Conference PaperPDF Available

Abstract and Figures

Personalization is essential for gameful systems. Past research showed that the Hexad user types model is particularly suitable for personalizing user experiences. The validated Hexad user types questionnaire is an effective tool for scientific purposes. However, it is less suitable in practice for personalizing gameful applications, because filling out a questionnaire potentially affects a person's gameful experience and immersion within an interactive system negatively. Furthermore, studies investigating correlations between Hexad user types and preferences for gamification elements were survey-based (i.e., not based on user behaviour). In this paper, we improve upon both these aspects. In a user study (N=147), we show that gameful applications can be used to predict Hexad user types and that the interaction behaviour with gamification elements corresponds to a users' Hexad type. Ultimately, participants perceived our gameful applications as more enjoyable and immersive than filling out the Hexad questionnaire.
Content may be subject to copyright.
HexArcade: Predicting Hexad User Types By
Using Gameful Applications
Maximilian Altmeyer
maximilian.altmeyer@dfki.de
German Research Center for Articial
Intelligence (DFKI)
Saarland Informatics Campus, Germany
Gustavo F. Tondello
gustavo@tondello.com
HCI Games Group, Games Institute and
School of Computer Science
University of Waterloo, ON, Canada
Antonio Krüger
krueger@dfki.de
German Research Center for Articial
Intelligence (DFKI)
Saarland Informatics Campus, Germany
Lennart E. Nacke
lennart.nacke@acm.org
HCI Games Group, Games Institute and
Stratford School of Interaction Design and Business
University of Waterloo, ON, Canada
Figure 1: Gameful applications used in the user study. Cloud Clicker (a) asks users to decide which statement is more important
to them. Snowball Shooter (b) provides several gamication elements, which can be interacted with.
ABSTRACT
Personalization is essential for gameful systems. Past research
showed that the Hexad user types model is particularly suitable
for personalizing user experiences. The validated Hexad user types
questionnaire is an eective tool for scientic purposes. However,
it is less suitable in practice for personalizing gameful applications,
because lling out a questionnaire potentially aects a person’s
gameful experience and immersion within an interactive system
negatively. Furthermore, studies investigating correlations between
Hexad user types and preferences for gamication elements were
survey-based (i.e., not based on user behaviour). In this paper, we
improve upon both these aspects. In a user study (N=147), we show
Permission to make digital or hard copies of part or all of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed
for prot or commercial advantage and that copies bear this notice and the full citation
on the rst page. Copyrights for third-party components of this work must be honored.
For all other uses, contact the owner/author(s).
CHI PLAY ’20, November 2–4, 2020, Virtual Event, Canada
©2020 Copyright held by the owner/author(s).
ACM ISBN 978-1-4503-8074-4/20/11.
https://doi.org/10.1145/3410404.3414232
that gameful applications can be used to predict Hexad user types
and that the interaction behaviour with gamication elements cor-
responds to a users’ Hexad type. Ultimately, participants perceived
our gameful applications as more enjoyable and immersive than
lling out the Hexad questionnaire.
CCS CONCEPTS
Human-centered computing Empirical studies in HCI.
KEYWORDS
Gamication; Hexad; Personalization; Prediction
ACM Reference Format:
Maximilian Altmeyer, Gustavo F. Tondello, Antonio Krüger, and Lennart E.
Nacke. 2020. HexArcade: Predicting Hexad User Types By Using Gameful
Applications. In Proceedings of the Annual Symposium on Computer-Human
Interaction in Play (CHI PLAY ’20), November 2–4, 2020, Virtual Event, Canada.
ACM, New York, NY, USA, 12 pages. https://doi.org/10.1145/3410404.3414232
CHI PLAY ’20, November 2–4, 2020, Virtual Event, Canada Altmeyer et al.
1 INTRODUCTION
Gamication, the use of game design elements in non-game con-
texts [
8
], has been used in various contexts, including health, ed-
ucation, commerce or crowdsourcing [
11
]. Researchers have used
gamication to motivate people to reach their goals, enhance their
user experience or turn unpleasant tasks into fun ones [
11
,
37
].
Often, this was done by adopting a “one-size-ts-all” approach
(i.e., using a static set of gamication elements) [
11
,
16
,
37
]. How-
ever, interpersonal dierences in the perception of gamication
elements exist according to prior research [
44
]. These interpersonal
dierences could threaten static gamication approaches. For in-
stance, researchers found that demographic factors such as age [
5
],
gender [
28
], and personality traits [
16
] inuence the perception of
gameful elements.
The inuence of demographic factors prompted gamication
research to study and model user preferences in gamied systems.
As a result, Marczewski [
20
,
44
] proposed the Hexad user types
model—a model that has been developed to understand and ex-
plain user preferences and behaviour in gameful systems [
30
,
43
].
The model consists of six user types, which dier in the degree
to which they need to experience autonomy, relatedness, compe-
tence, and purpose (which are core pillars of Self-Determination
Theory [
35
]). Tondello et al. [
44
] developed a survey, rened it,
and demonstrated its reliability and validity [
42
]. Subsequently, the
Hexad user types model was used to investigate user preferences
in gamied systems across dierent contexts, including physical
activity [
3
], education [
24
], energy conservation [
19
], health [
30
],
and others. These investigations highlighted the usefulness of the
Hexad user types model for personalizing gameful systems. The
correlations between the perception of gamication elements and
Hexad user types [
3
,
44
] enable dynamic adjustments to the ele-
ments of a gameful system.
However, determining the Hexad user type requires people to ll
out a 24-item questionnaire. While this is appropriate and necessary
in academic contexts to not break the psychometric properties of
the Hexad model and ensure scientic rigour, it may be disadvan-
tageous when using the Hexad model to tailor gamied systems
dynamically to its users. Because gamied systems usually aim
at providing an enjoyable and gameful user experience, requiring
users to ll out a survey upfront may break immersion, lead to frus-
tration and thus detrimentally aect the overall user experience of
a gamied system, as suggested by previous work in the context
of gamied surveys [
12
,
13
,
17
,
45
]. Therefore, researching ways to
tailor gamied systems without aecting the user experience of a
gameful system negatively is quintessential.
We contribute to this tailoring of gamied systems by investi-
gating whether gameful applications may be used to predict Hexad
user types, such that gamied systems, which aim to personalize
the set of gamication elements to their users, do not require users
to ll out questionnaires. We analyze whether the interaction be-
haviour of users in gameful applications corresponds to their Hexad
user type as a secondary aspect of our work. Specically, this means
whether users choose to interact with gamication elements that
should—according to their Hexad user type—be particularly rele-
vant to them. Previous studies investigating correlations between
preferences for gamication elements and Hexad user types were
survey-based. This survey methodology implied that participants
did not interact with gameful applications but instead rated their
perception subjectively. Thus, investigating whether we can repli-
cate these correlations within gameful applications is a necessary
next step.
We implemented two gameful applications to investigate the
aspects above in an online study with 147 participants. In “Cloud
Clicker” (see Figure 1a), participants were asked to select one state-
ment that is more important to them out of two statements in
total and then received gameful feedback. In the online study, we
expected that the choice that participants made in Cloud Clicker
corresponds to their Hexad user type. In “Snowball Shooter” (see
Figure 1b), participants experienced gamication elements (Points,
Achievements, Leaderboard, a Virtual Character, and Unlockables)
and could decide with which gamication elements they wanted
to interact. They could interact with the gamication elements
by shooting snowballs at dierent virtual items, which in turn in-
creased their progress in the corresponding gamication elements.
We expected that the amount of interactions with each gamication
element would reect a participants’ Hexad user type.
We found that Cloud Clicker can be used to predict Hexad user
types in a gameful way. We also found that Snowball Shooter does
not explain enough variance to predict Hexad user types reliably,
but that participants indeed mostly interacted with gamication
elements that correspond to their user type. Our results also show
that both gameful applications are more enjoyable and immersive
than completing the Hexad questionnaire.
Taking these ndings into account, we contribute an eective,
gameful way of assessing Hexad user types that is benecial for
enjoyment and immersion. In addition, our system shows that
interaction behaviour in gameful applications indeed corresponds
to a user’s Hexad type.
2 RELATED WORK
We contribute to the eld of personalized gamication. Conse-
quently, we begin our discussion of the literature by motivating
the eld’s importance. Next, we show that the Hexad model is
frequently used, provides consistent ndings across domains and
was found to be the most favourable model to personalize gameful
systems. Afterwards, we demonstrate that gamiying surveys was
successfully used to enhance the user experience while maintaining
the validity of the collected data. We conclude by situating our work
within previous literature.
2.1 Personalization in Gameful Systems
To account for the interpersonal dierences in the perception of mo-
tivational aordances, researchers studied which factors moderate
their perception. For instance, Jia et al. [
16
] investigated personality
traits and to which extend they can be used to tailor gamication el-
ements to users. Their results show that personality traits inuence
the perception of certain gamication elements, suggesting that tai-
loring gamied systems to the personality of users is benecial. Orji
et al. [
29
]’s research—which focused on persuasive strategies and
how personality traits inuence their perception—supports these
results. They found correlations between persuasive strategies and
personality traits.
HexArcade: Predicting Hexad User Types By Using Gameful Applications CHI PLAY ’20, November 2–4, 2020, Virtual Event, Canada
In addition to research on personality, the motivational impact
of demographic factors (e.g., age or gender) on game elements has
been studied. Birk et al. [
5
] found that play motives change with in-
creasing age when older adults focus more on enjoyment instead of
performance (i.e., it seems like—with increasing age—the relevance
of game performance decreases because enjoyment becomes para-
mount). Findings from Altmeyer et al. [
2
] support the decreased
importance of game performance among older adults, showing the
main reason to play is that older adults enjoy spending time with
other people. In addition, the eect of gender has been studied by
Oyibo et al. [
31
], who found that competition and rewards were
perceived as more persuasive by male participants. Similarly, eects
of age and gender have also been described by Orji et al. [
28
], who
investigated dierences in the perception of Cialdini’s persuasion
strategies in a study with 1,108 participants. The authors found that
females are more responsive to most of the strategies.
2.2 Hexad User Types and Gameful Elements
Although the factors mentioned above are useful to personalize
gameful systems, none of those were explicitly developed for this
purpose. The Hexad user type model [
20
] bridges this gap, because
it lays out ways to cluster users of gameful systems and provides
recommendations to inform and tailor a system’s design. It consists
of six user types that dier in how much they are driven by their
needs for autonomy, relatedness, competence and purpose (from
Self-Determination Theory (SDT) [35]).
Philanthropists (“PH”)
are socially-minded, like to take respon-
sibility, and share their knowledge with other users. Overall, they
are driven by purpose.
Socialisers (“SO”)
are also socially-minded but are more inter-
ested in user interaction. Consequently, relatedness is their main
motivation.
Free Spirits (“FS”)
strive for exploration and acting without ex-
ternal control, with autonomy being most important.
Achievers (“AC”)
are driven by overcoming obstacles and mas-
tering dicult challenges. They are motivated by competence.
Players (“PL”)
—maybe the least aptly named of the user types—
are focused on their benets, and are motivated by the will to win
and earn rewards. Hence, extrinsic rewards are most important for
them.
Disruptors (“DI”)
like to test a system’s boundaries and are driven
by triggering change, either positive or negative.
Tondello et al. [
44
] developed a questionnaire to assess Hexad
user types and thereby provide the foundation for further research.
