Detecting Play and Learning Styles for Adaptive Educational Games
Renny S. N. Lindberg and Teemu H. Laine
Department of Information and Computer Engineering, Ajou University, Suwon, South Korea
Keywords:
Learning Styles, Play Styles, Educational Games, Adaptation, Questionnaire.
Abstract:
Digital games have emerged as promising tools that aim to make learning more fun. As each person has
different learning and play styles, educational games can become more effective should they adapt content
delivery according to the user’s learning and play styles. To achieve this, models are needed to represent these
styles. Several learning and play style models exist today with varying levels of validity and complexity. With
the emergence of educational games, the importance of using such models for enhancing the effectiveness of
games as viable educational tools increases. However, not many studies have been conducted that simultane-
ously consider both style groups. We first analysed existing learning and play models. Based on the analysis
results, we created a questionnaire that uses Bartle’s Player Types model and Honey and Mumford’s Learning
Style Questionnaire, and tested it with 127 South Korean elementary school children. The results indicated
that within the play style model specific styles were clearly more preferred, whereas learning styles were dis-
tributed more evenly. There also were some differences between genders. The results can be used to inform us
on what sort of stimuli might yield better immersion in educational games and thereby facilitate the learning
process.
1 INTRODUCTION
Digital games have risen to new heights as general
entertainment to most age groups. They also have
expanded beyond their traditional field of leisure, as
they have been increasingly utilized as tools to sup-
port and provide education (Nygren et al., 2012; Co-
enen et al., 2013; Popescu et al., 2011), and as cat-
alysts for people do physical exercise (Morelli et al.,
2011; Macvean and Robertson, 2012). These types
of games that have other purpose than pure enter-
tainment, are generally referred to as serious games.
Within the umbrella of serious games, game-based
learning in particular is a heavily trodden research
field. It has been shown that utilizing games in ed-
ucation can yield positive results in learning and mo-
tivation (Connolly et al., 2012; Popescu et al., 2011;
Coenen et al., 2013; Arnab et al., 2013; Nygren et al.,
2012; Kim et al., 2009; Ke and Grabowski, 2007).
While researching education and especially learn-
ing, it is not uncommon to come across different
learning style models and discussions for and against
them. Researchers have proposed several models for
mapping out various learning styles with the pur-
pose of understanding the differences among hetero-
geneous learners and thereby optimizing the learning
experience for each learning style. Popular learn-
ing styles include Kolb’s Learning Style Inventory
(LSI) (Kolb, 1984), Honey and Mumford’s Learn-
ing Style Questionnaire (LSQ) (Honey and Mumford,
1982), and Felder and Silverman’s Learning Style
Model Felder1988.
There are several game genres – ranging from first
person shooters (FPS) to a large scale country govern-
ing simulations, and from role-playing games to sport
games – that cater for and expect vastly different play
styles. For example, a majority of FPS games expect a
combination of aggressiveness and cunning, whereas
a typical role-playing game might want the player to
have more of a strategic mindset. Perhaps due to the
youth of the gaming research field, there are not many
validated play style models that tackle the heterogene-
ity in player population. Examples of models that are
dedicated to categorize play styles are: Bartle’s Player
Types (BPT) (Bartle, 2015), Bateman’s Demographic
Game Design (DGD1, DGD2) (Bateman and Boon,
2005), and Lazzaro’s Four Keys (4K) (Gomez et al.,
2004). In addition, personality models have also been
suggested for identifying different player types, such
as Keirsey Temperament Sorter (KTS) (Keirsey and
Bates, 1984).
The reasoning for adapting educational games to
cater for both play and learning styles of a person is
threefold. Firstly, adaptive educational games could
Lindberg, R. and Laine, T.
Detecting Play and Learning Styles for Adaptive Educational Games.
In Proceedings of the 8th International Conference on Computer Supported Education (CSEDU 2016) - Volume 1, pages 181-189
ISBN: 978-989-758-179-3
Copyright
c
 2016 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
181

offer learning materials to the person in a way that
the person would be more attuned to receiving and
processing. Secondly, by offering a play style more
suitable for the person, the game mechanics would be
easier to understand and thus cut down the learning
curve at the beginning. Thirdly, a well-aligned play
and learning styles, through which learning materials
are provided, could increase the person’s motivation
to continue playing the game and increase the possi-
bility for the person to enter the flow state (Mirvis,
1991).
