in the experimental context; however, even
learning system. Although it is difficult to make generalizations
if there was an effect, this would not weaken the practical useful-
from a single evaluation with each system, and though this
ness of our findings. Participants were assigned to our experimental
reductionist approach may differ from a real design context in
conditions at random. Within the limits of sampling error, experi-
certain critical aspects, this exercise can provide some empirical
ence should have been distributed as it is in the sample population.
support for the feasibility of considering personality traits when
designing e-learning systems.
3.3. Conclusions and discussion

The first conclusion that could be drawn from this experiment
was that personality effects (at least extraversion/introversion)
could influence the ease with which a learning activity in an e-
learning system can be performed. In this light, the work of
Felder and Brent (2005) is worth noting. They contended that in-
troverts might need to be treated carefully in order to enhance
4.1. Learning material structure: depth-first vs. breadth-first
We developed a somewhat different experimental setting for
this part of the study, which was chosen in order to compare
learners’ personality traits and corresponding presentation styles.
Two presentation styles, i.e., Depth-first (Stash et al., 2004) and
456 J. Kim et al. / International Journal of Industrial Ergonomics 43 (2013) 450e461
Breadth-first structures (Brown et al., 2005), were considered in this
study.
Previous studies (e.g., Riding and Fanning, 1998; Riding and
Rayner, 1999; Zhang, 2006) suggested that different personality
types have preferences for different instruction content
sequencing. As demonstrated in our previous experiment, this is
likely because different personality traits have different cognitive
styles for processing given information (Blaylock and Rees, 1984).
For instance, extroverts tend to take in general ideas first, and then
move toward more detailed information. This cognitive style helps
them to draw attention to the whole learning content first (Soles
and Moller, 2001). In contrast, introverts appear to be inductive,
i.e., depth first, followed by breadth as necessary (Myers, 1993;
Myers et al., 1998).
Given this context, there are two content-sequencing alterna-
tives that can be exploited in this study: Breadth-first and depth-first
(Ford and Chen, 2000, 2001). The breadth-first strategy concen-
trates on establishing an overview of learning outcomes before
moving on to further detail. Hypothetically, this approach may be
well suited for extraverted learners, as it provides a big picture of
the learning content prior to addressing the detailed content. In
comparison, the depth-first style employs a bottom-up approach,
starting with low-level details (basic principles) and then pro-
ceeding to teach more abstract concepts, which may match the
inductive nature of introverts (Felder and Brent, 2005; Ford and
Chen, 2001). It should be noted that a number of other studies
have asserted that personality type itself has nothing to do with the
two design alternatives considered in this study (e.g., Stash and De
Bra, 2004). Felder et al., (2002), for instance, showed that learners
who use learning materials that mismatch their personality type
may perform better, because this paradoxically encourages them to
develop a new learning strategy that can cope with a wider range of
materials and experiences in the future.
Notwithstanding alternative perspectives, the experiment
described in this section further explores the possibility that per-
sonality type is indeed a relevant component of computer-based
learning systems design, and the results provide some empirical
evidence of this claim. For this experiment, we developed two
applications (hereafter referred to as ‘Haskell-1’ and ‘Haskell-2’)
Fig. 3. Haskell-1 is designed using a depth-first strategy. A user can only navigate this system using the “<(previous)” and “>(next)” buttons in the bottom right corner.
designed to teach aspects of the Haskell declarative programming
language as part of the Computer Science undergraduate courses at
Massey University (for more details on Haskell, see Thompson
(1999)).
‘Haskell-1’ was designed using a depth-first strategy, which we
hypothesized would be a good approach for introverts, and ‘Has-
kell-2’ was designed using a breadth-first structure for extraverted
learners. Fig. 3 depicts a screenshot of the ‘Haskell’ learning systems
used in this study. The left-hand side of the interface shows the
course map without headings. Although the content of both sys-
tems was the same, the sequence in which the content was deliv-
ered was not, as shown in Figs. 4 and 5. For the purpose of this
experiment, a depth-first strategy was implemented in ‘Haskell-1’
(Fig. 4) and a breadth efirst strategy for ‘Haskell-2’ (Fig. 5).