More recently, the authors [
42
] made slight adjustments to the
initial questionnaire and showed its reliability and validity. Addi-
tionally, Tondello [
41
, chapter 3] proposed a method for person-
alized gameful design that can use the Hexad user types to aid in
the selection of the preferred game design elements. As a result,
the Hexad user types model has been used successfully in vari-
ous domains, showing that the model can explain preferences for
gamication elements. For instance, the Hexad model was used to
investigate the perception of gamication elements in the physical
activity domain by using a storyboards-based approach [
3
]. The
authors found that a considerable majority of correlations between
the Hexad user types and preferences for gamication elements
established in [
44
] could be replicated. Similarly, Mora et al. [
24
]
investigated whether using the Hexad model to personalize learn-
ing experiences helps to motivate and engage students better. They
found that the personalized approach yielded higher engagement
of the students, underlining the usefulness of the Hexad model for
tailoring gameful systems.
Besides physical activity and education, Orji et al. [
30
] exam-
ined the Hexad model in the domain of unhealthy alcohol con-
sumption. The authors found that the Hexad user type inuences
the perceived persuasiveness of strategies. Overall, the reported
eects relate to the user type denitions and underline Hexad’s
applicability. Kotsopoulos et al. [
19
] investigated the perception of
certain gamication elements and potential correlations to Hexad
user types in the domain of energy eciency applications at the
workplace. The authors validated the Hexad user model in another
domain by showing that the user types can be used to explain
preferences towards gamication elements. They found similar cor-
relations between gamication elements and user types as Tondello
et al. [
44
]. Tondello et al. [
43
] proposed a conceptual framework for
classifying game elements based on an exploratory factor analysis
of people’s preferences in a general context. Supporting previous
results, expected correlations to the Hexad user types were shown.
Lastly, Hallifax et al. [
10
] analyzed which factors should be con-
sidered when personalizing gamied systems. They investigated
which user models should be used, and compared the BrainHex [
25
]
model, the Hexad model, and the Big-5 personality model [
22
]. They
concluded that the Hexad model is the most suitable typology and
should be used to tailor gamied systems as most of the results that
were found by the authors are in line with the denitions of the
Hexad user types. The authors state that this is likely because the
Hexad model was specically designed for gamication and most
of its user types are based on SDT [35].
2.3 Gamied Questionnaires
We reviewed literature in the domain of gamied questionnaires
because one of the goals of this research is to assess Hexad user
types in a gameful way. In this context, Triantoro et al. [
45
] com-
pared non-gamied surveys against gamied ones. They used the
well-established Big-5 personality model [
22
] and transformed the
Big-5 survey items from Likert scales into a binary, gameful deci-
sion for the gamied version of the questionnaire. In contrast, the
non-gamied questionnaire used the traditional scale. The results
revealed that the Big-5 responses that were assessed in a gameful
way could be used to predict the actual Big-5 responses of the tra-
ditional survey. This prediction strategy is an essential precursor
to our work because it shows that transforming a validated survey
into a gamied version using a binary choice is possible without
compromising the integrity of the collected data. The authors also
found positive eects of gamication on enjoyment and attention,
suggesting a better user experience and advantages for data quality.
Guiding the realization of gamied surveys, Harms et al. [
14
]
suggests a four-step design process. In a follow-up work [
12
], this
design process is extended, used and evaluated. Two HCI designers
went through the aforementioned process transforming a tradi-
tional sports survey into its gamied counterpart. As a result, they
came up with multiple sports-related minigames. The feedback of
CHI PLAY ’20, November 2–4, 2020, Virtual Event, Canada Altmeyer et al.
both designers supported the practical usefulness of the proposed
process. Next, the gamied survey was compared to a non-gamied
counterpart. Similar to the study from McCrae and John [
22
], it was
found that participants considered the gamied version as more
fun, provided more qualitative feedback and spent more time lling
out the survey. Another work by Harms et al. [
13
] even shows
that simplistic approaches to gamify questionnaires lead to positive
eects. In this work, badges could be unlocked by completing the
survey and results showed that participants preferred such a simple
gamied survey over traditional ones, that the gamied design was
found to be more attractive and stimulating, and that participants
perceived the use of badges very positively. Lastly, Keusch et al.[
17
]
conducted a literature review on comparative studies of gamied
surveys. They support the aforementioned positive outcomes, out-
lining that gamied surveys were perceived as more interesting,
more fun and that participants perceived them as more enjoyable
and more comfortable to ll out. They also found that most of the
survey questions were not negatively aected by gamication.
2.4 Summary
To sum up, related work has demonstrated that personalization is
a vital topic in gamication research, emphasizing the relevance
of our work. Previous research has identied numerous factors to
consider when personalizing gameful systems, including age [
2
,
5
],
gender [
28
,
31
] and personality traits [
16
,
29
]. To better understand
gamication preferences and how to tailor gameful systems, the
Hexad user types model has been developed [
20
]. It is the only
model specically designed for the gamication domain and a vali-
dated questionnaire assessing Hexad user types reliably has been
provided by Tondello et al. [42].
Although being comparably new, the Hexad model has been
investigated and used across various domains and contexts and
has been consistently shown to explain user preferences for gami-
cation elements [
3
,
19
,
24
,
30
,
43
,
44
]. These ndings support the
relevance of the Hexad model for personalizing gameful systems
and consequently researching alternative ways of assessing Hexad
user types. A study comparing the importance of dierent user
typologies for personalization of gameful systems conducted by
Hallifax et al. [
10
], highlights the importance and applicability of
the Hexad model even further, showing that the Hexad model ex-
plains preferences of users most reasonably and thus should be
used to eectively tailor gamied systems.
As a third fact, related work has shown that turning surveys
into gameful experiences has positive eects on cognitive and af-
fective reactions, including enjoyment, attention, involvement, and
ease-of-use [
12
,
13
,
17
,
45
]. It was shown that gamication mostly
had no or little eect on the validity of the questionnaire [
17
,
45
]
and that turning Likert-scale questions into gameful binary-choice
decisions provides comparable data that has favourable construct
reliability [
45
]. This evidence compellingly supports our endeavour
of assessing Hexad user types using gameful applications.
Taking all these ndings into account, we extend previous work
by investigating whether interaction behaviour with gamication
elements in gameful applications corresponds to Hexad user types.
Given that previous work considered self-reported preferences
based on storyboards [
3
,
30
] or textual descriptions [
19
,
44
], al-
lowing users to perceive, experience and interact with gamication
elements is a necessary extension. Next, we build upon previous lit-
erature showing the positive eects of gamifying questionnaires by
investigating a gameful way of assessing Hexad user types. Given
that the Hexad model is meant to be used within gamied systems,
providing a gameful way of assessing user types seems to be im-
perative to turn the high potential of the Hexad model for tailoring
gamied systems into practice.
3 GAMEFUL APPLICATIONS
We implemented two gameful web applications using AngularJS
1
and Phaser
2
. In the following, the concepts behind both gameful
applications are explained and discussed. For both applications,
interactive step-by-step tutorials were developed, explaining how
to interact with them. We used the systematic literature review by
Keusch et al. [
17
] to ensure that most relevant papers on gamied
surveys have been considered to inform the design of the gameful
applications.
3.1 Gameful Application 1: Cloud Clicker
Similar to Triantoro et al. [
45
], we decided to transfer the 7-point
Likert scales, which the Hexad questionnaire uses into binary choice
questions in the gameful application. The whole design of the rst
gameful application (“Cloud Clicker”) followed the process pro-
posed by Harms et al. [
12
,
14
] and considered recommendations
and lessons learned from relevant previous work [
12
,
13
,
17
,
45
], as
described in the following. In Cloud Clicker, users see two state-
ments in each of 15 rounds and then have to decide which of these
two statements is more relevant to them (see Figure 1a). We se-
lected one statement for each Hexad user type, which grasps its
underlying motivation. We based the statements on the denitions
given by Tondello et al. [
44
] and on items with a substantial factor
load in the conrmatory factor analysis in the validation study of
the Hexad user types questionnaire [
42
]. This follows the same
procedure as Triantoro et al. [
45
] proposed to translate the Big-5
survey into a gamied counterpart.
Figure 2: Illustrations and statements used in Cloud Clicker.
Next, as part of the “aesthetics and relationship” layer of the
design process for gamied surveys by Harms et al. [
12
,
14
], we
1https://angularjs.org/, last accessed August 5, 2020
2https://phaser.io/, last accessed August 5, 2020
HexArcade: Predicting Hexad User Types By Using Gameful Applications CHI PLAY ’20, November 2–4, 2020, Virtual Event, Canada
decided to present the two statements shown to the user using cloud
visualizations to create visual sensation [
12
] and a convincing and
motivational environment [
17
]. To enhance questions and support
the comprehensibility of the statements [
12
,
17
], we created visual
illustrations for each statement, explaining the statement by using a
gender-neutral avatar (Figure 2). For the Philanthropist, we decided
to focus on the aspect of helping others because the statements It
makes me happy if I am able to help others and I like helping others
to orient themselves in new situations had the highest factor load for
this type [
42
]. For Socialisers, we focused on being part of a team
because the corresponding statement I like being part of a team had
the highest factor load among this trait [42].
Similarly, we used the statements having the highest factor load
in a trait for the Free Spirit (It is important to follow my own path),
the Achiever (I like overcoming obstacles) and the Player (Rewards
are a great way to motivate me) [
42
]. Because the item having the
highest factor load for the Disruptor type (I see myself as a rebel)
was hard to illustrate visually, we decided to use the item I dis-
like following rules instead, which had the second highest factor
load [
42
]. Similar to Hallifax et al. [
10
], statements were presented
to users using a full paired-comparison design (i.e., each user was
asked to evaluate all possible pairs of statements, resulting in 15
rounds of comparisons in Cloud Clicker). The order of the state-
ments as well as whether a cloud with a statement was shown on
the left or on the right side of the screen in a round was randomized
to avoid biasing results [
12
]. Cloud Clicker provides a ranking of
the statements, in which scores of 0–5 are distributed across each
statement representing its corresponding user type, because each
statement is compared to every other statement.
Clouds were shaking and dropping coins when being clicked
(i.e., when a user decided on a particular statement) to increase
the gameful experience of Cloud Clicker. Also, we added sound
eects to indicate interactions with the gameful applications. Both
follow recommendations by Harms et al. [
12
], stating that gameful
feedback in surveys should be provided by using indicators such as
coins as rewards and supported by using auditory feedback. The
coins were colour-coded to represent the corresponding Hexad user
type and were showing a miniature version of the illustration of
the related user type. Similar to Triantoro et al. [
45
], we introduced
time pressure when participants were asked to decide between
two statements in each round. This mechanic had three reasons:
First, it emphasizes and stimulates the gameful experience of the
application [
45
]. Second, it supports spontaneous responses, which
was shown to increase the reliability of responses [
26
]. Third, it
limits the time it takes to complete the application and thus to assess
the Hexad user type, which might be important as we aim to provide
a practical way of assessing Hexad user types in gamied systems.
To allow researchers and practitioners to use Cloud Clicker, we
published the source code as well as all graphical assets on GitHub
3
.
3.2 Gameful Application 2: Snowball Shooter
Cloud Clicker aims at providing a gameful way of assessing Hexad
user types and, thus, builds on the original items from the Hexad
questionnaire [
42
]. In contrast, Snowball Shooter focuses on user
behaviour when interacting with gameful elements and whether it
3https://github.com/m-altmeyer/cloud-clicker, last accessed August 5, 2020
is possible to use this input to predict Hexad user types. As such,
Snowball Shooter provides some gamication elements that users
can interact with (see Figure 1b): The user controls a snowball
cannon and shoots snowballs at items representing each gamica-
tion element. These items are randomly positioned in each round.
Shooting at an item increases the internal score of the correspond-
ing gamication element. Similar to Cloud Clicker, the application
consists of 15 rounds in which users may shoot ve snowballs. It
uses feedback sounds when shooting.