This work is part of a three-year (2015-2018) Ko-
rean National Research Foundation project, which
aims at developing a learner-aware adaptation frame-
work for developing games for programming educa-
tion. In this paper, we presents the results of mul-
tiple research activities that contribute towards the
project’s goals. Firstly, we review existing models for
learning and play styles. Secondly, based on the re-
view, we select one learning style model and one play
style model, and develop a questionnaire for identi-
fying a child’s play and learning styles. Thirdly, we
present an analysis of data that were collected from
127 South Korean elementary school students using
the questionnaire. The purpose of this study is to
see how learning and play styles manifest themselves
in a sample population of South Korean elementary
school students.
2 BACKGROUND
2.1 Game-based Learning
As we briefly stated above, using games to convey ed-
ucational content has shown positive results in mul-
tiple areas. Connolly et al. (Connolly et al., 2012)
conducted a systematic literature review on 129 pa-
pers of which 82 presented educational games in a
range of fields such as business, engineering, ge-
ography, health, history, language and mathematics.
The authors discovered that many of the reviewed pa-
pers reported that students enjoyed a game-based ap-
proach and found it motivating. They also mentioned
that the diversity of research on positive impacts and
outcomes associated with playing digital games was
noteworthy.
As shown in Connolly et al’s survey and sev-
eral other studies (Garris et al., 2002; Eastern, 2009;
Bostan, 2009; Nygren et al., 2012), using games to
convey educational content shows clear results when
it comes to raising people’s motivation. However, an
educational game that does not respond well to play-
ers’ learning and play styles might damage motivation
and therefore negatively affect the learning experi-
ence. Moreover, as educational games’ main focus is
to educate the player on selected topics through game-
play, it is important to select appropriate pedagogical
methods, game mechanics and content modalities to
support the player’s personal play and learning styles.
Several studies have been conducted in the
field of game research that focus on either play
style(McMahon et al., 2012; Orji et al., 2014) or
learning style (Koops and Hoevenaar, 2012; Bontchev
and Vassileva, 2011). We were, however, unable to
find many studies that would consider both styles si-
multaneously. Finally, many models have been pro-
posed to organize various play and learning styles.
Thus, in the following sections we describe and com-
pare the most well-known and well-tested models to
see how they differ from one another.
2.2 Learning Styles
Learning styles have been researched by scholars for
decades. In the following, we present three popular
learning style models.
2.2.1 Kolb’s Learning Style Inventory
One of the best known learning style models was cre-
ated by Kolb, who published his Learning Style In-
ventory (LSI) in 1984 (Kolb, 1984). In his model,
Kolb defined four distinct learning styles as follows:
• Accommodator (CE / AE): Prefers practical
hands-on approach to problems.
• Converger (AC / AE): Prefers hands-on approach
to applying theories.
• Diverger (CE / RO): Prefers discussing and think-
ing the problem through.
• Assimilator (AC / RO): Prefers reasoning and the-
oretical approach.
Kolb’s learning styles are based on the four-stage
learning cycle, which consists of: Concrete Expe-
rience (CE), Reflective Observation (RO), Abstract
Conceptualization (AC) and Active Experimentation
(AE). Essentially, the four learning styles are com-
binations of these four stages. For instance, Accom-
modator is a combination of Concrete Experience and
Active Experimentation (Kolb, 1984). Since its emer-
gence, LSI has been developed further with the lat-
est iteration specifications having been published in
2005 (Kolb and Kolb, 2005).
CSEDU 2016 - 8th International Conference on Computer Supported Education
182

2.2.2 Honey and Mumford Learning Style
Questionnaire
Honey and Mumford (Honey and Mumford, 1982)
built their Learning Style Questionnaire (LSQ) on top
of Kolb’s LSI. Their purpose was to use the model
in the context of management of workspaces. Honey
and Mumford redefined LSI’s four learner types as:
• Activist: Learns by doing; prefers a practical ap-
proach.
• Theorist: Learns by an analytical approach; needs
to understand the theory first.
• Pragmatist: Needs to see a connection between
the theory and task at hand; prefers applying the-
ory in real life.
• Reflector: An observer who does not directly
interact, but rather collects data from multiple
sources.
Honey and Mumford’s model has a four-stage cyclical
form with the following stages: (i) Having an experi-
ence, (ii) Reviewing the experience, (iii) Concluding
from the experience, and (iv) Planning the next steps.
Their model suggests that different people are more
attuned to different stages, thus individual learning
processes should be adapted accordingly.
2.2.3 Felder-Silverman Learning Styles Model
Originally aimed specifically at engineering students,
Felder-Silverman Learning Styles Model was created
by Felder and Silverman (Kapadia, 2008) and later
improved upon by Felder and Soloman, who used the
model to derive a questionnaire instrument named as
the Index of Learning Styles (ILS) (Oxford, 2003;
Felder and Spurlin, 2005). The model maps the learn-
ers into four dimensions: (i) Active and Reflective,
(ii) Sensing and Intuitive, (iii) Visual and Verbal, and
(iv) Sequential and Global. The extremes of these di-
mensions are defined as:
• Active learner: Hands on approach, discussing
about the project. Likes group work.