As shown in Fig. 4, the user starts with ‘Introduction to Haskell’,
followed by all of the subsections e ‘1.1. Why Haskell?’, and ‘1.2.
Features of Haskell’, which introduces the main ideas (data types,
functions, and lists) underpinning the functional programming in
Haskell and discussing how to define a function and what it means
to prove that a function behaves in a particular way. After Section 1
(Introduction) has been fully covered, the system moves on to
Section 2, which details the data types in Haskell. This chapter
discusses two ways of building compound data in the Haskell
language; the tuple and the list. Along with the lists described in
Section 2, the next chapter shows how to define a variety of func-
tions over lists using a combination of list comprehensions and the
built-in list processing functions. Finally, Section 4 revisits the most
critical data and function type e lists e in order to help learners
understand the general operations on lists by examining how
generic or polymorphic functions are handled in Haskell.
In contrast, ‘Haskell-2’ first delivers an overview of Haskell for an
exposition of functional programming, as depicted in Fig. 5. It be-
gins with a preview of a wide variety of ideas in Haskell. This allows
learners to understand how Haskell differs from other program-
ming languages. As learners proceed with this information over-
view, they are given links to subsequent chapters where these ideas
are explained in more detail so that they can reference them and
learn more, i.e., breadth first, depth first as needed. This approach
seems to go quickly for some learners (presumably extroverted
J. Kim et al. / International Journal of Industrial Ergonomics 43 (2013) 450e461 457

Fig. 4. The depth-first content-sequencing structure of ‘Haskell-1’.

Fig. 5. The breadth-first content-sequencing structure of ‘Haskell-2’.

458 J. Kim et al. / International Journal of Industrial Ergonomics 43 (2013) 450e461

learners), because they can take advantage of the similarities be- Explain why you get the answer that you do). They then returned to
tween these and other things that they already know, and they the laboratory one week later and asked to draw the content
understand how the new ideas fit into a larger context. However, it structure on a blank sheet of paper to illustrate their recall of the
takes a great deal of real-world experience to truly understand the whole structure of Haskell.
nuances of a subject, so learners might do less in-depth reading,
and instead just ‘wing it’ with the new ideas. 4.2.3. Experimental design
We are aware that the depth-first vs. breadth-first preference is The experimental design was 2 by 2 between-subjects factorial,
closely related to the level at which learner is operating. If a learner where both personality types (extraverted vs. introverted) and the
is more experienced in programming, then breadth-first might be distinct content structure used (Haskell-1 vs. Haskell-2) served as
the ideal approach. However, if she or he is focusing on one very independent variables. The dependent variables were the number
low-level task, then the learner will require a depth-first approach. of correct answers and the number of revisited pages. To qualita-
To offset this possible nuisance factor, Haskell, which is a functional tively determine their comprehension of the course structure, the
programming language, was chosen over more traditional imper- participants were asked to construct a knowledge structure map
ative or object-oriented approaches. Furthermore, our participants (Smith and Riding, 1999). At the end of the experiment, the par-
(all of them first-year undergraduate computer science students) ticipants were asked to draw the structure of what they had learned
were told that their primary goal in this learning activity was to from Haskell-1 and Haskell-2. They were to make this diagram as
gain a foundation in Haskell, and not to burden them with excessive detailed as they could. The criterion used to judge the drawings was
coding practice. In addition, Haskell-2 provides previews of each how many levels they used to draw the course structure.
chapter, but Haskell-1 does not. This is for the simple, practical
reason that our participants needed to obtain an impression of the 4.2.4. Procedure
breadth-first alterative. First, the MBTI test was administered to all the participants by a
Notwithstanding our caveats related to the experimental design, licensed psychologist, and the participants were then randomly
this is the only such study that we are aware of to compare two seated in a laboratory where either Haskell-1 or Haskell-2 was
design alternatives in order to address the particular concerns of installed. On logging on to the experimental system, the partici-
our research. pants were provided with simple instructions regarding the
experiment. These instructions gave information about the
learning activity, the purpose of the study, and the data protection
4.2. Method
policy.