Figure 3: Visualizations of the score levels for Unlockables,
Achievements and Virtual Character in Snowball Shooter.
Following design suggestions from Harms et al. [
12
], we in-
tegrated progression loops in each gamication element. Conse-
quently, there are three score thresholds, which lead to a state
change of the corresponding element (e.g., unlocking a virtual
item). To ensure comparability, these thresholds were kept con-
stant across all gamication elements (given that a maximum score
of 75 can be reached, the rst state change happens at a score of
10, the second at 35 and the third at 60). As a result, it is impossible
to complete all gamication elements. This was explained to users
in the tutorial. It is important to note that—in contrast to previous
work—users could experience the gamication elements instead
of being given storyboards or textual descriptions only. Based on
the user type descriptions by Marczewski [
20
] and the proposed
gamication elements by Tondello et al. [
44
], we integrated the
following gamication elements (Figure 1b):
Unlockables
: Unlockables are expected to motivate
Free Spirits
because they are mainly driven by autonomy [
20
,
44
]. In Snowball
Shooter, we decided to provide treasure chests which could be
unlocked to obtain virtual items. Reecting the score thresholds,
there are three dierent types of treasure chests (wooden, silver,
golden) unlocking items of dierent rarity (common, rare, epic;
see Figure 3).
Achievements
: This element was shown to be especially suitable
for
Achievers
as it supports mastery [
20
,
44
]. In Snowball Shooter,
CHI PLAY ’20, November 2–4, 2020, Virtual Event, Canada Altmeyer et al.
three Achievements (using the score thresholds mentioned before)
can be unlocked: “Snowball Enthusiast” (bronze frame), “Snowball
Master” (silver frame) and “Snowball Guru” (golden frame), see
Figure 3.
Points
: Points have been shown to positively aect
Players
[
20
,
44
]. To underline the value of points as virtual currency (which is
important for Players [
44
]), points can be used to buy modications
for the Virtual Character (see below). The amount of points that
needs to be spent to buy all modications (change or customize
the colour of and add wings for the virtual character, see Figure 1b)
equals the maximum score threshold described above.
Leaderboard
: Social gamication elements such as Leaderboards
are relevant for
Socialisers
[
20
,
44
]. However, ndings by Ton-
dello et al. [
44
] and other researchers [
3
,
19
,
30
] consistently
demonstrate that Leaderboards are also positively correlated to
Players
,
Achievers
and
Disruptors
, which is why we expect to
also nd such correlations in Snowball Shooter. Similar to Mekler
et al. [
23
], we decided to show ctitious users, having scores based
on the thresholds established before, to ensure that all participants
have same chances to rise in ranks.
Virtual Character
:
Philanthropists
are driven by purpose and
like to care for others [
20
,
44
]. Although no signicant correlations
have been shown, we expect that this gamication element should
be particularly relevant for Philanthropists because it may induce
feelings of care-taking. Weused a virtual monster whose emotional
state is coupled to the amount of snowballs shot at the respective
item, a green doughnut. The three changes in its emotional state
(Figure 3) are coupled to the thresholds described before.
4 METHOD
We used the aforementioned gameful applications to investigate
the following hypotheses stemming from our review of the related
literature:
H1: Gameful applications can be used to predict Hexad types.
H1a
: The score of the statements in Cloud Clicker is correlated
to the corresponding Hexad user types and thus may be used to
predict them.
H1b
: The amount of interactions with gameful elements in Snow-
ball Shooter is correlated to the corresponding Hexad user types
and thus may be used to predict them.
H2
: The users’ perception of the gameful applications diers com-
pared to their perception of the Hexad questionnaire.
H2a
: Both applications are perceived as more enjoyable (mea-
sured by the IMI enjoyment subscale) than the Hexad question-
naire.
H2b
: Participants feel more competent (measured by the IMI
competence subscale) using both applications than using the
Hexad questionnaire.
H2c
: Participants feel more pressure (measured by the IMI pres-
sure subscale) in both applications than in the Hexad question-
naire.
H2d
: Both applications are perceived as more immersive (mea-
sured by the PXI immersion subscale) than the Hexad question-
naire.
H1
is motivated by previous work showing that questionnaires can
be transformed into gameful applications without heavily aecting
their validity [
45
]. Triantoro et al. [
45
] demonstrated that the Big-5
personality traits can be predicted based on gameful, binary choices
in their survey, which is similar to our approach and thus motivates
H1a
. The subjective assessments of preferences for gamication
elements using textual descriptions [
44
] or storyboards [
3
,
30
] are
correlated to the Hexad user types, which motivated us to nd
similar correlations when investigating actual interaction with im-
plemented gamication elements (H1b).
H2a
is mainly based on previous work in the domain of gamied
surveys [
12
,
13
,
45
], where positive eects on enjoyment-related
measures have been demonstrated.
H2b
relates back to feedback
provided by gamication elements having been shown to increase
perceived competence [
36
]. We expect to see an increase in per-
ceived pressure in the gameful applications mainly because of the
time pressure that is induced by Cloud Clicker and because of
the round-based nature of both applications (
H2c
). An increase
in pressure does not necessarily aect user experience negatively
but might help to shape optimally challenging systems [
12
]. This
supports users in reaching a ow state, which is described as a state
of increased concentration and enjoyment [
7
]. To better under-
stand whether the perceived pressure related to feelings of ow and
immersion, we also evaluated immersion as part of the PXI ques-
tionnaire and expected it to be higher in the gameful applications
(H2d).
4.1 Procedure
We conducted an online study on Prolic
4
, an online platform
specically targeted at recruiting participants for scientic research
studies. The only requirement was an understanding of the English
language. The study has been reviewed and received ethics clear-
ance through a University of Waterloo Research Ethics Committee
(ORE#41608). It took approximately 15–20 minutes to complete and
participants were paid £2 GBP. After giving informed consent, they
were asked to provide demographic data including age and gender.
Next, the 24-item Hexad user types questionnaire [
42
] was admin-
istered. The questionnaire consists of four items for each of the six
user types, being measured on 7-point scales. To obtain a baseline
for how participants perceived lling out the Hexad questionnaire,
they were asked to ll out the 22-item task evaluation question-
naire of the Intrinsic Motivation Inventory (“IMI”) [
21
,
34
] as well
as the “Immersion” subscale of the Player Experience Inventory
(“PXI”) [
1
]. Both the IMI items and the PXI items are measured on
7-point scales.
Next, participants were asked to interact with the gameful ap-
plications. The order of the gameful applications was randomized.
Before starting the actual application, participants had to complete
a tutorial explaining how to interact with them. In Cloud Clicker,
we measured how often participants chose each statement. Simi-
larly, the number of interactions with each gamication element
in Snowball Shooter was measured. After interacting with each
4https://www.prolic.co/, last accessed August 5, 2020
HexArcade: Predicting Hexad User Types By Using Gameful Applications CHI PLAY ’20, November 2–4, 2020, Virtual Event, Canada
gameful application, participants were asked to ll out the IMI
questionnaire and the PXI “Immersion” subscale.
4.2 Participants
After removing participants who preferred not to answer questions
of the Hexad questionnaire, 147 participants were considered for
the analysis. Of those, 49% self-reported their gender as female, 49%
as male, 0.7% as non-binary and 1.3% preferred not to answer this
question. The mean age was 33 years (SD=11.5, Mdn=30, Min=18,
Max=66). The Hexad user types average scores are similar to the av-
erages reported in the validation study of the Hexad questionnaire
by Tondello et al. [
42
]. Achievers showed the highest average scores
(M=23.6, SD=2.98), followed by Philanthropists (M=22.8, SD=3.18),
Players (M=22.8, SD=3.53) and Free Spirits (M=22.3, SD=3.52). So-
cialisers (M=18.7, SD=4.89) and Disruptors (M=15.0, SD=4.49) fol-
lowed with lower average scores.
5 RESULTS
In this section, we present results related to predicting Hexad user
types based on each gameful application as well as ndings related
to the enjoyment and perception of them.
5.1 Cloud Clicker and Hexad User Types
To analyze whether Cloud Clicker may be used to predict Hexad
user types, a canonical correlation analysis (“CCA”) was conducted
using the score of the six statements in the gameful application as
predictors of the six Hexad user types measured by the Hexad user
types questionnaire. A CCA is preferable when analyzing the associ-
ation strength between two sets of variables and allows to evaluate
the multivariate shared variance between them (i.e., between the
six statement scores of the gameful application and the six scores
of the Hexad subscales) [
38
]. Next, this method is explained based
on Sherry and Hanson’s guide on using CCA [38].
The core idea of CCA is that the set of predictor variables and the
set of criterion variables are combined into a synthetic variable each
(i.e., there is a synthetic predictor and a synthetic criterion variable).
The canonical correlation is the correlation between these synthetic
variables. Each pair of synthetic variables is called a canonical
function (“CF”). Canonical functions are comparable to principal
components in Principal Component Analyses (PCA) with the main
dierence that the CFs are composed of two dierent variable
sets and thus can be seen as an extension of PCA [
46
]. In line
with this, CCA was loosely dened as “a double-barreled principal
components analysis” [40].
As long as there is residual variance left in the two variable
sets which cannot be explained by the already derived canonical
functions, the above process is repeated. This continues until either
no residual variance is left to be explained or there are as many
canonical functions as there are variables in the smaller variable set.
Although CCA can accommodate variables without relying strictly
on multivariate normality [
46
], multivariate normality was assessed
by inspecting univariate Q-Q plots, skewness, and kurtosis of each
variable included in the CCA. The Q-Q plots mainly supported the
assumptions of normality, whereas some variables were shown to
be slightly skewed. However, all skewness and kurtosis values were
within the acceptable thresholds of skewness
<
3and kurtosis
<
8[
18
], given that the maximum absolute values of skewness and
kurtosis were found to be 2
.
5and 6
.
8respectively such that the
CCA could be conducted. Given that 10 participants per observed
variable are recommended to reach a reliability of 80% [
39
], our
sample size can be considered as adequate.
Overall, the full model across all CF was statistically signicant
using the Wilks’s
𝜆
=.256 criterion,
𝐹(
36
,
595
.
59
)=
6
.
01
, 𝑝 <.
001.
This shows that the variance unexplained by the model is 25.63%.
Consequently, the full model is able to explain 74.37% of the variance
(the
𝑟2
type eect size is
.
74) shared between the two variable
sets. Given that the recommended threshold for strong eects was
derived to be
𝑟2=.
64 [
9
], the model can be considered to explain a
substantial amount of variance between the two variable sets. Based
on this, we derive
R1: The score of statements in the gameful
application is substantially associated to the scores of the
Hexad user type questionnaire. This result shows that the two
variable sets are strongly related. As a next step of the CCA, we will
consider the results of the dimension reduction analysis to analyze
whether the predictor variables (the score of the statements in Cloud
Clicker) load on the same canonical functions as the corresponding
Hexad user types. This is important to investigate whether the
statements we have chosen for a certain user type actually represent
this user type, given our data.
The dimension reduction analysis yielded six canonical func-
tions (CF1–CF6) with squared canonical correlations of .36, .33,
.22, .17, .09 and .00 each. The rst ve canonical functions were
statistically signicant whereas CF6 did not explain a statistically
signicant amount of shared variance between the variable sets
(CF1–CF4:
𝑝<.
001, CF5:
𝑝=.
011, CF6:
𝑝=.
99). Therefore, CF6
will not be interpreted as part of the analysis. Figure 4 presents
the structure coecients for CF1–CF5 being stronger than
|.