• Reflective learner: Thinks about the problem qui-
etly at first. Prefers working alone.
• Sensory learner: Likes to learn facts. Enjoys solv-
ing problems with well established methods.
• Intuitive learner: Enjoys discovering connections
between the problems.
• Visual learner: Remembers best anything that
they have seen, such as images, flow charts, di-
agrams.
• Verbal learner: Gets more out of words, both writ-
ten and spoken.
• Sequential learner: Approaches a problem in a
stepwise manner, solving one subproblem at a
time.
• Global learner: Move forward in large jumps, ab-
sorbing materials in a seemingly random manner.
It is worth noting that “Verbal” was “Auditory” in
the original model, and the term “Verbal” was pro-
posed by Felder and Solomon in ILS (Oxford, 2003;
Felder and Spurlin, 2005). Furthermore, within one
dimension, such as Active-Reflective, a person may
converge towards one of the extremes or diverge away
from both extremes.
2.3 Play Styles
Compared to learning styles, play styles for games is
a much less researched area. In the following section,
we describe three play style models and one person-
ality model, which have been used in game studies.
2.3.1 Bartle’s Player Types
Perhaps the best known play style model was de-
scribed by Bartle in his paper “Hearts, Clubs, Di-
amonds, Spades: Players who suit MUDs” (Bartle,
2015). The Bartle’s Player Types (BPT) model was
originally created for multiplayer games, from where
the term MUD (Multi-User Dungeon) has been de-
rived. Since its emergence, the model has been cited
often in game-related studies across game genres.
The BPT model defines four player types as Killers,
Achievers, Explorers and Socializers. These four
player types are distributed in the corners of a two-
dimensional space where the X-axis represents Player
vs. World and the Y-axis denotes Acting vs. Interact-
ing. The characteristics of the four types are:
• Killers: Aggressive predators of other players
who get their enjoyment from harassing other
players. To some extent they do socialize with
other Killers, explore the game and gather re-
sources, but these are only means to an end. Also
being in control and dominating the game are ap-
pealing to Killers.
• Achievers: Solely focused on beating the game as
perfectly as possible. In a multiplayer game this
typically means gathering resources and level-
ups, whereas in a single player game the focus
would be to clear every achievement and the game
perfectly.
Detecting Play and Learning Styles for Adaptive Educational Games
183

• Explorers: Interested in mapping the game me-
chanics inside and outside. Inside means discov-
ering every corner of the game and outside refers
to utilizing possible bugs that might give benefits
to the player.
• Socializers: Are into the game nearly solely to use
it as a communication forum to connect and talk
with other people. The other game mechanics rep-
resent merely a backdrop to achieve this.
Despite originally creating the model for research-
ing and mapping socio-centric behaviour of players
in multiplayer games, Bartle’s model can also, with
ease, used to map the dominant play style of a in-
dividual playing a single player game. This view is
supported by the emergence of newer models that are
very similar to Bartle’s model, such as DGD1 and
Four Keys. These models will be discussed in greater
depth on the following sections.
While Bartle did not create a questionnaire for his
model, the Bartle Test was proposed by Andreasen
and Downey (Andersen and Downey, 2001). The Bar-
tle Test is based on binary choice questions following
the style “I would rather do this or that”, and it has
gained significant popularity as a play style mapping
tool within the gaming community.
2.3.2 Demographic Game Design Model (DGD1
and DGD2)
The first Demographic Game Design Model (DGD1)
was revealed in the book “21st-Century Game De-
sign” by Bateman and Boon (Bateman and Boon,
2005). DGD1 comprises four player styles:
• Wanderer: Associated with experience and iden-
tity. They enjoy new experiences and discovering
the game area or story elements rather than chal-
lenging gameplay, which they tolerate in order to
advance in the game.
• Conqueror: Aim to defeat the games they play.
They enjoy challenges introduced by the game
and the pay off when overcoming those chal-
lenges.
• Manager: Aim to master the game and understand
its gameplay mechanics. They do not care much
about winning the game, but rather to reach point
where there is nothing new to learn.
• Participant: Associated with immersing them-
selves into a game world mainly through social
interaction with other players or with game char-
acters. Any game that gives emotional feedback
(good or bad) appeals to this player type.
The DGD1 model is dependant on the Myers-Briggs
Type Indicator that has 16 different personality types,
which, in turn, are different combinations of 8 core
types: Extraversion/Introversion, Sensing/Intuition,
Thinking/Feeling, Judging/Perceiving (Myers, 1962).