Two sessions (i.e., the learning session and the knowledge
4.2.1. Participants
structure drawing session) were needed to complete the experi-
Thirty-three first-year students from Massey University took
ment. In the first session, each participant was taught using their
part in this study. We recruited students with high grades (avg.
respective system for half an hour and then asked to answer twenty
GPA ¼ 3.5 out of 4.5) and who were Computer Science majors. They
test questions. Another session followed one week later. In this
were virtually homogenous in terms of their previous learning
time-delayed session, each individual was asked to create a
outcomes in other computer science courses before this experi-
cognitive map showing the structure of Haskell on a blank sheet of
ment. None of them had any previous experience with Haskell. All
paper.
of the participants completed the MBTI questionnaires before the
Note that both Haskell-1 and Haskell-2 only provided the
study (see Table 5 for the details). This experiment was adminis-
learners with “previous” and “next” buttons at the bottom of each
tered as part of the ‘functional programming’ course, and the par-
page to navigate the content, so they could not directly proceed to
ticipants were given 5% course credit.
specific content (see Fig. 3).
4.2.2. Apparatus
4.2.5. Data analysis
Two versions for teaching Haskell were implemented. Haskell-1
The analysis of the data in Experiment II was exactly the same as
was designed with a depth-first strategy, whereas Haskell-2 took a
in Experiment I, except that no statistical analyses were made of the
breadth-first approach, as described above. These two systems had
data collected due to the small sample sizes in some experimental
the same course content, except that Haskell-2 included previews of
conditions. Instead, the cognitive maps showing what the partici-
each chapter (see Figs. 4 and 5). The major difference between them
pants learned were analyzed using the Group Embedded Figures Test
can thus be considered to be the learning approach.
(GEFT; Oh and Lim, 2005), which assesses maps based on how
At the end of the study, each participant completed two types of
many levels are used to represent a knowledge structure. Put
tests comprised of ten multiple-choice questions (e.g., Haskell is a
simply, the more levels are drawn, the more likely that a good
declarative programming language that is (a) more efficient than C
understanding of the content has been obtained. Thus, if the
and Cþþ, (b) a functional programming language, (c) more time
consuming to write programs, (d) less efficient but requires less time to
program) and ten open-ended questions (e.g., Define a function
three Different:: Int / Int / Int / Bool, so that the result Table 6
Task performance in Experiment II.
of three Different m n p is True only if all three of the numbers m,
n, p are different. What is your answer for three Different 3 4 3? Correct answers (mean: s.d.) No. of
participants
Multiple-choice Open-end
revisited
(max: 10) (max: 10)
previous pages
Table 5
The participants of Experiment II. Extra. Intro. Extra. Intro. Extra. Intro.
Haskell-1 6.78 6.14 4.11 7.43 7/9a 2/7a
System Extraverted Introverted Total
(0.83) (1.06) (0.78) (0.97)
Haskell-1 9 7 16 Haskell-2 7.80 6.00 6.40 4.57 3/10a 4/7a
Haskell-2 10 7 17 (0.72) (1.29) (0.72) (1.39)
Total 19 14 33 a
Total number of participants.
 J. Kim et al. / International Journal of Industrial Ergonomics 43 (2013) 450e461
Fig. 6. Knowledge structure map: an example of weak recall (a) and good recall (b).
participants were only able to describe one level of the structure,
they were classified as weak performers. If they described more
than two levels, they were classified as average performers. Par-
ticipants who drew more than two levels were considered to be
good performers.
4.3. Results
Table 6 shows the mean learning performance and the number
of participants who revisited the content with the two different
instructional structures (i.e., Haskell-1 or Haskell-2). In particular,
the last measure was expected to reveal that learners who used a
well-matched material structure would have fewer revisits.
Due to the small sample size for each system, statistical com-
parisons are inappropriate. In addition, minor differences in
learning content (note that Haskell-2 provided previews for each
lesson, but Haskell-1 did not) could have contributed to more cor-
rect answers for the extraverted participants using Haskell-2.
Nonetheless, this trial does demonstrate that the introverted
learners using Haskell-1 (i.e., depth-first) were at least as effective
as the extraverted learners using Haskell-2 (i.e., breadth-first).
These results are in line with other studies (e.g., Ford and Chen,
2001; Moallem, 2003) that identified a relationship between the
learning process and recall of conceptual knowledge. This was also
confirmed by the fact that the extraverted learners using Haskell-1
had more revisits (around 77.8%, i.e., 7 out of 9) than their coun-
terparts (around 28.5%, i.e., 2 out of 7), although these results are
not amenable to statistical analysis.