35
|
(an
upper threshold for weak factor loads established in [
47
]). Dotted
and transparent lines indicate relationships which diered between
the predictor and criterion variables. All standardized canonical
function coecients and structure coecients can be found in
Table 1.
CF 1 CF2 CF3 CF4 CF5
Pred. co rs co rs co rs co rs co rs
G_DI .10 .46 .04 -.21 .19 -.03 -.06 -.10 1.34 .76
G_FS .38 .73 .11 -.12 .32 .36 -.12 -.22 .49 -.17
G_AC -.53 -.37 .21 .01 .89 .85 .34 .36 .76 -.07
G_PL -.11 .18 .99 .91 -.13 -.19 .45 .31 .90 .01
G_PH -.22 -.30 -.21 -.64 -.29 -.46 .72 .45 .69 .10
G_SO -.64 -.67 .40 -.04 .10 -.30 -.57 -.65 .98 .04
Crit.
Hex_DI .30 .41 -.31 -.37 -.18 -.08 -.08 -.17 1.00 .78
Hex_FS .75 .39 -.01 -.34 -.09 .19 -.50 -.41 -.70 -.08
Hex_AC -.56 -.33 -.07 -.05 1.24 .59 -.11 -.19 .37 .30
Hex_PL .24 -.21 .92 .60 -.40 -.12 .26 .05 .16 .36
Hex_PH -.16 -.31 -.57 -.51 -.21 -.12 .99 .36 -.11 .07
Hex_SO -.64 -.63 -.19 -.17 -.58 -.35 -.87 -.46 .03 .13
Table 1: Structure coecients (rs) and standardized canon-
ical function coecients (co) for predictor variables (state-
ment scores in Cloud Clicker: G_DI etc.) and criterion vari-
ables (user type scores: Hex_DI etc.) for the canonical func-
tions. Bold entries represent loads higher than |.
35
|, under-
lined entries represent loads higher than |.50|.
CHI PLAY ’20, November 2–4, 2020, Virtual Event, Canada Altmeyer et al.
Figure 4: Structure coecients for CF1–CF5 being stronger
than |.
35
|for Cloud Clicker. Dotted and transparent lines
indicate relationships which diered between the predictor
and criterion variables.
While standardized canonical function coecients represent the
weights applied to the observed variables to combine the unob-
served synthetic variables, structure coecients are simple bivari-
ate correlations between observed variables and synthetic vari-
ables [
27
]. It can be seen that most predictor and criterion variables
have large structure coecients loading substantially (i.e.,
>|.
5
|
according to [
6
]) on the same canonical functions. This is supported
by the symmetry of the relationships, which can be seen in Figure 4.
In terms of strength of the correlation, the Free Spirit subscale is
an exemption to this as it is the only variable having no structure
coecient higher than |.5|but loads moderately [6] on CF1.
Thus, we formulate
R2: All predictor and criterion variables
have medium to large structure coecients loading on the
same canonical functions
. This result shows that there is not
only a substantial relationship between the variable sets (
R1
) but
also that the predictor variables load on the same canonical func-
tions as the criterion variables. This means that the statements we
have chosen represent the corresponding Hexad user types.
DI_R FS_R AC_R PL_R SO_R PH_R
G_DI .32 -.23
G_FS .21 .36 -.36
G_AC .35
G_PL .50
G_SO -.32 .44
G_PH -.22 .56
DI .44 -.22 -.37
FS .52 -.38
AC .45
PL -.37 .60 -.32
SO -.52 -.31 .64
PH -.31 -.23 -.26 .54
Table 2: Spearman’s rank correlation coecients between
the ranked Hexad user type scores (DI_R etc.) and the score
of each statement in Cloud Clicker (G_DI etc.) as well as be-
tween the ranked Hexad user type scores and the absolute
Hexad user type scores (DI etc.). All 𝑝<.01.
To further analyze the correlations between the Hexad user types
questionnaire scores and the scores obtained through the gameful
application, we calculated bivariate Spearman’s rank correlation
coecients. Since the gameful application requires users to make a
binary choice whereas the original Hexad user types questionnaire
allows to have the same score in multiple user types, we ranked
the Hexad user type scores by assigning values from 0–5 to ensure
comparability. These ranked Hexad scores are denoted by “_R” in
the results. As a reference for interpretation, we also added the
correlations between the absolute score of each user type of the
Hexad and the ranked Hexad user type. The results are shown in
Table 2.
Providing further support for the results of the CCA, it can be
seen that there are medium to large size correlations [
6
] between
the scores of each statement of the gameful application and the
ranked Hexad user types. This leads to
R3: The ranked Hexad
scores are positively correlated to the corresponding scores
of each statement of Cloud Clicker having medium to large
eect sizes
. This provides further support for the suitability of
the statements and visualizations used in Cloud Clicker. It can also
be seen that the correlations between the ranked Hexad user type
scores and the score of each statement in the gameful application
are similar to the correlations between the absolute Hexad user type
scores calculated using the Hexad questionnaire and the ranked
Hexad user type scores concerning both strength and direction of
the correlations.
Thus, we formulate
R4: The correlations between the scores
of the statements of Cloud Clicker and the ranked Hexad
scores are similar to the correlations between the absolute
Hexad user type scores and the ranked Hexad user type scores
.
R4
is reected visually by the two highlighted diagonals in Table 2.
In line with the results of the CCA, this indicates that assessing the
ranking of Hexad user types with Cloud Clicker is comparable to as-
sessing the ranking of Hexad user types based on the questionnaire.
Consequently, taking
R1–R4
together, our results demonstrate that
Cloud Clicker explains a substantial amount of shared variance be-
tween the predictor and criterion variable sets, that the statements
and visualizations used in Cloud Clicker successfully represent
their corresponding Hexad user types and that Cloud Clicker can
be used to assess the ranking of a user’s Hexad type scores.
5.2 Snowball Shooter and Hexad User Types
Again, a CCA was conducted to investigate the shared variance
between the amount of interactions with each gamication ele-
ment in Snowball Shooter as predictor variables and the Hexad
scores of each user type. The analysis yielded ve canonical func-
tions with squared canonical correlations of .20, .12, .07, .01, and
.00. The full model was statistically signicant (Wilks’s
𝜆=.
659
criterion,
𝐹(
30
,
546
.
00
)=
2
.
00
, 𝑝 =.
001
)
. This leads to result
R5:
The amount of interactions with gamication elements in
Snowball Shooter is moderately associated to the scores of
the validated Hexad user type questionnaire.
Similar to Cloud
Clicker, this means that the amount of interactions with gamica-
tion elements and the Hexad user types are related. However, the
variance shared between the two sets of variables was considerably
HexArcade: Predicting Hexad User Types By Using Gameful Applications CHI PLAY ’20, November 2–4, 2020, Virtual Event, Canada
CF 1 CF1
Predictor co rs Criterion co rs
Collectibles .42 -.05 Hex_DI .37 .37
Achievements .20 .01 Hex_FS -.35 .10
Points .37 .06 Hex_AC .62 .69
Leaderboard 1.08 .89 Hex_PL .51 .79
Virtual Character -.06 -.72 Hex_PH -.40 .09
Hex_SO .22 .46
Table 3: Structure coecients (rs) and standardized canon-
ical function coecients (co) for predictor variables (num-
ber of interactions with gamication elements in Snowball
Shooter) and criterion variables (score of each user type:
Hex_DI etc.) for CF1. Bold entries represent loads higher
than |.35|, underlined ones higher than |.50|.
lower than in Cloud Clicker as the model of Snowball Shooter ac-
counts for 34.1% of the shared variance. This indicates a moderate
eect size [
9
]. As part of the dimension reduction analysis, it was
found that solely CF1 was explaining a signicant amount of vari-
ance. Therefore, only CF1 was considered for the interpretation of
the canonical correlation analysis. An overview of the structure
coecients and standardized canonical function coecients for
CF1 can be found in Table 3.
Looking at the CF1 coecients, it can be seen that the score
in the gamication elements Leaderboard and Virtual Character
strongly contribute to the synthetic predictor variable. While the
score in Leaderboard contributes positively to CF1, Virtual Charac-
ter contributes negatively. Regarding the criterion variable set in
CF1, Achiever and Player were the primary contributors to the crite-
rion synthetic variable, with secondary contributions by Socialiser
and Disruptor. All of the aforementioned variables add positively
to CF1. Thus, the amount of interactions with the Leaderboard is
positively related to the score in the Achiever, Player, Disruptor and
Socialiser factors of the Hexad. This is in line with previous results
based on self-reported preferences for gamication elements [
44
].
That the Virtual Character is negatively contributing to CF1 indi-
cates that participants interacting with it were likely not interested
in interacting with Leaderboards and tended to score lower on the
Player, Achiever, Socialiser and Disruptor types.
Analyzing the relationships between interaction with gamica-
tion elements and Hexad user types further, we again calculated
bivariate Spearman’s rank correlations. Similar to Cloud Clicker,
we considered the absolute score in each Hexad user type as well as
the ranked Hexad user type. Table 4 shows signicant correlation
coecients.
DI AC PL SO PH_R
Collectibles
Achievements .19*
Points
Leaderboard .19* .30** .33** .23** -.23**
Virtual Character -.25** -.22** .19*
Table 4: Spearman’s rank correlation coecients between
the amount of interactions with each gamication element
in the Snowball Shooter application and the absolute Hexad
user type scores (DI etc.) as well as the ranked Hexad scores
(PH_R). Hexad user types having at least one signicant cor-
relation are shown. *𝑝<.05, **𝑝<.01
Overall, the correlations support the results from the canonical
correlation analysis. It can be seen that most correlations were
found for the Leaderboard and Virtual Character gamication ele-
ments. The positive correlation between the ranked Philanthropist
score and the amount of interactions with the Virtual Character
gamication element indicates that, as expected based on the def-
inition of the user type [
44
], the Virtual Character seems to be
particularly relevant for the Philanthropist. Also, the positive cor-
relations between the Achiever, Player, Disruptor and Socialiser
user types and the Leaderboard are in line with the results from the
canonical correlation analysis and were expected based on previous
work [
44
]. In addition, the positive correlation between Achieve-
ments and the Achiever was expected and is in line with previous
ndings [44].
Taking these results all into account together, we establish
R6:
The amount of interactions with gamication elements cor-
relates to their corresponding Hexad user types
. On a more
abstract level,
R5
and
R6
mean that users interact with gamica-
tion elements that correspond to their Hexad user types. This is an
important result for the validity of the Hexad model, as previous
research did not consider actual user behaviour within gameful
applications, as far as we know. However, it should be noted that
we could not nd correlations between Collectibles and Free Spirits
as well as between Points and Players.
5.3 Perception of Cloud Clicker and Snowball
Shooter
To analyze the perception of the gameful applications compared
to completing the Hexad user types questionnaire, a repeated mea-
sures Friedman ANOVA was calculated for the IMI and PXI factors
(the responses in the IMI and PXI responses were not normally
distributed). The Durbin-Conover method was used for post-hoc
analysis and the Benjamini-Hochberg false discovery rate [4] was
used to adjust signicance values for multiple comparisons. For
this analysis, all participants who decided to not answer either the
PXI or IMI questions after the Hexad questionnaire, Cloud Clicker,
or Snowball Shooter, were excluded. Thus, the responses of 113
participants were considered.
Table 5 provides an overview of the mean and standard devia-
tions of the IMI subscales and the PXI Immersion subscale in each
condition (Hexad questionnaire, Cloud Clicker, Snowball Shooter).
When analyzing the Competence subscale of the IMI, we did not nd
signicant dierences between the conditions (
𝜒2(
2
)=
5
.
54
, 𝑝 =
.