The DGD1 model was developed further into
DGD2, which aimed at improving upon DGD1 (Bate-
man et al., 2011). Furthermore, the Brain Hex model
has been proposed that combines the two older DGD
models and adds new types to them (Seeker, Sur-
vivor, Daredevil and Mastermind) while discarding
the Manager type (Nacke et al., 2014).
2.3.3 Keirsey Temperament Sorter
In their book “Please understand me” (Keirsey and
Bates, 1984), Keirsey and Bates proposed the Keirsey
Temperament Sorter (KTS), a personality question-
naire for identifying temperaments. This widely used
personality mapper is quite similar to Myers-Briggs
Type Indicator (Myers, 1962), but they have some dif-
ferences. KST defines four temperaments as:
• Artisan: Very practical who like working with
their hands. They are adventurous in the nature
and always looking for new challenges.
• Guardian: Social persons who are serious about
their duties and responsibilities. They believe in
following the rules and cooperating with others.
• Rational: Tend to be pragmatic and focus on prob-
lem solving. They also have insatiable hunger to
accomplish their goals and work tirelessly on any
project that they are focused on.
• Idealist: Believe that friendly cooperation is the
best way to achieve one’s goals. They shy away
from conflicts as these upset them.
“Please Understand Me II” (Keirsey and Bates,
1998) placed these temperaments into a two-
dimensional space akin to Bartle’s model. The X-axis
represents Internals versus Externals and the Y-Axis
comprises Change versus Structure.
It is important to note that KST was not designed
to be a play style model, but rather to map out observ-
able personality traits. Therefore, it can help game re-
searchers understand the player’s personality, which
has a direct influence on their play style. Following
this train of thought, KST could also be placed under
the category of learning styles because it can be used
to understand the personality traits of heterogeneous
learners.
2.3.4 Four Keys
Lazzaro published her Four Keys (4K) model in a
conference paper “Why We Play Games: Four Keys
to More Emotion without Story” (Gomez et al., 2004)
CSEDU 2016 - 8th International Conference on Computer Supported Education
184

in which she discussed the emotions created by game-
play. The 4K model’s approach to mapping players is
through different emotions that games may evoke in
them:
• Hard Fun: Competitive and playing to beat the
game that also requires strategy.
• Easy Fun: Exploration and wanting to figure the
game out.
• Altered States: Game is a tool to pass time, to
clear thoughts, or to change one’s mental state.
• The People Factor: Play the game to socialize and
spend time with friends. Game itself is not impor-
tant.
3 EVALUATION
As we have shown above, there are several learning
and play style models which could be used for mak-
ing educational games more adaptive. In this section,
we present an evaluation of play and learning styles
among South Korean elementary school students. Be-
fore presenting the results, we explain the details of
research design, including the formation of a ques-
tionnaire.
3.1 Research Design
3.1.1 Participants
The test group consisted of 127 South Korean elemen-
tary school students of whom 33 were from the 3th
grade (10 years old) and remaining 94 were from the
5th grade (12 years old). The ages follow the Korean
age reckoning system where a person is one year old
at the time of birth. The test group had a nearly equal
gender distribution with one undisclosed, 61 females
and 65 males.
3.1.2 Instrument Development
The questionnaire, based on the models proposed by
Honey and Mumford (LSQ) and Bartle (BPT), was
first created in English and later translated to Korean.
A teacher from the target school helped in validating
the questionnaire in two ways. First, she confirmed
the clarity of the questions in English format. Sec-
ond, after translation of the questionnaire to Korean,
she validated the question to ensure that the language
used was appropriate to the target population’s com-
prehension ability.
Honey and Mumford’s learning style model was
chosen over Kolb’s model mainly due to having sim-
pler terms describing the four styles. The comparison
conducted by Coffield et al. (Coffield et al., 2004) in-
dicated the two models to be standing on a very equal
footing.
Bartle’s Player Types model was chosen mainly
due to the lack of validated competing models and
also to avoid unnecessary complexity that would en-
sue from using models that define play styles overly
detailed, such as the Brain Hex model, that has seven
styles in total (Nacke et al., 2014). Despite hav-
ing originally been intended mainly as a tool for
analysing socio-centric interaction of different people
within multiplayer games, Bartle’s player types can
with relative ease be used to define play styles within
single player games, as was shown in section 2.3.1.
It should be noted that our questionnaire does not
expect everyone to play games, but if someone plays
even irregularly, they will naturally adopt a certain
play style. It is to be expected that a player’s play
style might vary depending on his or her’s familiarity
with the game being played.