We also predicted that Haskell-1 would provide the introverts
with better learning outcomes because its in-depth structure is
much more suitable for them, with Haskell-2 acting similarly for the
extraverts. To demonstrate whether matching a learner’s person-
ality with the preferred instructional design might be of central
importance, we compared what they could recall one week later by
means of a Knowledge Structure Map (Fig. 6).
Table 7
Cognitive map for Haskell-1/Haskell-2.
System Personality type No. of participants
Weak recall Good recall
Haskell-1 Introvert 2 5
Extravert 7 2
Haskell-2 Introvert 5 2
Extravert 2 8
459
Fig. 7. A proposed user model for adaptive e-learning systems. The shaded areas are
covered in the empirical studies.
Table 7 lists the number of cases for each contingency using the
Group Embedded Figures Test (GEFT; Oh and Lim, 2005). Again, the
results are not amenable to statistical analysis. However, the two
systems appeared to produce broadly different outcomes. Haskell-1
can be viewed as effective for the introverted participants (5 good
recall of 7 vs. 2 with Haskell-2 of 7) and Haskell-2 for the extraverted
participants (8 good recall of 10 vs. 2 with Haskell-1 of 9).
5. General conclusions and discussion
This article examined how personality traits might be applied to
the design of e-learning systems in order to provide a pragmatic
approach to user modeling for practitioners. The first experiment
demonstrated that the level of extraversion could be such a feature.
Based on this, Experiment II intended to empirically demonstrate
an approach to incorporating personality traits in the design of
structured learning content, which is a novel contribution.
5.1. Embodying personality in the user model
The diagram shown in Fig. 7 illustrates that personality traits
can mediate between the user model and the instructional design,
perhaps dictating the appropriate structure and content. An un-
derstanding of this relationship would allow for the practical
design of computer-based learning systems based on user per-
sonality. Even though the Big Five theory (Costa and McCrae, 1992)
is the more academically dominant approach, rather than MBTI, our
study was based on MBTI, which has been commercially successful
and which simplifies the definition of the learning styles of each
individual learner. It is therefore suggested that designers of
computer-based learning systems could benefit from the practi-
cality of MBTI, rather than employing the Big Five theory.
Our findings regarding the relationship between user modeling
and personality traits are not new. In particular, Wicklein and
Rojewski (1995) claimed that a better understanding of personal-
ity could lead to improved satisfaction of individual learning needs
and could allow educators to provide an optimal learning envi-
ronment. Lauridsen (2001) further contended that adaptive e-
learning systems should focus not only on technologies, but also on
learning styles and personal approaches. However, the user models
employed in most adaptive e-learning systems do not consistently
apply specific content or support for students’ learning activities in
terms of different knowledge acquisition strategies.
Additionally, we sought to provide practical guidance for
designing e-learning experiences optimized for individual learners’
personalities. Our personalized user model can monitor adaptive
effects, so as to refine itself for subsequent adaptations (in this
460 J. Kim et al. / International Journal of Industrial Ergonomics 43 (2013) 450e461
study, in terms of the structure and content of the learning
materials).
5.2. Using the results and further research
The empirical data presented in Section 4 supports the feasi-
bility of implementing a personality-adaptive e-learning system and
some insights as to why ours might be more effective in some
circumstances. One possible approach to designing ‘personality-
adaptive’ e-learning systems would be the development of content
sequencing congruent with the personality traits of the learner.
However, we have not yet applied this approach in real computer-
based learning systems, and thus the success of this method is
another empirical question for further study.
No user modeling for adaptive e-learning systems can encom-
pass all design issues. This study addressed personality level,
determined according to MBTI standards, for both user preferences
and goals as assumed based on MBTI temperaments, and thus
incorporated personality, which we believe to be a less-examined
but crucial factor. Hence, the same results may not be observed
with other personality theories (such as clinical personality theory,
HDM e Human Dimensions Model (Khan and Radcliffe, 2005)) or
other tutoring systems with the same experimental contexts. The
major limitation of this empirical study may be that the partici-
pants were only computer science students, and none were stu-
dents of other disciplines which may have given more diverse
personality types (Bayne, 2004). For example, according to Myers
and McCaulley (1985, p. 42), almost 90% of fine art students were
found to be intuitive, compared with 32% of the general US popu-
lation. In addition, 86% of a sample of US independent studies
students preferred feelingeperceiving, compared to 28% of the
general population.