063). Similarly, no signicant dierences were found for the Choice
subscale of the IMI (
𝜒2(
2
)=
4
.
94
, 𝑝 =.
085). However, the Enjoy-
ment score diered signicantly (
𝜒2(
2
)=
35
.
8
, 𝑝 <.
001). The
post-hoc analysis revealed that both gameful applications were
signicantly more enjoyable than completing the Hexad question-
naire (
𝑝<.
001 each). Also, Snowball Shooter was perceived as
more enjoyable than Cloud Clicker (𝑝=0.012).
We summarize these eects by
R7: Both gameful applications
are perceived as more enjoyable than completing the Hexad
questionnaire
. This result indicates that gameful approaches might
be more suitable to be used within gameful systems than the Hexad
questionnaire, when a gameful experience is important. In addition,
we found that the perceived pressure diered signicantly between
CHI PLAY ’20, November 2–4, 2020, Virtual Event, Canada Altmeyer et al.
Hexad G1 G2
IMI Competence 5.00 /1.05 5.08 / 1.13 4.84 / 1.32
IMI Choice 5.29 / 1.47 5.29 / 1.37 5.48 / 1.36
IMI Enjoyment* 3.87 / 1.34 4.56 / 1.45 4.81 / 1.62
IMI Pressure* 1.96 / 0.98 2.44 / 1.30 2.34 / 1.27
IMI Immersion* 4.91 / 1.10 5.24 / 1.21 5.40 / 1.23
Table 5: Mean / Standard Deviation for each condition of
the study. All variables are measured on 7-point scales. Vari-
ables for which the Friedman ANOVA was signicant are
marked (*). G1=Cloud Clicker, G2=Snowball Shooter.
the conditions (
𝜒2(
2
)=
11
.
7
, 𝑝 =.
003). Both gameful applications
scored higher in the Pressure factor of the IMI (both
𝑝<.
001),
whereas no dierence was found between the gameful applications
themselves (
𝑝=
1
.
00), leading to
R8: The perceived pressure is
signicantly higher in both gameful applications
. This nd-
ing is likely related to the timed and round-based nature of the
gameful applications (i.e., the fact that we used a timer in Cloud
Clicker and 15 rounds of interaction in both gameful applications).
Finally, we analyzed whether the immersion, as measured by the
Immersion subscale of the PXI, diers across the conditions. The
Friedman ANOVA revealed a signicant eect (
𝜒2(
2
)=
26
.
5
, 𝑝 <
.
001). Both gameful applications scored signicantly higher on the
Immersion subscale of the PXI (each p<.001) while there were no
dierences between the two gameful applications (
𝑝=.
127). Thus,
we derive
R9: Both gameful applications were perceived as
more immersive than completing the Hexad questionnaire
.
Taking
R7–R9
together, it seems like the higher pressure is not
perceived negatively but may cause a feeling of higher immersion
leading to a more enjoyable experience [7].
6 DISCUSSION
Our results show the scores of the statements (“predictor variables”)
in the Cloud Clicker application and the Hexad user type scores
(“criterion variables”) are substantially related to each other. They
share 74.37% of their variance (
R1
). We also found the structure
coecients of the predictor variables and the structure coecients
of the criterion variables load on the same canonical functions, with
large eect sizes (R2). In addition, we found there are medium-to-
large correlations between the ranked Hexad scores and the scores
of the statements in Cloud Clicker (
R3
), which were shown to be
comparable to the correlations between the absolute Hexad scores
and the ranked Hexad scores (
R4
). Taking
R1–R4
together, we
conclude that Cloud Clicker can be used to predict Hexad user
types in a gameful way, when an order of user types is sucient
(which is likely the case when personalizing gamication elements
set in a gameful system).
We suggest using the order of scores, since Cloud Clicker uses
a binary choice instead of allowing users to rate their agreement
with statements on an ordinal scale (as was done in the validated
Hexad questionnaire). Ultimately, these results support H1a: The
score of the statements in Cloud Clicker is correlated to the
corresponding Hexad user types and thus may be used to
predict them
. This is explainable because we used statements
which were similar to the Hexad questionnaire items with the
highest factor load.
Regarding the Snowball Shooter application, in which we ana-
lyzed whether the amount of interactions with gamication ele-
ments (“predictor variables”) could be used to predict Hexad user
types (“criterion variables”), we found that in general, there is a
relationship between the predictor and criterion variables (
R5
).
This is an important nding on its own because it shows that
the correlations between the Hexad user types and preferences
for gamication elements—which have been identied based on
self-reports using textual descriptions or storyboards in previous
work [
3
,
44
]—can be replicated based on actual user behaviour. This
nding supports the suitability of the Hexad model to explain user
behaviour in gameful systems. When analyzing correlations be-
tween the amount of interactions with gamication elements and
Hexad user types further, we found correlations that were expected
based on previous work (
R6
). This is in line with the ndings by
Hallifax et al. [10], and supports the validity of the Hexad model.
However, it should be noted that two correlations that were ex-
pected could not be found (between Free Spirits and Unlockables
as well as between Players and Points). A reason might be that
Snowball Shooter did not motivate a specic behaviour (as gameful
systems usually do [
11
]) but rather encouraged users to try out
dierent gamication elements. This would be likely for Free Spir-
its who like to explore [
44
] and thus might explain the absence
of correlations for this user type. The unlocked items could only
be collected and not be used for anything else. This might have
aected the engagement of users negatively. Also, an incentive for
collecting points was missing, which might have been detrimental
to Players’ motivation to collect points. Considering that the shared
variance between predictor and criterion variable sets was mod-
erate (34.11%), we do not recommend deriving Hexad user types
based on interaction behaviour alone in Snowball Shooter. How-
ever, the amount of interaction with gamication elements might
still be a useful factor for dynamic adjustments of gameful systems.
Based on
R5
and
R6
, we consider
H1b: The amount of inter-
actions with gamication elements in Snowball Shooter is
correlated to the corresponding Hexad user types and thus
may be used to predict them
partially supported. Although we
found that there are correlations to gamication elements that
match the corresponding Hexad user types, the amount of shared
variance between the two sets of variables is too low to reliably
predict Hexad user types.
Furthermore, our results show that both gameful applications
were perceived as more enjoyable (
R7
) and more immersive (
R9
)
than the traditional Hexad questionnaire. Based on these results,
H2a: Both applications are perceived as more enjoyable
and
H2d: Both applications are perceived as more immersive
are
supported. We also found a signicant eect on perceived pressure
(
R8
), which might be related to the higher immersion and arguably
a higher sense of ow [
7
,
12
]. Based on this,
H2c: Participants
feel more pressure in both applications is supported.
We conclude that both gameful applications provide a more
pleasurable gameful experience than completing the Hexad ques-
tionnaire.
H2b
is not supported because no eects were found
regarding perceived competence. Contrary to previous work [
36
], it
seems like the gamication elements did not enhance the perceived
competence through feedback as much as expected. A potential
explanation might be that the user interface elements (such as
HexArcade: Predicting Hexad User Types By Using Gameful Applications CHI PLAY ’20, November 2–4, 2020, Virtual Event, Canada
radio buttons) provide visual feedback on their own, potentially
enhancing the perceived competence in the baseline condition.
On a more abstract level, our results show that Hexad user types
can be assessed by using binary choices. These could be easily
adapted to dierent contexts or could even be turned into concrete
choices a player needs to make in a more game-like setting. Also,
we show that interacting with gameful design elements is related
to a user’s Hexad type. This could be used to infer Hexad user types
dynamically when interacting with a gameful system and provides
huge potential for further research.
6.1 Limitations and Future Work
Our work has several limitations. First, transforming the 7-point
Likert scales into binary decisions and considering only one par-
ticularly relevant item per Hexad user type as was done in Cloud
Clicker unavoidably leads to a loss of information. Consequently,
we recommend using Cloud Clicker as a practical tool when person-
alizing gameful systems while ensuring a gameful user experience
and to prevent a loss of immersion. For scientic purposes, we ac-
knowledge that Cloud Clicker cannot replace the validated Hexad
questionnaire [
42
]. Second, we used statements in Cloud Clicker
that were similar to the statements with high factor loads in the
Hexad questionnaire validation study (but not the same). We de-
cided for one statement instead of all four statements. This means
that even though we used statements that had a substantial factor
load, using other statements for the corresponding user types might
lead to dierent results.
Next, regarding the Snowball Shooter application, it did not mo-
tivate a real-life behaviour but allowed users to interact with the
gamication elements. Although this allows users to experience
how certain gamication elements work, their perception might
be dierent when motivating concrete real-life goals in specic
domains such as physical activity. The free exploration of gami-
cation elements within Snowball Shooter could be particularly
appreciated by Free Spirits, which might explain why we could not
nd any correlations for this user type (because they might have
tried several dierent gamication elements).
Also, we decided to randomize the order of the gameful applica-
tions, but not of the Hexad questionnaire. This was done to avoid
detrimental eects of removing gamication on the perception of
the Hexad questionnaire [
11
]. Since the IMI and PXI scales are the
only shared dependent variables across these conditions, potential
ordering eects would not primarily concern the main goal of the
study, i.e.predicting Hexad user types from interaction behaviour.
In addition, we assume the chance of ordering eects is low since
lling out a questionnaire and interacting with gameful applications
can be considered dierent tasks, reducing the chance of practice
eects [
32
]. Nevertheless, the fact that participants always started
by completing the Hexad questionnaire should be considered.
Since participants were asked to answer roughly 100 items in
total, we cannot rule out fatigue eects. However, considering the
number of items and that the duration of the study is within a
maximum length of 20 minutes [
33
], no practically relevant eects
on data quality are to be expected [15].
Last, we acknowledge that, although the design of the applica-
tions is based on previous research, certain decisions are inherently
a matter of interpretation, which might aect the external validity.
Future work should investigate whether dierent game controls
(e.g., allowing for continuous user input) will enhance the accuracy
of using gameful applications to predict Hexad user types. Further
research should be conducted into correlations between actual
user behaviour and Hexad user types to replicate the ndings of
previous research. Also, our ndings should be validated in dierent
domains.
7 CONCLUSION
To reduce the barrier of having to ll out Hexad questionnaires
for user types assessments, we conceptualized and implemented
two gameful applications to predict Hexad user types in a gameful
way. We contribute two main ndings: First, we show that the rst
gameful application, Cloud Clicker, indeed could be used to predict
Hexad user types. The interaction within the gameful application
and the validated Hexad scores are correlated for all user types
and share a substantial amount of variance. Cloud Clicker could
be used to tailor gameful systems without having to complete a
questionnaire.
Second, we showed that the interaction behaviour with gameful
elements correlates to the corresponding Hexad user types. This is
important because previous work only used self-reported measures
based on storyboards or textual descriptions explaining the gameful
elements. Our results highlight that enjoyment and immersion are
signicantly higher in both gameful applications than when lling
out the Hexad questionnaire.
8 ACKNOWLEDGEMENTS
We thank Karina Arrambide, Cayley MacArthur and the anonymous
reviewers for their valuable feedback to improve this paper.
REFERENCES
[1]
Vero Vanden Abeele, Katta Spiel, Lennart Nacke, Daniel Johnson, and Kathrin
Gerling. 2020. Development and Validation of the Player Experience Inventory: A
Scale to Measure Player Experiences at the Level of Functional and Psychosocial
Consequences. International Journal of Human Computer Studies 135 (2020).
https://doi.org/10.1016/j.ijhcs.2019.102370
[2]
Maximilian Altmeyer and Pascal Lessel. 2017. The Importance of Social Relations
for Well-Being Change in Old Age - Do Game Preferences Change As Well?
