Clustering on Player Types of Students in Health Science - Trial and Data Analyses

Abstract

Gamification has many positive effects, such as increased motivation, engagement, and well-being of users. For this purpose, a wide field of game mechanics is already available that can be used in teaching. For the development of gamified teaching methods, it's important to adapt the mechanics used to the students. There are different models that divide target groups of games and gamification into player types to understand what motivates the respective users. This paper describes a study of player types among students of health-related disciplines and analyses the data by a K-Means clustering procedure. The player types Socializer, Player and Achiever are found, and game elements for this groups are suggested. Thus, in the field of health education, game mechanics can be used, which are suitable for students of this domain.

Full text

Clustering on Player Types of Students in
Health Science – Trial and Data Analyses
Lea C BRANDLa,1 and Andreas SCHRADERa
aInstitute of Telematics, University of Lübeck, Lübeck, Germany
Abstract. Gamification has many positive effects, such as increased motivation,
engagement, and well-being of users. For this purpose, a wide field of game
mechanics is already available that can be used in teaching. For the development of
gamified teaching methods, it’s important to adapt the mechanics used to the
students. There are different models that divide target groups of games and
gamification into player types to understand what motivates the respective users.
This paper describes a study of player types among students of health-related
disciplines and analyses the data by a K-Means clustering procedure. The player
types Socializer, Player and Achiever are found, and game elements for this groups
are suggested. Thus, in the field of health education, game mechanics can be used,
which are suitable for students of this domain.
Keywords. Gamification, Player type, Health education
1. Introduction
During their studies, students are often confronted with the challenge of acquiring
extensive knowledge in a short period of time. Miller's pyramid describes four levels of
competence learning. To reach the top level "Does" of competence, the lowest level
"Knows" must be fulfilled first [1].
For example, in medical studies, extensive factual knowledge is therefore already
required in the first semesters. The M1 exam includes 320 questions on medical factual
knowledge and is administered after the 4th semester [2]. However, the way in which
this basic knowledge is taught at universities is ineffective for storing the knowledge in
the brain. According to Dale's cone of experience, learners can only remember 50% of
the information that is conveyed visually and auditorily in a lecture [3].
As a result, lecture content should either be conveyed by other means or repeated by
the students. From a motivational and learning psychology point of view, gamification
can be used for this purpose. Gamification has many positive effects, such as increased
motivation, engagement, and well-being of users [4–6]. For this purpose, a wide field of
game mechanics is already available that can be used in teaching [7, 8].
Despite the already prevalent use of gamification in teaching, there is a lack of
scientific methodology in the development of gamified teaching methods, as well as the
evaluation of them [5]. For the development of gamified teaching methods, it’s important
to adapt the mechanics used to the students [5]. There are different models that divide
1
Corresponding Author: Lea Brandl, Institute of Telematics, University of Lübeck, Ratzeburger Allee
160, 23562 Lübeck, Germany, Mail: lea.brandl@uni-luebeck.de
German Medical Data Sciences 2023 — Science. Close to People.
R. Röhrig et al. (Eds.)
© 2023 The authors and IOS Press.
This article is published online with Open Access by IOS Press and distributed under the terms
of the Creative Commons Attribution Non-Commercial License 4.0 (CC BY-NC 4.0).
doi:10.3233/SHTI230698
89

target groups of games and gamification into player types to understand what motivates
the respective users [9–11].
This paper describes a study of player types among students of health-related
disciplines, and proposes a new method for evaluating them. Here, the goal is to identify
the player types among the students, considering not only the strongest player type, but
all player types of their personalities.
1.1. Theoretical Background and Related Work.
There are several papers that describe how gamification is used in health-related
education. For example, Huber et al. [12] use an audience response system to test
students, playfully. Unfortunately, the player types are often ignored.
Bartle's model is among the widely used models of player types. It defines four
different player types based on two dimensions.
The first dimension deals with the relevance of the environment of the game or being
surrounded by players. The second dimension extends from the preference of acting to
that of reacting [9]. Critical to the model is that players in games must satisfy multiple
properties of different types. Although it is often used, a player cannot be assigned
exclusively to one type, but has characteristics of different types. Furthermore, the model
is not based on sociologically collected data, but only on experiences and reports of
players of the genre Multiuser Dungeon [8, 9].
1.1.1. The Gamification User Types Hexad Framework
To create a model that is not limited to specific game genres, but to enable the analysis
of player types to a broader application group, Marczewski created the Gamification User
Types Hexad Framework. The framework describes six different types, of which four (1
to 4) are intrinsically motivated, one (5) is extrinsically motivated, and one (6) is neither
extrinsically nor intrinsically motivated [8, 11]:
1. Socializer: This type of player wants to interact with other players. They are
interested in parts of the system that allow them to connect with other
participants.
2. Free spirit: They do not want to be restricted in exploring the system or creating
new things within the system. For them freedom of choice is very important.
3. Achievers: They want to use and master the system 100%. This is about
improving themselves.
4. Philanthropist: This type is motivated by the feeling of being part of something
bigger. Without expecting anything in return, they want to give to other players.
5. Player: This is the only extrinsic player type. These players are looking for
rewards or quid pro quos from other players.
6. Disruptor: This type is neither extrinsically nor intrinsically motivated for the
system. He is motivated by disrupting the system or fellow players.
1.1.2. The five player types of Gaalen et al.
Based on Bartle's model, a less genre-specific model for identifying gamer types was
therefore conducted in a study among medical and dental students at the University of
Groningen in the Netherlands. Subjects in the study were asked to sort a total of 49
statements regarding preferred game mechanics by providing information on whether
L.C. Brandl and A. Schrader / Clustering on Player Types of Students in Health Science90