After a demographics section, the questionnaire is
divided into three portions. The first portion is meant
to find out about the gaming habits of the respondents,
including how much they play games, their preferred
game genres, and their favourite game titles. The pur-
pose of this portion is to get an overview of the land-
scape of digital gaming among South Korean elemen-
tary school students.
The second portion focuses on identifying play
styles of the respondents. It is based on the afore-
mentioned Bartle Test. We transformed the Bartle
Test’s binary choice questions into four-point Likert
scale (“Strongly Agree”, “Somewhat Agree”, “Some-
what Disagree”, “Strongly Disagree”) to allow greater
flexibility and depth for answering and data analy-
sis. Furthermore, as the test group consisted of ele-
mentary school students, we simplified the questions
in order to decrease the chances of misunderstanding
the questions or simply not understanding them at all.
Another reason why rewording the questions was nec-
essary was due to the Bartle Test’s original purpose of
being used by multiplayer gamers, whereas our ques-
tionnaire did not make any assumptions of the types
of games the respondent might like to play.
The third portion of the questionnaire focuses on
mapping learning styles of the respondents. It is
a shorter version of the original Honey and Mum-
ford Learning Style Questionnaire with simplified
language. We applied Likert scale statements instead
of yes/no statements, which were originally used to
determine a dominant learning style. Rewording the
Detecting Play and Learning Styles for Adaptive Educational Games
185

questions was necessary so as to make the statements,
which were originally aimed at adults, easier for chil-
dren to understand. This rewording, however, was
done carefully so as to retain the original meanings
of the claims used by Honey and Mumford in their
questionnaire.
Our questionnaire contains 16 questions per
model, with 4 dedicated for each style. The origi-
nal Bartle’s Test has 41 questions and the number of
questions in LSQ is 40 or 80 depending on the ver-
sion. Both original questionnaires were deemed to be
too long because children are known to have short at-
tention spans(Abadzi, 2008).
3.1.3 Data Collection and Preprocessing
After receiving appropriate permissions to conduct
an experiment at a South Korean public elementary
school, the questionnaire was distributed to students
in four classes with the aid of teachers who briefly ex-
plained the purpose of the questionnaire to their stu-
dents. The students were encouraged to ask questions
should they find any question difficult to understand.
The teachers remained in the classrooms for the dura-
tion of filling in the questionnaire in order to answer
any questions raised by the students.
The questionnaire data consisted of 127 re-
sponses, which were pruned down to 100 (46 Fe-
males, 53 Males, 1 undisclosed) for play style and
learn style analyses. Pruning was done by disregard-
ing questionnaires that were not properly filled, that
is multiple blank answers or “Strongly Disagree” or
“Strongly Agree” in all statements. For play time
analysis, we were able to use the entire data set.
3.2 Results
The following sections present the results of the ques-
tionnaire data analysis. After describing the play time
analysis, we show the results on how the play and
learning styles manifested themselves among the re-
spondents.
3.2.1 Play time
In the gaming habits portion of the questionnaire, we
investigated how much the respondents spend time
playing video games. Figure 1 illustrates the respon-
dents’ estimations of their video game playing time
within a scale of 1-6 where 1 depicts daily playing
and 6 indicates never playing. Mean (µ) and stan-
dard deviation (σ) values are also reported. Answers
within the range of 1 to 3 can be considered regular
to semi-regular playing and 4-6 to be from irregular
to very rarely or never. Out of all respondents there
were clearly more males who played games regularly,
with 52 out of total of 65 answering between 1 to 3,
whereas among females the difference was practically
50/50 as 30 out of 61 answered between 1 to 3.
Figure 1: How regularly do respondents play video games.
3.2.2 Play Styles
The length of the play style portion of the question-
naire was 16 statements in total. It was divided into
four sections based on Bartle’s play styles: Killer,
Achiever, Explorer and Socializer. There was no op-
tion given to skip any of the statements separately, but
if the respondent stated that they do not play games at
all, they could jump directly to the learning style por-
tion of the questionnaire.
In order to identify a dominant play style for each
of the respondents, we used a point system in which
we assigned 1, 0.5, -0.5 and -1 points for answers
of Strongly Agree, Somewhat Agree, Somewhat Dis-
agree and Strongly Disagree, respectively. The points
in each style category were then added up, and the
style with the highest points was designated as the
respondent’s dominant play style. This was done so
as to see which style would be the strongest style for
each respondent. If there were two or more equally
strong styles, all of them were included in the total
counts. A possible secondary style can also identified
through the point system. Figure 2 shows a histogram
of dominant play styles of the respondents. Socializer
Figure 2: Distribution of dominant play styles.