Proceedings of the Positive Gaming: Workshop on Gamication and Games for
Wellbeing. Amsterdam, The Netherlands (2017).
[3]
Maximilian Altmeyer, Pascal Lessel, Linda Muller, and Antonio Krüger. 2019.
Combining Behavior Change Intentions and User Types to Select Suitable Gami-
cation Elements for Persuasive Fitness Systems. In International Conference on
Persuasive Technology. Springer.
[4]
Yoav Benjamini and Yosef Hochberg. 1995. Controlling the False Discovery Rate:
A Practical and Powerful Approach to Multiple Testing. Journal of the Royal
Statistical Society 58, 4 (1995), 619–678.
[5]
Max V. Birk, Maximilian A. Friehs, and Regan L. Mandryk. 2017. Age-Based
Preferences and Player Experience: A Crowdsourced Cross-sectional Study. Pro-
ceedings of the Annual Symposium on Computer-Human Interaction in Play - CHI
PLAY ’17 (2017), 157–170. https://doi.org/10.1145/3116595.3116608
[6]
Jacob Cohen. 1988. Statistical Power Analysis for the Behavioral Sciences, Hills-
dale. NJ: Lawrence Erlbaum Associates (1988).
[7]
Mihaly Csikszentmihalyi. 1997. Finding Flow: The Psychology of Engagement with
Everyday Life. Basic Books, New York, NY, US. ix, 181–ix, 181 pages.
[8]
Sebastian Deterding, Dan Dixon, Rilla Khaled, and Lennart Nacke. 2011. From
Game Design Elements to Gamefulness: Dening Gamication. Proceedings of
the 15th International Academic MindTrek Conference. ACM, 2011. (2011), 9–15.
https://doi.org/10.1145/2181037.2181040
CHI PLAY ’20, November 2–4, 2020, Virtual Event, Canada Altmeyer et al.
[9]
Christopher J. Ferguson. 2009. An Eect Size Primer: A Guide for Clinicians and
Researchers. Professional Psychology: Research and Practice 40, 5 (2009), 532–538.
https://doi.org/10.1037/a0015808
[10] Stuart Hallifax, Audrey Serna, Jean-charles Marty, Guillaume Lavoué, and Elise
Lavoué. 2019. Factors to Consider for Tailored Gamication. Proceedings of
the Annual Symposium on Computer-Human Interaction in Play - CHI PLAY ’19
(2019).
[11]
Juho Hamari and Harri Sarsa. 2014. Does Gamication Work? - A Literature
Review of Empirical Studies on Gamication. Hawaii International Conference on
System Sciences. (2014), 3025–3034. https://doi.org/10.1109/HICSS.2014.377
[12]
Johannes Harms, Stefan Biegler, Christoph Wimmer, Karin Kappel, and Thomas
Grechenig. 2015. Gamication of Online Surveys: Design Process, Case Study,
and Evaluation. Lecture Notes in Computer Science (including subseries Lecture
Notes in Articial Intelligence and Lecture Notes in Bioinformatics) 9296, Section 4
(2015), 219–236. https://doi.org/10.1007/978-3-319-22701-6_16
[13]
Johannes Harms, Dominik Seitz, Christoph Wimmer, Karin Kappel, and Thomas
Grechenig. 2015. Low-Cost Gamication of Online Surveys: Improving the User
Experience through Achievement Badges. CHI PLAY 2015 - Proceedings of the
2015 Annual Symposium on Computer-Human Interaction in Play (2015), 109–114.
https://doi.org/10.1145/2793107.2793146
[14]
Johannes Harms, Christoph Wimmer, Karin Kappel, and Thomas Grechenig.
2014. Gamication of Online Surveys: Conceptual Foundations and a Design
Process based on the MDA Framework. Proceedings of the NordiCHI 2014: The
8th Nordic Conference on Human-Computer Interaction: Fun, Fast, Foundational
(2014), 565–568. https://doi.org/10.1145/2639189.2639230
[15]
A. Regula Herzog and Jerald G. Bachman. 1981. Eects of Questionnaire Length
on Response Quality. Public Opinion Quarterly 45, 4 (1981), 549–559. https:
//doi.org/10.1086/268687
[16]
Yuan Jia, Bin Xu, YaminiKaranam, and Stephen Voida. 2016. Personality-Targeted
Gamication: A Survey Study on Personality Traits and Motivational Aordances.
Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems -
CHI ’16 (2016), 2001–2013. https://doi.org/10.1145/2858036.2858515
[17]
Florian Keusch and Chan Zhang. 2017. A Review of Issues in Gamied Surveys.
Social Science Computer Review 35, 2 (2017), 147–166. https://doi.org/10.1177/
0894439315608451
[18]
Rex B. Kline. 2011. Principles and Practice of Strctural Equation Modeling.
Vol. 53. 1689–1699 pages. https://doi.org/10.1017/CBO9781107415324.004
arXiv:arXiv:1011.1669v3
[19]
Dimosthenis Kotsopoulos, Cleopatra Bardaki, Stavros Lounis, and Katerina Pra-
matari. 2018. Employee Proles and Preferences towards IoT-enabled Gamica-
tion for Energy Conservation. International Journal of Serious Games 5, 2 (2018),
65–85. https://doi.org/10.17083/ijsg.v5i2.225
[20]
Andrzej Marczewski. 2015. Even Ninja Monkeys Like to Play: Gamication, Game
Thinking and Motivational Design. CreateSpace Independent Publishing Platform.
[21]
Edward D. McAuley, Terry Duncan, and Vance V. Tammen. 1989. Psychometric
Properties of the Intrinsic Motivation Inventoiy in a Competitive Sport Setting:
A Conrmatory Factor Analysis. Research Quarterly for Exercise and Sport 60, 1
(1989), 48–58. https://doi.org/10.1080/02701367.1989.10607413
[22]
Robert R. McCrae and Oliver P. John. 1992. An Introduction to the Five-Factor
Model and Its Applications. Journal of Personality 60, 2 (1992), 175–215. https:
//doi.org/10.1111/j.1467-6494.1992.tb00970.x
[23]
Elisa D. Mekler, Florian Brühlmann, Alexandre N. Tuch, and Klaus Opwis. 2017.
Towards Understanding the Eects of Individual Gamication Elements on In-
trinsic Motivation and Performance. Computers in Human Behavior 71 (2017),
525–534. https://doi.org/10.1016/j.chb.2015.08.048
[24]
Alberto Mora, Gustavo F. Tondello, Lennart E. Nacke, and Joan Arnedo-Moreno.
2018. Eect of Personalized Gameful Design on Student Engagement. IEEE
Global Engineering Education Conference, EDUCON 2018-April (2018), 1925–1933.
https://doi.org/10.1109/EDUCON.2018.8363471
[25]
Lennart E. Nacke, Chris Bateman, and Regan L. Mandryk. 2014. BrainHex: A
Neurobiological Gamer Typology Survey. Entertainment Computing 5, 1 (2014),
55–62. https://doi.org/10.1016/j.entcom.2013.06.002
[26]
Aljoscha C. Neubauer and Bertram F. Malle. 1997. Questionnaire Response
Latencies: Implications for Personality Assessment and Self-Schema Theory.
European Journal of Psychological Assessment 13, 2 (1997), 109–117. https://doi.
org/10.1027/1015-5759.13.2.109
[27]
Kim Nimon, Robin K. Henson, and Michael S. Gates. 2010. Revisiting Inter-
pretation of Canonical Correlation Analysis: A Tutorial and Demonstration of
Canonical Commonality Analysis. Multivariate Behavioral Research 45, 4 (2010),
702–724. https://doi.org/10.1080/00273171.2010.498293
[28]
Rita Orji, Regan L Mandryk, and Julita Vassileva. 2015. Gender, Age, and Respon-
siveness to Cialdini’s Persuasion Strategies. International Conference on Persuasive
Technology 9072, June (2015). https://doi.org/10.1007/978-3- 319-20306-5
[29]
Rita Orji, Lennart E. Nacke, and Chrysanne Di Marco. 2017. Towards Personality-
driven Persuasive Health Games and Gamied Systems. Proceedings of the 2017
CHI Conference on Human Factors in Computing Systems - CHI ’17 (2017), 1015–
1027. https://doi.org/10.1145/3025453.3025577
[30]
Rita Orji, Gustavo F Tondello, and Lennart E Nacke. 2018. Personalizing Persua-
sive Strategies in Gameful Systems to Gamication User Types. Proceedings of
the SIGCHI Conference on Human Factors in Computing Systems - CHI ’18 (2018).
https://doi.org/doi.org/10.1145/3173574.3174009
[31]
Kiemute Oyibo, Rita Orji, and Julita Vassileva. 2017. The Inuence of Culture
in the Eect of Age and Gender on Social Inuence in Persuasive Technology.
Adjunct Publication of the 25th Conference on User Modeling, Adaptation and Per-
sonalization - UMAP ’17 (2017), 47–52. https://doi.org/10.1145/3099023.3099071
[32]
Paul Christopher Price, I-Chant A. Chiang, and Rajiv Jhangiani. 2015. Research
Methods in Psychology. BCcampus 40, 5 (2015), 551–552. https://openlibrary-
repo.ecampusontario.ca/jspui/handle/123456789/246
[33]
Melanie Revilla and Carlos Ochoa. 2017. Ideal and Maximum Length for a
Web Survey. International Journal of Market Research 59, 5 (2017), 557–565.
https://doi.org/10.2501/IJMR-2017- 039
[34]
Richard M. Ryan. 1982. Control and Information in the Intrapersonal Sphere:
An Extension of Cognitive Evaluation Theory. Journal of Personality and Social
Psychology 43, 3 (1982), 450–461.
[35]
Richard M Ryan and Edward L Deci. 2000. Self-Determination Theory and
the Facilitation of Intrinsic Motivation, Social Development, and Well-Being.
American Psychologist 55, 1 (2000), 68–78. https://doi.org/10.1037/0003- 066X.55.
1.68
[36]
Michael Sailer, Jan Ulrich Hense, Sarah Katharina Mayr, and Heinz Mandl. 2017.
How Gamication Motivates: An Experimental Study of the Eects of Spe-
cic Game Design Elements on Psychological Need Satisfaction. Computers
in Human Behavior 69 (2017), 371–380. https://doi.org/10.1016/j.chb.2016.12.033
arXiv:arXiv:1011.1669v3
[37]
Katie Seaborn and Deborah Fels. 2015. Gamication in Theory and Action:
A Survey. International Journal of Human-Computer Studies 74 (2015), 14–31.
https://doi.org/10.1016/j.ijhcs.2014.09.006
[38]
Alissa Sherry and Robin K. Henson. 2005. Conducting and Interpreting Canonical
Correlation Analysis in Personality Research: A User-Friendly Primer. Jour-
nal of Personality Assessment 3891 (2005), 37–48. https://doi.org/10.1207/
s15327752jpa8401_09
[39]
Barabara G. Tabachnick and Linda S. Fidell. 2013. Using Multivariate Statistics
(sixth edit ed.). Pearson. 1–1018 pages.
[40]
Bruce Thompson. 1984. Canonical Correlation Analysis: Uses and Interpretation.
Number 47. Sage.
[41]
Gustavo Fortes Tondello. 2019. Dynamic Personalization of Gameful Interactive
Systems. Ph.D. Dissertation. University of Waterloo. http://hdl.handle.net/10012/
14807
[42]
Gustavo F. Tondello, Alberto Mora, Andrzej Marczewski, and Lennart E. Nacke.