Another question is how this approach will scale up to complex,
real e-learning system design projects. As mentioned in the intro-
duction, computer-based e-learning systems should be cost effec-
tive and easy to use by teachers and educational providers.
Considering personality traits would require extensive effort,
which may not be pragmatic outside the confines of a research
study. To confirm that this approach is practical, further large-scale
studies will be required.
Acknowledgments
Major funding for this work was from the National Research
Foundation of Korea (NRF) grant funded by the Ministry of Edu-
cation, Science and Technology (MEST) (2012-0008152; 2011-
0028992).
References
Barcellini, F., Françoise, D., Burkhardt, J.-M., 2009. Participation in online interaction
spaces: design-use mediation in an open source software community. Inter-
national Journal of Industrial Ergonomics 39 (3), 533e540.
Bayne, R., 2004. Psychological Types at Work: An MBTI Perspective. International
Thomson Business, London.
Blaylock, B., Rees, L., 1984. Cognitive style and the usefulness of information. De-
cision Science 15 (1), 74e91.
Brown, E., Cristea, A., Stewart, C., Brailsford, T., 2005. Patterns in authoring of
adaptive educational hypermedia: taxonomy of learning styles. Educational
Technology and Society 8 (3), 77e90.
Brusilovsky, P., 2001. Adaptive hypermedia. User Modeling and User-Adapted
Interaction 11 (1e2), 87e110.
Brusilovsky, P., Schwarz, E., Weber, G., 1996. A tool for developing adaptive elec-
tronic textbooks on www. Paper presented at the Webnet (’96) Conference. San
Francisco, CA.
Carver, C., Howard, R., Lane, W., 1999. Enhancing student learning through hyper-
media courseware and incorporation of student learning style. IEEE Transaction
on Education 42 (1), 33e38.
Corno, L., Snow, R., 1986. Adapting teaching to individual differences among
learners. In: Wittrock, M. (Ed.), Handbook of Research on Teaching, third ed.
Macmillan, New York, pp. 605e629.
Costa, P., McCrae, R., 1992. NEO PI-R: Professional Manual: Revised NEO PI-R and
NEO-FFI. Psychological Assessment Resources, Florida.
DiTiberio, J., 1998. Uses of type in education. MBTI Manual: A Guide to the Devel-
opment and Use of the Myers-Briggs Type Indicator. Consulting Psychologists
Press, Palo Alto.
Felder, R., Brent, R., 2005. Understanding student differences. Journal of Engineering
Education 94 (1), 57e72.
Felder, R., Felder, G., Dietz, E., 2002. The effects of personality type on engi-
neering student performance and attitudes. Journal of Engineering Educa-
tion 91, 3e17.
Ford, N., Chen, S., 2000. Individual differences, hypermedia navigation, and
learning: an empirical study. Journal of Educational Multimedia and Hyper-
media 9 (4), 281e311.
Ford, N., Chen, S., 2001. Matching/mismatching revisited: an empirical study of
learning and teaching styles. British Journal of Educational Technology 32 (1),
5e22.
Germanakos, P., Tsianos, N., Lekkas, Z., Mourlas, C., Samaras, G., 2008. Capturing
essential intrinsic user behaviour values for the design of comprehensive web-
based personalized environments intrinsic user behaviour values for the design
of comprehensive web-based personalized environments. Computers in Hu-
man Behavior 24 (4), 1434e1451.
Gerjets, P., Scheiter, K., Opfermann, M., Hesse, F.W., Eysink, T.H.S., 2009. Learning
with hypermedia: the influence of representational formats and different levels
of learner control on performance and learning behavior. Computers in Human
Behavior 25 (2), 360e367.
Gilbert, J., Han, C., 1999. Adapting instruction in search of ‘a significant difference’.
Journal of Network and Computer Application 22 (3), 149e160.