they agreed, disagreed, or were neutral towards them. Five genre independent player
types were identified [10]:
1. Social achiever: This player type prefers collaborative achievement of goals and
successes in a game.
2. Explorer: Players of this player type favor exploring and changing a game world
on their own.
3. Competitor: These players attach great importance to winning in a game. They
look for competition with other players, but also with computer-controlled
opponents.
4. Socializer: They use the activity of gaming to meet and exchange ideas with
other players.
5. Troll: This type of players compensates their own lack of interest in a game by
using cheats, exploiting bugs in the game to their own advantage, and annoying
or harassing other players.
Among the 109 participants, 30 were assigned to a type. 12 Social Achievers, seven
Explorers, four Competitors, five Socializers and two Trolls were identified [10].
2. Methods
In order to identify the types of players, the students of the University of Lübeck were
asked by e-mail to fill out an online survey between 25.01.23 and 08.02.23. . From the
field of health sciences this concerns students of occupational therapy/logopedics, health
and care research, midwifery, human medicine, nursing, and physiotherapy 3]. These
were included in the study. There was no reward offered to the students, other than the
fact at the end of the survey to know their own player type. All participants confirmed
that they had not already completed the survey, otherwise the data set was excluded from
the analysis.
2.1. Sample size calculation
The sample size calculation was based on the University of Lübeck's 2021 university
figures. According to these, 2098 students are enrolled in the previously listed degree
programs 3]. Based on Cochran's formula for population-adjusted sample size 4] (see Eq.
(1)) 92 responses are required for a representative sample, assuming a significance level
of 95% and an error of 10%.
݊=
௡
బ
ଵା(
೙బషభ
ಿ
)
with ݊
଴
=
௭
మ
௣(ଵି௣)
௘
మ
(1)
2.2. Measuring instrument
Tondello et al. developed and validated a questionnaire that identifies player types
according to the Hexad Framework. The questionnaire includes 24 items, which are
answered on a 7-point Likert scale. Each player type is assigned four items [8]. This was
L.C. Brandl and A. Schrader / Clustering on Player Types of Students in Health Science 91