2018. Empirical Validation of the Gamication User Types Hexad Scale in English
and Spanish. International Journal of Human-Computer Studies (2018). https:
//doi.org/10.1016/j.ijhcs.2018.10.002
[43]
Gustavo F. Tondello, Alberto Mora, and Lennart E. Nacke. 2017. Elements of
Gameful Design Emerging from User Preferences. Proceedings of the Annual
Symposium on Computer-Human Interaction in Play - CHI PLAY ’17 (2017), 129–
142. https://doi.org/10.1145/3116595.3116627
[44]
Gustavo F Tondello, Rina R Wehbe, Lisa Diamond, Marc Busch, Andrzej Mar-
czewski, and Lennart E Nacke. 2016. The Gamication User Types Hexad Scale.
The ACM SIGCHI Annual Symposium on Computer-Human Interaction in Play -
CHI PLAY ’16 (2016). https://doi.org/10.1145/2967934.2968082
[45]
Tamilla Triantoro, Ram Gopal, Raquel Benbunan-Fich, and Guido Lang. 2019.
Would You Like to Play? A Comparison of a Gamied Survey with a Traditional
Online Survey Method. International Journal of Information Management 49,
April (2019), 242–252. https://doi.org/10.1016/j.ijinfomgt.2019.06.001
[46]
Hao-Ting Wang, Jonathan Smallwood, Janaina Mourao-Miranda, Cedric Huchuan
Xia, Theodore D. Satterthwaite, Danielle S. Bassett, and Danilo Bzdok. 2018.
Finding the Needle in High-Dimensional Haystack: A Tutorial on Canonical
Correlation Analysis. arXiv preprint arXiv:1812.02598 (2018). arXiv:1812.02598
http://arxiv.org/abs/1812.02598
[47]
Carmen Ximénez. 2007. Eect of variable and subject sampling on recovery of
weak factors in CFA. Methodology 3, 2 (2007), 67–80. https://doi.org/10.1027/1614-
2241.3.2.67
... Whereas, a dual personalization approach resulted in positive motivational outcomes in [67], providing a valuable contribution on the potential of considering more than one personalization criterion. From such context, we expect multidimensional personalization 2 to approximate the positive effects of customization (e.g., [42,66]) because users are not reduced to a single dimension; while avoiding the burden of asking their game element preferences [2,54]. However, the only study (to our best knowledge) applying multidimensional personalization of gamification with users did not compare it to the OSFA approach [67]. ...
... First, they differ in context (i.e., learning activities [46,48] and social networks [23]), which might be a reason for the contradictory findings. Second, personalizing based on interaction instead of survey preferences might be another one [2,54]. Consequently, Hajarian et al. [23] personalized based on data from the same context and activity, while others [46,48] implemented recommendations from general preferences. ...
... Accordingly, one cannot ensure whether someone has experienced that motivation level or if they carelessly selected that option due to, e.g., hypothesis-guessing. Therefore, we chose not to systematically inspect or remove data based on those patterns following similar studies [2,40,46,66,67,74] while we complemented our dataset with qualitative data from semi-structured interviews to enrich our findings [15]. Second, our sample size is limited to 45 data points, which is associated with low statistical power and often large confidence intervals. ...
Article
Personalization of gamification is an alternative to overcome the shortcomings of the one-size-fits-all approach, but the few empirical studies analyzing its effects do not provide conclusive results. While many user and contextual information affect gamified experiences, prior personalized gamification research focused on a single user characteristic/dimension. Therefore, we hypothesize if a multidimensional approach for personalized gamification, considering multiple (user and contextual) information, can improve user motivation when compared to the traditional implementation of gamification. In this paper, we test that hypothesis through a mixed-methods sequential explanatory study. First, 26 participants completed two assessments using one of the two gamification designs and self-reported their motivations through the Situational Motivation Scale. Then, we conducted semi-structured interviews to understand learners' subjective experiences during these assessments. As result, the students using the personalized design were more motivated than those using the one-size-fits-all approach regarding intrinsic motivation and identified regulation. Furthermore, we found the personalized design featured game elements suitable to users' preferences, being perceived as motivating and need-supporting. Thus, informing i) practitioners on the use of a strategy for personalizing gamified educational systems that is likely to improve students' motivations, compared to OSFA gamification, and ii) researchers on the potential of multidimensional personalization to improve single-dimension strategies. For transparency, dataset and analysis procedures are available at https://osf.io/grzhp.
... Altmeyer et al. [45] conducted a study with 147 participants to analyze the potential of two gameful applications to predict the Hexad user types. In one application, the respondents were asked to select one statement and then they received gameful feedback. ...
... Year Player/User Typology AUP AUS Hallifax et al. [30] 2019 BrainHex and Hexad • Lopez et al. [42] 2019 Hexad • Tondello et al. [19] 2019 Hexad • Altmeyer et al. [45] 2020 Hexad • Busch et al. [36] 2016 BrainHex • • Our study 2020 Hexad • • Key: AUP: Accessed the user type of the participants; AUS: Analyzed if the player/user type is stable over time. ...
... Since each Hexad sub-scale is formed by four questions arranged in a 7-point Likert Scale, the maximum value a Hexad sub-scale can be is 28. Similar to other studies that accessed the user type through the Hexad scale [38,19,45], the Philanthropists and Achievers presented the higher average score while the Disruptors presented the lower average score. After testing the normality of the data using the Shapiro-Wilk test, we measured the bivariate correlation coefficients using Kendall's , since the user type scores were nonparametric. ...
Conference Paper
Full-text available
In recent years, different studies have proposed and validated user models (e.g., Bartle, BrainHex, and Hexad) to represent the different user profiles in games and gamified settings. However, the results of applying these user models in practice (e.g., to personalize gamified systems) are still contradictory. One of the hypotheses for these results is that the user types can change over time (i.e., user types are dynamic). To start to understand whether user types can change over time, we conducted an exploratory study analyzing data from 74 participants to identify if their user type (Achiever, Philanthropist, Socialiser, Free Spirit, Player, and Disruptor) had changed over time (six months). The results indicate that there is a change in the dominant user type of the participants, as well as the average scores in the Hexad sub-scales. These results imply that all the scores should be considered when defining the Hexad's user type and that the user types are dynamic. Our results contribute with practical implications, indicating that the personalization currently made (generally static) may be insufficient to improve the users' experience, requiring user types to be analyzed continuously and personalization to be done dynamically.
... Altmeyer et al. [45] conducted a study with 147 participants to analyze the potential of two gameful applications to predict the Hexad user types. In one application, the respondents were asked to select one statement and then they received gameful feedback. ...
... Year Player/User Typology AUP AUS Hallifax et al. [30] 2019 BrainHex and Hexad • Lopez et al. [42] 2019 Hexad • Tondello et al. [19] 2019 Hexad • Altmeyer et al. [45] 2020 Hexad • Busch et al. [36] 2016 BrainHex • • Our study 2020 Hexad • • Key: AUP: Accessed the user type of the participants; AUS: Analyzed if the player/user type is stable over time. ...
... Since each Hexad sub-scale is formed by four questions arranged in a 7-point Likert Scale, the maximum value a Hexad sub-scale can be is 28. Similar to other studies that accessed the user type through the Hexad scale [38,19,45], the Philanthropists and Achievers presented the higher average score while the Disruptors presented the lower average score. After testing the normality of the data using the Shapiro-Wilk test, we measured the bivariate correlation coefficients using Kendall's , since the user type scores were nonparametric. ...
Preprint
Full-text available
In recent years, different studies have proposed and validated user models (e.g., Bartle, BrainHex, and Hexad) to represent the different user profiles in games and gamified settings. However, the results of applying these user models in practice (e.g., to personalize gamified systems) are still contradictory. One of the hypotheses for these results is that the user types can change over time (i.e., user types are dynamic). To start to understand whether user types can change over time, we conducted an exploratory study analyzing data from 74 participants to identify if their user type (Achiever, Philanthropist, Socialiser, Free Spirit, Player, and Disruptor) had changed over time (six months). The results indicate that there is a change in the dominant user type of the participants, as well as the average scores in the Hexad sub-scales. These results imply that all the scores should be considered when defining the Hexad's user type and that the user types are dynamic. Our results contribute with practical implications, indicating that the personalization currently made (generally static) may be insufficient to improve the users' experience, requiring user types to be analyzed continuously and personalization to be done dynamically.
... As reported in supplementary Table S2, the higher average scores were from Philanthropists, Achievers, and Free Spirits, and the lower average scores were from Disruptors. These values are similar to other recent studies about the Hexad user types 9,14,17,39 . We also calculated the dominant user types (i.e. the strongest tendency of the respondents 10,40 ), which distribution results were: Philanthropist = 34%, Achiever = 30%, Free Spirit = 13%, Player = 12%, Socialiser = 11%, and Disruptor = 1%. ...
... Considering the distribution of the scores, our results are similar to prior research 9,14,17,39 , demonstrating that Philanthropists, Achievers, and Free Spirits are the strongest tendencies of the users regarding the Hexad user types, while Disruptor is the lower tendency. We also calculated the dominant user types, indicating that Achiever and Philanthropist were responsible for more than 60% of the dominant user types of the respondents. ...
Article
Full-text available
Gamification has become a significant direction in designing technologies, services, products, organizational structures, and any human activities towards being more game-like and consequently being more engaging and motivating. Albeit its success, research indicates that personal differences exist with regards to susceptibility to gamification at large as well as to different types of gamification designs. As a response, models and measurement instruments of user types when it comes to gamification have been developed. One of the most discussed related instruments is the Hexad user types scale. However, there has been paucity of research related to the validity and reliability of the Hexad instrument in general but also of its different formulations and language versions. To face this gap, our study focused on analyzing the psychometric properties of the Hexad scale in Brazilian Portuguese by conducting two confirmatory factor analyses and two multi-group confirmatory factor analyses. The survey was answered by 421 Brazilian respondents (52% self-reported women, 47% self-reported men, 0.5% preferred not to provide their gender, and 0.5% checked the option “other”), from the five Brazilian regions (23 different states and the Federal District), and aged between 10 and 60 years old. Findings support the structural validity of the scale as an oblique model and indicate opportunities for small improvements. Further research, both at academy and practice, may use this study as the source of measurement of user types related to gamification (in Brazilian Portuguese), as well as, as a theoretical and practical source for further studies discussing personalized gamification.
... The decision of the final player type-and thus the most appropriate game element-usually relies on the predominant rating [4] or a combination of them [22]. As an alternative to questionnaires, a recent study [23] proposed to predict HEXAD player types through gameful applications. The authors created two applications. ...
... Our adaptive method needs information about the player type of the user at the beginning of the experience, so we used the HEXAD player types questionnaire proposed by [20]. It is worth noting that a recent study that describes a new method to predict player types using gameful applications [23]. As we mentioned in Section 3, we selected 14 game elements of the 52 presented in [24]. ...
Article
Full-text available
The design of gamified experiences following the one-fits-all approach uses the same game elements for all users participating in the experience. The alternative is adaptive gamification, which considers that users have different playing motivations. Some adaptive approaches use a (static) player profile gathered at the beginning of the experience; thus, the user experience fits this player profile uncovered through the use of a player type questionnaire. This paper presents a dynamic adaptive method which takes players’ profiles as initial information and also considers how these profiles change over time based on users’ interactions and opinions. Then, the users are provided with a personalized experience through the use of game elements that correspond to their dynamic playing profile. We describe a case study in the educational context, a course integrated on Nanomoocs, a massive open online course (MOOC) platform. We also present a preliminary evaluation of the approach by means of a simulator with bots that yields promising results when compared to baseline methods. The bots simulate different types of users, not so much to evaluate the effects of gamification (i.e., the completion rate), but to validate the convergence and validity of our method. The results show that our method achieves a low error considering both situations: when the user accurately (Err = 0.0070) and inaccurately (Err = 0.0243) answers the player type questionnaire.