Graesser, A.C., Jackson, G.T., McDaniel, B., 2007. AutoTutor holds conversations with
learners that are responsive to their cognitive and emotional states. Educational
Technology 47, 19e22.
Graf, S., Lin, T., Kinshuk, J., 2008. The relationship between learning styles and
cognitive traits getting additional information for improving student modelling.
Computers in Human Behavior 24 (2), 122e137.
Heckmann, D., 2006. Ubiquitous User Modelling. IOS Press, Germany.
Honey, P., Mumford, A., 1986. The Manual of Learning Styles. Peter Honey
Associates.
Horikoshi, W., 1998. Cross-cultural conflict of Asians in the U.S.: applications of
MBTI in a multicultural environment. Paper presented at the 2nd Multicultural
Research Symposium. Gainesville, FL, pp. 29e43.
Hsu, C.-M., Yeh, Y.-C., Yen, J., 2009. Development of design criteria and evaluation
scale for web-based learning platforms. International Journal of Industrial Er-
gonomics 39 (1), 90e95.
Keirsey, D., 1998. Please Understand Me II. Prometheus Nemesis Book Company, Del
Mar, CA.
Khan, A., Radcliffe, K., 2005. Mind Shapes: Understanding the Differences in
Thinking and Communication. Paragon House, St. Paul, MN.
Kilmann, T., 1998. Psychological Type and Culture-East and West. Honolulu, Hawaii.
Kinshuk, T., Lin, T., 2004. Application of learning style adaptivity in mobile learning
environments. Third Pan Commonwealth Forum on Open Learning, Dunedin,
New Zealand.
Kogan, N., 1971. Educational implications of cognitive styles. In: Lesser, G.S. (Ed.),
Psychology and Educational Practice. Scott Foresman, Illinois, pp. 242e292.
Kolb, D., 1984. Experiential Learning: Experience as the Source of Learning and
Development. Prentice Hall, Englewood Cliffs, New Jersey.
Kolb, D., Boyatzis, R., Mainemelis, C., 2000. Experiential learning theory: previous
research and new directions. In: Sternberg, R., Zhang, L. (Eds.), Perspectives on
Cognitive, Learning, and Thinking Styles. Lawrence Erlbaum, NJ, pp. 1e40.
Kwok, M., Jones, C., 1985. Catering for different learning styles. Association for
Learning Technology Journal 3 (1), 5e11.
Lauridsen, O., 2001. Some like it soft. Retrieved October 2006 from. www.hba.dk/
fsk/pdfs/0003102.pdf.
Lawrence, G., 1984. A synthesis of learning style research involving the MBTI.
Psychological Type 8 (2), 2e15.
Lawrence, G., 1997. Looking at Type and Learning Styles. Center for Applications of
Psychological Type, Gainesville, FL.
Meisgeier, C., Murphy, E., Meisgeier, C., 1989. A Teacher Guide to Type: A New
Perspective on Individual Differences in the Classroom. Consulting Psycholo-
gists Press, Palo Alto, CA.
Messick, S., 1970. The criterion problems in the evaluation of instruction: assessing
possible, not just intended outcomes. In: Wittrock, M.C., Wiley, D. (Eds.),
The Evaluation of Instruction: Issues and Problems. Rinehart and Winston,
New York.
Messick, S., 1976. Personality Consistency in Cognition and Creativity, Individuality
in Learning. Jossey-Bass, San Francisco.
Moallem, M., 2003. Applying learning styles in an online course. Academic Ex-
change 7 (4), 209e215.
Myers, I., 1993. Introduction to Type. Consulting Psychologists Press, Palo Alto, CA.
Myers, I., McCaulley, M., 1985. Manual: A Guide to the Development and Use of the
Myers-Briggs Type Indicator. Consulting Psychologists Press, Palo Alto, CA.
Myers, I., McCaulley, M., Quenk, N., Hammer, A., 1998. MBTI Manual: A Guide to the
Development and Use of the Myers Briggs Type Indicator, third ed. Consulting
Psychologists Press, Palo Alto, CA.
 J. Kim et al. / International Journal of Industrial Ergonomics 43 (2013) 450e461
Myers, I., Myers, K., Kirby, L., 1993. Introduction to Type: A Guide to Understanding
Your Results on the Myers-Briggs Type Indicator. Consulting Psychologists
Press, Palo Alto, CA.