translated into German and validated by Krath and von Korflesch [7]. Compared to the
instrument used by Gaalen et al., it can be completed in a shorter time. In addition, a
simple evaluation of all expressions is possible. For the evaluation of the measuring
instrument, the items of the respective characteristics are added up [8]. It is suggested to
assign the player type whose proficiency is the highest. Since the proficiencies of other
player types among the respondents are not considered, the data in this paper were
evaluated using a cluster procedure that takes all proficiencies into account. To ensure
the comparability of the data sets and to consider only the proficiency of the player types,
the data were standardized before the cluster procedure. Three cluster procedures were
compared using the Calinski-Harabasz [15] and Davies-Bouldin scores [16]. The
unsupervised clustering algorithms K-Means, Meanshift, and agglomerative hierarchical
clustering (Euclidean distance, ward linkage) were compared to analyze the data. In
addition, preprocessing by PCA for K-Means was analyzed. The number of clusters was
evaluated using the Silhouette score as well as the Elbow plot or a histogram. The
clustering and the analyses were realized with python (v. 3.10).
The Hexad framework was used to assign individual game mechanics to the
identified clusters.
3. Results
There were 190 students participating in the survey. The standardization resulted in
values between -2.236 and 2.122. The best values were obtained for clustering with three
clusters in the K-Means and hierarchical procedure and four clusters in the Meanshift
procedure. For the latter, the estimator of Comaniciu and Meer [17] was used.
3.1. Clusters
The best results are obtained with K-Means clustering and previous PCA with two axes
(see Table 1). The Silhouette score for three clusters is 0.44.
Table 1. Comparison of the clustering algorithms K-Means, Meanshift and agglomerative hierarchical
clustering (Euclidean distance, ward linkage) using the Calinski-Harabasz score and the Davies-Bouldin score.
Algorithm Calinski-Harabasz score Davies-Bouldin score
K-Means 66.617 1.314
K-Means + PCA 164.127 0.796
Hierarchical Clustering 60.119 1.343
Meanshift 24.638 1.165
Table 2. Description of the clusters by mean, variance and standard deviation of the manifestations.
Cluster Socializer Free spirit Achievers Philanthropist Player Disruptor
1
(n=94)
Mean 0.726 0.022 0.285 0.762 0.063 -1.858
Std 0.424 0.536 0.541 0.434 0.514 0.396
Var 0.180 0.287 0.292 0.188 0.265 0.157
2
(n=52)
Mean -0.889 0.361 0.746 0.340 0.730 -1.289
Std 0.643 0.587 0.583 0.636 0.491 0.716
Var 0.405 0.345 0.340 0.409 0.240 0.513
3
(n=44)
Mean 0.012 0.659 0.586 0.821 -1.235 -0.841
Std 0.808 0.610 0.604 0.612 0.383 0.681
Var 0.652 0.371 0.364 0.374 0.147 0.464
L.C. Brandl and A. Schrader / Clustering on Player Types of Students in Health Science92

Figure 1. Polar plot of cluster 1 (left) and 2 (right).
The results of the clustering process can be seen in Table 2. The first cluster includes 94
subjects. The low expression of the player type Disruptor and the strong expression of
the Socializer and the Philanthropist is noteworthy. The other types are on average less
strong to neutral in their expression, see figure 1 (left).
The second cluster consists of 52 subjects. In contrast to the first cluster, the
Socializer player type is weak to neutral, as is the Disruptor type. In comparison, the
Achiever and the Player types are remarkably strong, see figure 1 (right).
The last cluster comprises 44 subjects. The only striking feature of this cluster is the
very weak expression of the Player type. The other types within this cluster show a high
variance, shown in table 2 and figure 2, which is why no statement can be made here.
Figure 2. Polar plot of cluster 3.
3.2. Game elements
Since the player type Disruptor is found to be notably weak in the clusters and not notably
absent in other clusters, this type can be neglected when selecting game mechanics.
Socializers, and Achievers, on the other hand, should receive more attention. Due to the
high level of the Player in cluster 2, it should be considered to include elements for this
type. Also, elements to support the Philanthropist and Free Spirit types can also be
incorporated. Based on Tondello et al. [8] and Karth and von Korflesch [7], the use of
challenges or quests that are to be solved together is particularly suitable for Achievers
and Socializers. For example, teams that compete against each other. For Players,
L.C. Brandl and A. Schrader / Clustering on Player Types of Students in Health Science 93

rewarding elements such as points, or achievements represent an opportunity. Easter
Eggs for Free Spirits and administrative roles for Philanthropists can be integrated also.
4. Discussion
The choice of possible game mechanics is large and should be adapted to the students.
For example, in the case of Huber et. al. [12] the mechanics used match partly not the
player types evaluated in this work and other mechanics could further amplify the
measured effects. Compared to the results of Gaalen et al. no clear differences can be
seen, if only the most expressed player types are considered. But this paper extends
previous evaluations by clustering them to account for all characteristics of player types.
Based on this, it shows which mechanics are most suitable for students in the healthcare
sector.
The sample size is significant for the population studied and a silhouette score of
0.44 could be obtained, showing a weak cluster strength. The clustering procedure could
probably be further improved by a larger sample and stronger clusters could be achieved.
Also, bootstrap methods could be used to compensate for uncertainties regarding cluster
properties. Since only students from the University of Lübeck were surveyed, the study
should be repeated with the inclusion of additional institutions.
Furthermore, it should be noted that in a survey with voluntary participation, a self-
selection bias can be assumed, which may have had an influence on the characteristics
of the player types.
Declarations
Conflicts of Interest: None declared.
Contributions of the authors: LB, AS: conception of the study, LB: conduct of the study,
data collection, data analysis and interpretation; LB: writing of the manuscript, AS
substantial revision of the manuscript. All authors have approved the manuscript as
submitted and accept responsibility for the scientific integrity of the work.
Acknowledgement
Thanks are due to all study participants, as well as D. Wessel for his helpful comments
on the study design.
References
1. Miller GE. The assessment of clinical skills/competence/performance. Acad Med J Assoc Am Med Coll.
1990 Sep;65(9):63–7. DOI: 10.1097/00001888-199009000-00045
2. Institut für medizinische und pharmazeutische Prüfungsfragen. Gliederung des Medizinstudiums
[Internet]. Institut für medizinische und pharmazeutische Prüfungsfragen 2023 [cited 2023 Mar 17].
Available from: https://www.impp.de/pruefungen/medizin.html
3. Davis B, Summers M. Applying Dale’s Cone of Experience to increase learning and retention: A study
of student learning in a foundational leadership course. Proceedings of the Engineering Leaders
Conference 2014 on Engineering Education, 2014 Nov 8-11; Doha, Qatar: (HBKU Press); 2015:6. DOI:
10.5339/qproc.2015.elc2014.6
4. Faiella, F. & Ricciardi, M. Gamification and learning: a review of issues and research. Journal of e-
Learning and Knowledge Society. 2015 11(3) DOI: 10.20368/1971-8829/1072
L.C. Brandl and A. Schrader / Clustering on Player Types of Students in Health Science94

