You are on page 1of 4

Perfectionism in Games - Analyzing Playing

Behaviors in an Educational Game


Greta Hoffmann Raphael Martin Christof Weinhardt
Institute of Information & Market Institute of Information & Market Institute of Information & Market
Engineering (IISM) Engineering (IISM) Engineering (IISM)
Karlsruhe Institute for Technology Karlsruhe Institute for Technology Karlsruhe Institute for Technology
Karlsruhe (KIT) Karlsruhe (KIT) Karlsruhe (KIT)
Germany Germany Germany
greta.hoffmann@kit.edu uuddl@student.kit.edu christof.weinhardt@kit.edu

Abstract — Perfectionism, if it is not pathological, can evoke would respond to the setup. Comparing our results to similar
desirable playing behaviors for games and applications that aim studies, we found a consistent but smaller number of players
for high performance in their target audience. In a field trial of that were incentivized to aim for perfectionism (7% in
an educational game about waste sorting that accumulated over comparison to ~10-20% depending on the study). In the
1600 players, we analyzed a cluster of players that achieved
course of the analysis we identified another corresponding
perfect scores in terms of performance, effort and long-term
commitment. During the analysis a complementary player type behavior pattern we named “rushers” (players that do not
cluster emerged, that while seemingly not showing interest in the correspond to any score optimizing game design elements).
score at all, showed higher long-term commitment. The findings Looking at performance, Perfectionists exceeded the Rushers
of our study suggest that perfectionism is a personality trait that significantly. However, more Rushers continued to play the
can be fostered with specific gamification-design elements game to the end. We derive from this data that the Rusher
within a certain percentage of a user-base in order to optimize group focuses on seeing more of the games content instead of
user interaction with a platform, application or interface. aiming its efforts on a high score. These insights lead us to
believe that there might be benefits in further reinforcing
Keywords—Gamification, Games User Research, Educational
Games
affordances for both types of players.
With our study we hope to enrichen game-design as well
I. INTRODUCTION as gamification research by looking at emerging player
Since its introduction in 2002 [1] gamification as well as behaviors as a result of implementing specific design
serious games have shown great results as motivational elements targeting optimal result-oriented playing behavior.
support tools for learning, behavioral change and II. LITERATURE
participation in fields of health, education and economics
[2]–[6]. One of the fields still lacking in services that When designing the research artefact, in order to better
incentivize intrinsic, long-term participation is the topic of anticipate our target audience, we conducted an explorative
domestic waste sorting. In its plan to raise EU-wide recycling literature review on the topic of player types and playing
to 55% and decrease landfill use to 10% by 2025, the incentives. Our research yielded a large amount and a broad
European Union acknowledges the importance of waste variance of different types of playing incentives and
sorting in this process. However, many recycling and waste clustering suggestions [10], [11]. We found that many papers
sorting facilities are not yet able to reach their maximum that look at playing incentives originate from a player type
efficiency without pre-sorting measures, a problem that has analysis approach. Since Bartle proposed his four player
been tackled in countries like Germany, Austria and types in 1996 [12], personality-related playing incentives
Switzerland by making domestic pre-sorting a citizen’s have dominated the research field [13]–[15]. Other
responsibility [7]. Achieving the necessary behavioral change approaches derive playing incentives based on studies around
as well as providing an incentive to do so will be expensive certain games [16], [17]. In the course of our research we
and unlikely to happen without fostering intrinsic motivation found that the most popular design elements seemed to
[8] as well as teaching necessary information to citizens. revolve around the themes of achievement, exploration and
In an attempt to achieve both, we developed a serious social interaction. However, in the different definitions of the
game that we tested in a field trial. We focused the design of Achiever-playertype and their corresponding incentives, we
our core game mechanics on gameplay elements that support did not find a type representing all the tropes we were looking
training of declarative as well as procedural knowledge [9] for. While an Achiever plays to collect, earn and achieve
by incentivizing repetition as well as long exposure to the points, badges and other rewards, a Perfectionist plays to
content. We further targeted long-term commitment for create states of perfection and/or completion [18], [19].
deeper involvement with the topic with explorative design Since this was the type of play we aimed to incentivize,
elements that include a storyline as well as side-quests we further extended our review towards studies looking at
running parallel to the main gameplay (however, research on perfectionistic traits in their target audiences. We used the
long-term commitment is not part of this study). following search terms (perfectionis* OR idealism OR
With the aim to teach the correct waste-bin for 250 compul* OR obsess* OR "perfect score" OR personality OR
different waste items, we focused on design elements that are "player type") AND ("design element" OR "game element"
meant to incentivize high/highest performances in the target OR "design mechanic" OR "game mechanic" OR "design
audience. Looking at studies where perfect or optimized play pattern" OR "game pattern" OR affordance OR "grading
was targeted, a consistent number of players emerged that system" OR "academic grading") looking on the following
platforms: ACM, IEEE Xplore, JSTOR, ScienceDirect, Web
Authorized licensed use limited to: Universidad de Sonora. Downloaded on February 08,2024 at 21:51:58 UTC from IEEE Xplore. Restrictions apply.
978-1-7281-4540-2/19/$31.00 ©2019 IEEE
of Science and Google Scholar. With a total of 1.198 the core gameplay, where the players are tasked to sort a
examined titles (Title, Keywords & Abstract) we narrowed certain number of incoming waste-items and 2. the game
the search down to 48 potential results. In the following we world where after each level the players can solve quests and
will shortly elaborate on the three publications provided the play minigames.
most relevant insights for our research. In the core gameplay, a number of the waste items
The first key paper “Game Mechanics Used for Achieving (starting with 20 at the beginning of the game and ending with