CSEDU 2016 - 8th International Conference on Computer Supported Education
186

was clearly the most common play style, followed by
Explorer. Interestingly, only five Killers were iden-
tified among females, whereas the number of Killers
was more than twice of that among males. Neverthe-
less, Killer was the least common style. This result
is exemplified by the Killer style statement “I enjoy
teasing other players in a multiplayer game”, which
was rejected by 72% and 89% of males and females,
respectively.
3.2.3 Learning Styles
Similarly to the play style portion of the question-
naire, the learning style portion also consisted of 16
statements, four for each learning style in LSQ: Ac-
tivist, Reflector, Theorist and Pragmatist. We iden-
tified the dominant learning styles using the same
point system than what was applied to the play styles
above. Figure 3 illustrates the distribution of domi-
nant learning styles. Activist and Reflector styles are
clearly more common compared to Pragmatist and the
least popular Theorist. Moreover, the results indicate
that males are more practical and active than females,
whereas females tend to prefer theories and contem-
plation than compared to males.
Figure 3: Distribution of dominant learning styles.
4 DISCUSSION
The evaluation presented above sheds some light to
the play and learning styles among South Korean el-
ementary school students. The results were as we ex-
pected in the sense that the respondents did not have a
single very strong learning and play style while other
styles would have been marginal. Instead, most re-
spondents had several fairly strong styles. A conse-
quence to this is that an adaptive educational game
does not necessarily have to follow strictly one learn-
ing/play style combination throughout the game, but
it can alternate between multiple strong styles that
were identified for a given player. This allows us
to create versatile adaptive games that could change
even between repeated game sessions.
Finding the underlying reasons for highly nega-
tive answers to the Killer statement “I enjoy teasing
other players in a multiplayer game” requires fur-
ther studies, but we have initial ideas to investigate
upon. On one hand, it is reasonable to assume that
some of the students might have felt the need to mask
their negative behaviour. On the other hand, a third
grade teacher commented that her students are partic-
ularly eager to help each other, thus indicating altruis-
tic mindsets, which is clearly against what the Killer
style represents. It would be interesting to conduct
a future study to see whether the results would be
similar with older students who are likely to consume
more “killer” games, such as first-person shooters.
The popularity of the Socializer style can be par-
tially explained by the collective Korean culture and
the general tendency for children to like to play games
together. As for the Explorer style, its popularity is
possibly to have influenced by the natural curiosity of
children. To draw any relevant conclusions on what
caused these two styles to be the most common dom-
inant styles would require additional research within
the same age group and also among older learners.
Currently, the questionnaire is aimed at elemen-
tary school children, which is why we limited the
number of questions to four per learning/play style.
On one hand, using a low number of questions bears
some disadvantages, such as suboptimal accuracy and
lack internal consistency validation. On the other
hand, the time taken for filling in a concise ques-
tionnaire is less likely to exceed the children’s atten-
tion span. If the questionnaire would be aimed at
teenagers or adults, it could be reasonable to include
more questions to get a greater amount of data for
analysis, hence increasing the accuracy and enabling
thorough internal consistency validation. However, if
the questionnaire is to be used within a game, then
a lengthy questionnaire might severely damage the
gameplay experience.
The lack of validation for reliability and internal
consistency is certainly something that we must con-
sider in the future evolutions of the questionnaire. Re-
liability metrics, such as Cronbach Alpha, can only be
performed on data set having statements that measure
the same property. In its current form, our question-
naire is minimized so that the statements for any given
style are measuring different properties related to that
style. However, since we based our questionnaire on
previously tested models, it is reasonable to assume
that the chosen questions, albeit simplified for the use
of children, also have a sufficient degree of reliability.
In the Evaluation section, we were able to suf-
ficiently map the students within learning and play
style models, thus enabling us to tell what kind of
Detecting Play and Learning Styles for Adaptive Educational Games
187

stimuli might enhance the respondents’ immersion
and therefore increase the likelihood for a higher de-
gree of learning. In the next step of this research
project, we envision using what we learned here for
making an adaptive educational game prototype and
evaluate it with elementary school children. Essen-
tially, for such an adaptive game to use learning and
play style models in an effective way, it should present
a questionnaire at the beginning of the game, make
calculations to deduce the learner and play styles for
the player, and utilize the identified styles to adapt the
gameplay and the game content. The questionnaire
could be disguised in form of a character building dia-
logue in order to connect it closely to the game world.
It would be even better if the game could be done in
such a way that it would automatically learn about the
play and learning styles as the player plays the game,
and consequently generate the game area and educa-
tional content to match what it infers to be the best
options for the player. Doing so, the game would not
lose its initial fun factor as the player would not have
to go through a questionnaire at the beginning.