Oh, E., Lim, D., 2005. Cross relationships between cognitive styles and learner
variables in online learning environment. Journal of Interactive Online Learning
4 (1), 53e65.
Oztekin, A., Kong, Z.J., Uysal, O., 2010. UseLearn: a novel checklist and usability
evaluation method for eLearning systems by criticality metric analysis. Inter-
national Journal of Industrial Ergonomics 40 (4), 455e469.
Papanikolaou, K., Grigoriadou, M., Kornilakis, H., Magoulas, G., 2002. INSPIRE: an
intelligent system for personalized instruction in a remote environment. Hy-
permedia: Openness, Structural Awareness, and Adaptivity, 215e225.
Papatheocharous, E., Belk, M., Germanakos, P., Samaras, G., 2012. Proposing a fuzzy
adaptation mechanism based on cognitive factors of users for web personali-
zation. Artificial Intelligence Applications and Innovations, 135e144.
Piombo, C., Batatia, H., Ayache, A., 2003, July. A framework for adapting instruction
to cognitive learning styles. In: Advanced Learning Technologies, 2003. Pro-
ceedings. The 3rd IEEE International Conference on. IEEE, pp. 434e435.
Rao, Z., 2002. Bridging the gap between teaching and learning styles in east Asian
context. TESOL Journal 11 (2), 5e11.
Riding, R., Fanning, P., 1998. Cognitive Styles and Learning Strategies. Fulton, London.
Riding, R., Rayner, S., 1999. Cognitive styles and learning strategies: under-
standing differences in learning and behavior. Educational Technology 15
(3), 222e241.
Shuck, A., Phillips, C., 1999. Assessing pharmacy students’ learning styles and per-
sonality types: a ten-year analysis. American Journal of Pharmaceutical Edu-
cation 10 (63), 27e33.
Sloan, E., Jens, K., 1982. Differences and Implications in Faculty and Student Types
on the Myers-Briggs Type Indicator. ASEE, Washington, DC.
Smith, E., Riding, R., 1999. Cognitive style and instructional preferences. Instruc-
tional Science 27, 355e371.
461
Soles, C., Moller, L., 2001. Myers-Briggs type preferences in distance learning edu-
cation. International Journal of Educational Technology 2 (2), 160e167.
Stash, N., Cristea, A., De Bra, P., 2004. Authoring of learning styles in adaptive hy-
permedia: problems and solutions, Paper presented at the 13th International
World Wide Web Conference. New York, NY.
Stash, N., De Bra, P., 2004. Incorporating cognitive styles in AHA the adaptive hy-
permedia architecture. Paper presented at the Web-Based Education LASTED.
Innsbruck, Austria.
Stewart, G., 2006. MBTI distributions for first year IT students. Transforming IT
Education: Promoting a Culture of Excellence, 371e389.
Thompson, S., 1999. Haskell: The Craft of Functional Programming. Addison-Wesley,
Reading, MA.
Vincent, A., Ross, D., 2001. Personalize training: determine learning styles, per-
sonality types and multiple intelligences online. The Learning Organization 8
(1), 36e43.
Vandewaetere, M., Desmet, P., Clarebout, G., 2011. The contribution of learner
characteristics in the development of computer-based adaptive learning envi-
ronments. Computers in Human Behavior 27 (1), 118e130.
Weber, G., Brusilovsky, P., 2001. ELM-ART: an adaptive versatile system for web-
based instruction. International Journal of Artificial Intelligence in Education
12 (4), 351e384.
Whittington, D., Dewar, T., 2000. Type Indicators and Online Learners. The OTiS e-
Workshop. University of Glasgow, UK.
Whitworth, B., 2005. Is There a Relationship between Personality Type and
Preferred Conflict-Handling Styles? An Exploratory Study of Registered Nurses
in Southern Mississippi. Unpublished doctoral dissertation, Capella University.
Wicklein, R., Rojewski, J., 1995. The relationship between psychological type and
professional orientation among technology education teachers. Journal of
Technology Education 7 (1), 57e74.
Zhang, L., 2006. Does studenteteacher thinking style match/mismatch matter in
students’ achievement? Educational Psychology 26 (3), 395e409.