... In 2013, Andrzej Marczewski developed the Hexad gamification user type model [5,23] for gamified systems. Currently, it has become one of the most used models to determine gamification user typologies [2,4]. The Hexad model consists of six user types and their associated motivations based on [37] differing in the degree to which they are driven by their needs for autonomy, competence, purpose and relatedness (as defined by the Self-Determination Theory (SDT) [32]): ...
... They used social network properties, the text of their status updates and their frequencies and time of posting. Altmeyer et al. [4] developed a gameful application named "Cloud Clicker" to predict Hexad user types. They found that the interaction with the game elements of "Cloud Clicker" and the validated Hexad scores of the application's users are correlated for all user types and share a substantial amount of variance. ...
Conference Paper
The Hexad user types model is often used in the gamification community to tailor gamified systems. However, most often, it requires users to fill out a questionnaire, preventing an automated adaptation of the interactive system. For this reason, we explored the potential of using mobile banking data to automate the profiling of Hexad user types. In our study, we conducted an expert consensus study to research whether a group of experts (N=11) in the field of gamification and banking perceive there is a relation between the Hexad user types and banking data. The results show that experts find this relation present, indicating that automating the calculation of Hexad user types from banking data could be feasible.
... years, SD = 8.27 years). Participants' average scores in the Hexad user types followed a similar distribution as reported in previous studies [5,[57][58][59]. ...
Article
The perception of gamification elements differs across users, which is why personalizing gamified systems is important. Past research showed that the Hexad user types model is particularly suitable for this purpose by demonstrating correlations between user types and gamification elements. However, previous studies were mostly survey-based, i.e. relied on participants' rating of gamification elements based on e.g. textual descriptions or storyboards. Thus, the question whether personalization based on Hexad user types provides benefits in implemented gameful systems was neglected. We contribute to this by investigating the effects of (contra-) tailoring the set of gamification elements to a user's Hexad type on performance and user experience, assessed with survey and physiological measures, in an image tagging context. In a lab study (N=29), we found that gamification increases performance and affects psychophysiological measures of arousal. Moreover, we demonstrate that personalization increases enjoyment, positively-valenced affective experiences and participants' absorption in the task at hand.
Conference Paper
Full-text available
Gamification is widely used to foster user motivation. Recent studies show that users can be more or less receptive to different game elements, based on their personality or player profile. Consequently, recent work on tailored gamification tries to identify links between user types and motivating game elements. However findings are very heterogeneous due to different contexts, different typologies to characterize users, and different implementations of game elements. Our work seeks to obtain more generalizable findings in order to identify the main factors that will support design choices when tailoring gamification to users' profiles and provide designers with concrete recommendations for designing tailored gamification systems. For this purpose, we ran a crowdsourced study with 300 participants to identify the motivational impact of game elements. Our study differs from previous work in three ways: first, it is independent from a specific user activity and domain; second, it considers three user typologies; and third, it clearly distinguishes motivational strategies and their implementation using multiple different game elements. Our results reveal that (1) different implementations of a same motivational strategy have different impacts on motivation, (2) dominant user type is not sufficient to differentiate users according to their preferences for game elements, (3) Hexad is the most appropriate user typology for tailored gamification and (4) the motivational impact of certain game elements varies with the user activity or the domain of gamified systems.
Thesis
Full-text available
Gameful design, the process of creating a system with affordances for gameful experiences, can be used to increase user engagement and enjoyment of digital interactive systems. It can also be used to create applications for behaviour change in areas such as health, wellness, education, customer loyalty, and employee management. However, existing research suggests that the qualities of users, such as their personality traits, preferences, or identification with a task, can influence gamification outcomes. It is important to understand how to personalize gameful systems, given how user qualities shape the gameful experience. Current evidence suggests that personalized gameful systems can lead to increased user engagement and be more effective in helping users achieve their goals than generic ones. However, to create these kinds of systems, designers need a specific method to guide them in personalizing the gameful experience to their target audience. To address this need, this thesis proposes a novel method for personalized gameful design divided into three steps: (1) classification of user preferences, (2) classification and selection of gameful design elements, and (3) heuristic evaluation of the design. Regarding the classification of user preferences, this thesis evaluates and validates the Hexad Gamification User Types Scale, which scores a person in six user types: philanthropist, socialiser, free spirit, achiever, player, and disruptor. Results show that the scale’s structural validity is acceptable for gamification studies through reliability analysis and factor analysis. For classification and selection of gameful design elements, this thesis presents a conceptual framework based on participants’ self-reported preferences, which classifies elements in eight groups organized into three categories: individual motivations (immersion and progression), external motivations (risk/reward, customization, and incentives), and social motivations (socialization, altruism, and assistance). And to evaluate the design of gameful applications, this thesis introduces a set of 28 gameful design heuristics, which are based on motivational theories and gameful design methods and enable user experience professionals to conduct a heuristic evaluation of a gameful application. Furthermore, this thesis describes the design, implementation, and pilot evaluation of a software platform for the study of personalized gameful design. It integrates nine gameful design elements built around a main instrumental task, enabling researchers to observe and study the gameful experience of participants. The platform is flexible so the instrumental task can be changed, game elements can be added or removed, and the level and type of personalization or customization can be controlled. This allows researchers to generate different experimental conditions to study a broad range of research questions. Our personalized gameful design method provides practical tools and clear guidelines to help designers effectively build personalized gameful systems.
Article
Full-text available
Gamification can be used to effect positive behavioral change in various fields, including energy conservation. This paper reports on a survey we conducted to formulate a holistic view of users’ profiles and preferences in an Internet of Things (IoT)-enabled gamified mobile application that provides personalized energy-saving recommendation tips to employees, towards conserving energy and adopting a more green behavior at the workplace. The collected insight dictates that a gamified app promoting energy saving at the workplace may become a daily habit for its users if it at least includes three game design elements: progression, levels and points. Additionally, we complement existing design guidelines about the requirements of each HEXAD gamification user type (Philanthropist, Socializer, Free Spirit, Achiever, Disruptor, Player) in game elements, specifically for energy efficiency applications. Thus, we apply and validate the HEXAD gamification user typology in a new context – multicultural workplaces – as well as domain – energy-saving. Moreover, the collected insight inspired us to devise a modular, rule-based mechanism for formulating personalized energy-saving recommendation tips tailored to the users’ profiles and game design choices. This research may assist researchers, as well as practitioners, in designing personalized gamified behavioral interventions, especially towards energy conservation in workplaces.
Conference Paper
Full-text available
Many recent studies of gamification applied to higher education have demonstrated a wide range of positive results. However, most of them fail to consider any personalization factor for the student experience, despite recent studies having shown that gameful systems may be more engaging when they are personalized to each user. Therefore, the goal of this work is to investigate if gameful learning experiences can better motivate and engage students if they are personalized. In this way, we present the design and analysis of a personalized gameful learning experience within a Computer Network Design course. The general purpose of this study is to determine whether a personalized gameful learning experience affects both the students' behavioral and emotional engagement. The results of a descriptive analysis reveal that personalization works better than generic approaches in all items regarding the behavioral and emotional engagement of the students, being a promising standpoint to further investigate in subsequent studies.
Conference Paper
Full-text available
Persuasive gameful systems are effective tools for motivating behaviour change. Research has shown that tailoring these systems to individuals can increase their efficacy; however, there is little knowledge on how to personalize them. We conducted a large-scale study of 543 participants to investigate how different gamification user types responded to ten persuasive strategies depicted in storyboards representing persuasive gameful health systems. Our results reveal that people's gamification user types play significant roles in the perceived persuasiveness of different strategies. People scoring high in the 'player' user type tend to be motivated by competition, comparison, cooperation, and reward while 'disruptors' are likely to be demotivated by punishment, goal-setting, simulation, and self-monitoring. 'Socialisers' could be motivated using any of the strategies; they are the most responsive to persuasion overall. Finally, we contribute to CHI research and practice by offering design guidelines for tailoring persuasive gameful systems to each gamification user type.
Article
Using the stimulus-organism-response (S-O-R) framework and signaling theory, we evaluated the signaling effect of gamification in online survey systems. Based on the Big Five personality assessment instrument, we developed an experimental study with two surveys – a traditional online survey with Likert scales, and a gamified survey powered by game mechanics. Then we evaluated the effect of both surveys on the users’ cognitive and affective reactions, as well as their preference toward the signaler. We also identified game elements that influence the individuals’ reactions when interacting with gamified surveys. The results suggest that gamification serves as a positive signal and increases affective reactions. These findings have theoretical and practical implications to improve the design of existing online surveys.
Conference Paper
Gamified systems may help older adults to remain physically, cognitively and socially active, which has positive effects on well-being. However, social aspects of psychological well-being change during life course, i.e., the importance of positive social relationships for well-being increases between young or middle aged persons and seniors. In this paper, we explore whether these changes are reflected in the game preferences of seniors aged 75 and older. We report findings of a semi-structured interview and a preliminary player classification survey (N=18, mean age=84.61). We found indications that there are differences in game preferences and the perception of game elements that are related to the increased importance of social relationships for well-being in older ages.
Conference Paper
The motivational impact of gamification elements differs substantially across users. To account for these differences, we investigate Hexad user types and behavior change intentions as factors to personalize gamifed, persuasive fitness systems. We conducted an online study (N = 179), measuring the perceived persuasiveness of twelve gamification elements using storyboards. Results show the applicability of the Hexad user type in the Physical Activity domain. Besides replicating correlations between gamification elements and user types, we also found correlations which were hypothesized in literature, but not yet shown. Our main contribution is to show that behavior change intentions influence the perception of gamification elements in general and affect the set of relevant elements for each user type. Since a static set of elements has been suggested for each user type so far, this is an important finding, leading to potentially more effective personalization approaches.
Article
The common approach to the multiplicity problem calls for controlling the familywise error rate (FWER). This approach, though, has faults, and we point out a few. A different approach to problems of multiple significance testing is presented. It calls for controlling the expected proportion of falsely rejected hypotheses — the false discovery rate. This error rate is equivalent to the FWER when all hypotheses are true but is smaller otherwise. Therefore, in problems where the control of the false discovery rate rather than that of the FWER is desired, there is potential for a gain in power. A simple sequential Bonferronitype procedure is proved to control the false discovery rate for independent test statistics, and a simulation study shows that the gain in power is substantial. The use of the new procedure and the appropriateness of the criterion are illustrated with examples.
Article
Gamification, the use of game elements in non-game systems, is now established as a relevant research field in human-computer interaction (HCI). Several empirical studies have shown that gameful interventions can increase engagement and generate desired behavioural outcomes in HCI applications. However, some inconclusive results indicate that we need a fuller understanding of the mechanisms and effects of gamification. The Gamification User Types Hexad scale allows us to parse different user motivations in participants’ interactions with gameful applications, which are measured using a self-report questionnaire. Each user type represents a style of interaction with gameful applications, for example, if the interactions are more focused on achievements, socialization, or rewards. Thus, by scoring an individual in each one of the user types of the Hexad model, we can establish a profile of user preferences for gameful interactions. However, we still lack a substantial empirical validation of this scale. Therefore, we set out to validate the factor structure of the scale, in both English and Spanish, by conducting three studies, which also investigated the distribution of the Hexad's user types in the sample. Our findings support the structural validity of the scale, as well as suggesting opportunities for improvement. Furthermore, we demonstrate that some user types are more common than others and that gender and age correlate with a person's user types. Our work contributes to HCI research by further validating the utility of the Gamification User Types Hexad scale, potentially affording researchers a deeper understanding of the mechanisms and effects of gameful interventions.