Making a game as we envisioned above would re-
quire a lot of attention on gameplay elements, such
as the type of gameplay (e.g., asymmetric, emergent,
cooperative), and the modes that can be played (e.g.,
isometric, first-person). Another great challenge is to
cater for various learning styles by presenting educa-
tional content in different ways. Finally, some play
and learning styles require the game to offer a multi-
player option to some extent.
Our questionnaire is based on two well-known
learning and play style models, which have been used
and validated by scholars across disciplines. Despite
this and a lot of work invested in ensuring that the
questionnaire was as clearly worded as possible to be
suitable for elementary school children, it was evident
from the feedback that a few of the respondents did
not fully understand all of the questions. We intend to
further enhance the questionnaire in order to ensure
its full comprehensibility among the target audience.
As this was our first iteration of the combined
questionnaire, we wanted to keep the number of styles
as low and as clear as possible. In the future we
may consider further modifications to the question-
naire such as adding more play styles from models
such as Brain Hex (Nacke et al., 2014). Further-
more, once we have a working game prototype, taking
into consideration additional factors such as the Self-
Determination Theory (Ryan and Deci, 2000) would
be relevant.
5 CONCLUSIONS
In this study we crafted a questionnaire to map
out elementary school students’ play and learning
styles in order to make more adaptive educational
games. After analysing and comparing popular learn-
ing and play style models, we decided to use Bartle’s
Player Types questionnaire and Honey and Mum-
ford’s Learning Style Questionnaire. Both of these
questionnaires were modified to use Likert scale state-
ments and language used was first simplified and
then translated to Korean. The data collected with
the questionnaire showed us the distribution of learn-
ing and play styles among South Korean elementary
school students.
Our questionnaire is significant because until now
most of the previous learning style and play style
mapping instruments were created for older respon-
dents and the two style groups were rarely combined.
While the questionnaire yielded results that are useful
for understanding the types of learners and players in
South Korea, we intend to perform further studies in
other countries such as Finland so as to compare the
results between cultures. Moreover, in order to un-
derstand the interplay among learning styles and play
styles, a study on the intra-model and inter-model re-
lationships is needed.
The next research step is to first map out possi-
ble relationships between the two style models from
the gathered data. Simultaneously, we will start mak-
ing a prototype game for programming education that
will utilize these findings. This will require tackling
important design issues such as how to create an ed-
ucational game that caters for different learning and
play styles.
ACKNOWLEDGEMENTS
This work was supported by the Korean Na-
tional Research Foundation project NRF-
2015R1C1A1A02036469, and the Korean Ministry
of Science, ICT, and Future Planning under the
Global IT Talent support program (NIPA-2014-
H0904-14-1004) supervised by the National IT
Industry Promotion Agency. We would also like
express our deepest gratitude to Sunglae Kim and
Gloria Hyun for their invaluable help with translating
the questionnaire.
CSEDU 2016 - 8th International Conference on Computer Supported Education
188

REFERENCES
Abadzi, H. (2008). Efficient Learning for the Poor: New
Insights into Literacy Acquisition for Children, vol-
ume 54. Washington, DC: The World Bank, Washing-
ton D.C.
Andersen, E. and Downey, B. (2001). The MUD personality
test. The MUD Companion, 1:33–35.
Arnab, S., Brown, K., Clarke, S., Dunwell, I., Lim, T.,
Suttie, N., Louchart, S., Hendrix, M., and De Fre-
itas, S. (2013). The development approach of a
pedagogically-driven serious game to support Rela-
tionship and Sex Education (RSE) within a classroom
setting. Computers and Education, 69:15–30.
Bartle, R. (2015). Hearts , Clubs , Diamonds , and Spades
: Players who suit MUDs HEARTS , CLUBS , DIA-
MONDS , SPADES :.
Bateman, C. and Boon, R. (2005). 21st Century Game De-
sign. Charles River Media, 1st edition.
Bateman, C., Lowenhaupt, R., and Nacke, L. E. (2011).
Player Typology in Theory and Practice. Proceed-
ings of DiGRA 2011 Conference: Think Design Play.,
pages 1–24.
Bontchev, B. and Vassileva, D. (2011). Learning Objects
Types Dependability on Styles of Learning. Proc. of
8th WSEAS International Conference . . . , pages 227–
234.
Bostan, B. (2009). Player Motivations: A Psychological
Perspective. Comput. Entertain., 7(2):22:1—-22:26.
Coenen, T., Mostmans, L., and Naessens, K. (2013).
MuseUs. Journal on Computing and Cultural Her-
itage, 6(2):1–19.
Coffield, F., Moseley, D., Hall, E., and Ecclestone, K.