5. van Gaalen AEJ, Brouwer J, Schönrock-Adema J, Bouwkamp-Timmer T, Jaarsma ADC, Georgiadis JR.
Gamification of health professions education: a systematic review. Adv Health Sci Educ. 2021 May
1;26(2):683–711. DOI: 10.1007/s10459-020-10000-3
6. Lampropoulos G, Keramopoulos E, Diamantaras K, Evangelidis G. Augmented Reality and
Gamification in Education: A Systematic Literature Review of Research, Applications, and Empirical
Studies. Appl Sci. 2022 Jan;12(13):6809. DOI: 10.3390/app12136809
7. Krath J, von Korflesch HFO. Player Types and Game Element Preferences: Investigating the
Relationship with the Gamification User Types HEXAD Scale. In: HCI in Games: Experience Design
and Game Mechanics. Proceedings of the 23rd International Conference on Human-Computer
Interaction (HCI International) 2021 July 24-29, Virtual Event: Springer Nature; 2021 p.219-238. DOI:
10.1007/978-3-030-77277-2_18
8. Tondello GF, Wehbe RR, Diamond L, Busch M, Marczewski A, Nacke LE. The Gamification User
Types Hexad Scale. In: Proceedings of the 2016 Annual Symposium on Computer-Human Interaction in
Play-Austin Texas USA: ACM; 2016 p. 229–43. DOI: 10.1145/2967934.2968082
9. Bartle R. Hearts, clubs, diamonds, spades: Players who suit MUDs. Journal of MUD Research 1. 1996 1
10. Gaalen AEJV, Schönrock-Adema J, Renken RJ, Jaarsma ADC, Georgiadis JR. Identifying Player Types
to Tailor Game-Based Learning Design to Learners: Cross-sectional Survey using Q Methodology. JMIR
Serious Games. 2022 Apr 4 10(2):e30464. DOI: 10.2196/30464
11. Marczewski A. Even Ninja Monkeys Like to Play: Gamification, Game Thinking and Motivational
ateSpace Independent Publishing Platform; 2015. Design. Cre
12. Huber J, Witti M, Schunk M, Fischer MR, Tolks D. The use of the online Inverted Classroom Model for
digital teaching with gamification in medical studies. GMS J Med Educ. 2021 Jan 28;38(1):Doc3.
13. Gillessen-Kaesbach G, Münte T, Hartmann E, Fischer S. Universitätskennzahlen 2021, Universität zu
Lübeck, 2022 April. [cited 2023 March 24] Available from: https://www.uni-
luebeck.de/fileadmin/uzl_qm/PDF/Universitaetskennzahlen/ Unikennzahlen2021_Web.pdf
14. Cochran, W. G. Sampling Techniques, 2nd Ed., New York: John Wiley and Sons, Inc; 1963
15. Caliński T, Harabasz J. A dendrite method for cluster analysis. Communications in Statistics. 1974 Jan
1;3(1):1–27.
16. Davies DL, Bouldin DW. A Cluster Separation Measure. IEEE Transactions on Pattern Analysis and
Machine Intelligence. 1979 Apr; PAMI-1(2):224–7.
17. Comaniciu D, Meer P. Mean shift: a robust approach toward feature space analysis. IEEE Trans Pattern
Anal Mach Intell. 2002 May;24(5):603–19. DOI: 10.1109/34.1000236
L.C. Brandl and A. Schrader / Clustering on Player Types of Students in Health Science 95