(2004). LSRC LearningStyles. Learning, page 84.
Connolly, T. M., Boyle, E. A., Macarthur, E., Hainey, T.,
and Boyle, J. M. (2012). Computers & Education A
systematic literature review of empirical evidence on
computer games and serious games. Computers & Ed-
ucation, 59(2):661–686.
Eastern, J. S. (2009). Making Learning Fun.
Felder, R. M. and Spurlin, J. (2005). Applications, Reliabil-
ity and Validity of the Index of Learning Styles. Inter-
national Journal of Engineering Education, 21(1):103
– 112.
Garris, R., Ahlers, R., and Driskell, J. E. (2002). Games,
Motivation, and Learning: A Research and Practice
Model. Simulation & Gaming, 33(4):441–467.
Gomez, R., Ceballos, L., Coburn, M., Corriere, J., Dixon,
C., Lobel, B., and Mcaninch, J. (2004). Consen-
sus statement on bladder injuries. BJU International,
94(1):27–32.
Honey, P. and Mumford, A. (1982). The Manual of Learn-
ing Styles. Peter Honey Publications, 3th edition.
Kapadia, R. J. (2008). Teaching and learning styles in en-
gineering education. 2008 38th Annual Frontiers in
Education Conference, 78(June):T4B–1–T4B–4.
Ke, F. and Grabowski, B. (2007). Gameplaying for maths
learning: cooperative or not? British Journal of Edu-
cational Technology, 38(2):249–259.
Keirsey, D. and Bates, M. (1984). Please understand
me: Character and temperament types,, volume null.
Prometheus Nemesis Book Company, Del Mar, 5th
edition.
Keirsey, D. and Bates, M. (1998). Please understand me:
Temperament, character, intelligence. Prometheus
Nemesis Book Company, Del Mar.
Kim, B., Park, H., and Baek, Y. (2009). Not just fun,
but serious strategies: Using meta-cognitive strategies
in game-based learning. Computers and Education,
52(4):800–810.
Kolb, A. Y. and Kolb, D. A. (2005). The Kolb Learning
Style Inventory - Version 3.1 2005 Technical Specifi-
cations.
Kolb, D. A. (1984). The Process of Experiential Learn-
ing. Experiential learning: Experience as the source
of learning and development, (1984):20–38.
Koops, M. and Hoevenaar, M. (2012). Conceptual Change
During a Serious Game: Using a Lemniscate Model
to Compare Strategies in a Physics Game. Simulation
& Gaming, 44(4):544–561.
Macvean, A. and Robertson, J. (2012). iFitQuest. Proceed-
ings of the 14th international conference on Human-
computer interaction with mobile devices and services
- MobileHCI ’12, page 359.
McMahon, N., Wyeth, P., and Johnson, D. (2012). Person-
ality and player types in Fallout New Vegas. Proceed-
ings of the 4th International Conference on Fun and
Games - FnG ’12, pages 113–116.
Mirvis, P. H. (1991). Flow: The Psychology of Optimal
Experience., volume 16. Harper & Row, 1st edition.
Morelli, T., Foley, J., and Lieberman, L. (2011). Pet-N-
Punch : Upper Body Tactile / Audio Exergame to
Engage Children with Visual Impairments into Physi-
cal Activity. Proceedings of Graphics Interface 2011,
pages 223–230.
Myers, I. B. (1962). The Myers-Briggs Type Indicator:
Manual (1962). US: Consulting Psychologists Press,
Palo Alto.
Nacke, L. E., Bateman, C., and Mandryk, R. L. (2014).
BrainHex: A neurobiological gamer typology survey.
Entertainment Computing, 5(1):55–62.
Nygren, E., Sutinen, E., Blignaut, A. S., Laine, T. H., and
Els, C. J. (2012). Motivations for Play in the UFrac-
tions Mobile Game in Three Countries. International
Journal of Mobile and Blended Learning, 4(2):30–48.
Orji, R., Vassileva, J., and Mandryk, R. L. (2014). Mod-
eling the efficacy of persuasive strategies for different
gamer types in serious games for health. User Model-
ing and User-Adapted Interaction, 24(5):453–498.
Oxford, R. L. (2003). Language learning styles and strate-
gies.
Popescu, M. M., Romero, M., Usart, M., and National, C. I.
(2011). Using Serious Games in adult education Se-
rious Business for Serious People-the MetaVals game
case study. Learning, 2(1):68–72.
Ryan, R. and Deci, E. (2000). Self-determination theory
and the facilitation of intrinsic motivation. American
Psychologist, 55(1):68–78.
Detecting Play and Learning Styles for Adaptive Educational Games
189