Better Results of Massive Online Courses” [20] consists of 35 items in the last waves) come in on a conveyer belt on the
three experiments with two of them investigating the effects right side of the screen and have to be dragged and dropped
of perfectionism based on the participants performance in an onto the correct bin before they drop off the other side (in the
online course (n=11.319, n=417 finishing all exercises). The first waves the movement of one item from right to left is 15
course was designed allowing for penalty-free repetition to seconds. A new item spawns every 4 seconds. Both spawn
reach a high or best score. The participants were clustered speed and item movement speed increase by 1.2% per wave.
with 13%, of the participants inherent of the idealist group. These parameters were chosen and tested beforehand to
When testing for studiousness of the participants the study afford for game-flow). To ensure training effectiveness, each
showed that participants, which finished an exercise with a wave the educational material (waste items) are divided into
perfect score, did spend more time on the exercise, showed sets of new items and already learned ones. At the end of a
more effort in looking for additional materials, as well as wave the players are shown the results of their sorting efforts
more repeated attempts in order to achieve a better score. and are given the choice to repeat the wave without negative
In the second key paper “Gamification - using elements consequences. We refer to this mechanic as the “repeat
of video games to improve engagement in an undergraduate option” in the course of this publication. If the wave is
physics class” [21], different effects of game design elements finished perfectly, the player is rewarded with a “perfect!”-
on student are measured in an online quiz with an a/b test. stamp as well as a gratifying sound effect. In this case no
The possibility to repeat penalty-free was implemented for repeat option is offered, and the player moves straight on to
both groups. The results of the paper showed, that there was the next level. These elements were designed to encourage
a significantly higher number of participants in the research repetition in case of non-perfection in order to raise the
group that repeated until they reached a perfect score, mnemonical depth as well as ensuring the quality of the
demonstrating that the gamified version of the quiz improved memorized information. According to the elaborated
the participants’ dedication to accomplish a score of 100%. literature, we expect a certain number of players to be
The third key paper “To Three or not to Three: Improving effectively incentivized to aim for optimal play by the design
Human Computation Game Onboarding with a Three-Star element of explicitly rewarding perfection (remunerated with
System” [22] analyzes the effects of a three-star reward the animation of a perfect stamp as well as a gratifying piece
system implemented into the serious game Foldit [23]. The of music) made possible by the option of repetition without
experiment compared player behavior in three different negative repercussion.
versions of the game: NO-STAR (basic game without three- To prevent a premature loss of interest due to potential
star system), 3-STAR (star rewards according to their number repetitiveness of the gameplay we added a game world (stage
of moves) and 3-STAR-R (star rewards according to their 2) with immersive, explorative design elements (a story, side-
number of moves including a forced reset after a specific quests, unlockable areas and minigames set to teach further
number of moves). waste management related topics like garden-waste, glass- &
Significant differences were observed in the amounts of paper-recycling) that can be accessed after completing each
extra moves (number of moves above the perfect value) and wave. Benefiting the story, the core gameplay is slightly
time per move between the NO-STAR and the 3-STAR altered at some points of the game (we designed “special
version. In total, the three-star mechanism doubled the waves” where e. g. the players have to play with a limited
amount of recompleted levels (Gaston & Cooper, 2017, p. amount of light due to a power outage). (Gameplay example:
5037). http://tinyurl.com/appdx1)
To evaluate the success of our design decisions in terms
III. METHOD/CONDUCT of our research goals - to incentivize perfect and prolonged
With the two main goals of the game, to effectively teach play - we published the game as a field trial. Beforehand, we
waste sorting as well as foster long-term commitment to sort conducted a preliminary study (referred to as pretest in the
at home, we translated all gathered insights of the following publication), gathering 139 identifiable players and
aforementioned literature review into in-game mechanics. 118 sets of uncorrupted data over a time period of three
In order to ensure easy access, especially for young months. After fixing some design issues (visual work-overs,
participants, we designed an application for mobile devices, adjusting the number of items per wave to grow instead of
developed for the three big platforms (Android, iOS, staying the same number over all waves and reworking the
Windows). For the core gameplay we chose a format that tutorial) we then launched the field trial that over a period of
underlies many currently successful game-applications – a six months received 799 downloads of which 743 data-sets
wave (or level)-based gameplay that doesn’t exceed playing could be used in the analysis. Using Unity Analytics, we first
times longer than a few minutes. identified unique players by IDs and filtered out incomplete
According to gamification research, strong incentives for or inconsistent data-sets, then we analyzed triggered events,
continuous play (in a single player game) are mechanics that specifically wave repetition and performance per wave per
afford for immediate feedback, a sense of growth and ID. The first 5 waves of the game were discounted as test
accomplishment as well as curiosity and discovery [24]–[26]. rounds as being part of the tutorial of the game. Players who
Based on this, we divided the gameplay into two stages: 1. quit during this time have not been included in the analysis.

Authorized licensed use limited to: Universidad de Sonora. Downloaded on February 08,2024 at 21:51:58 UTC from IEEE Xplore. Restrictions apply.
For the purposes of identifying and isolating playstyles we well as the Rushers were slightly higher in the field test
defined that players fall under the group of “Perfectionists” if compared to the pretest. In terms of long-term commitment -
their performance was 100% in more than 80% of the waves how long the players continued to interact with the game -
they played before quitting or finishing the game (we allowed more players in the Rusher group played the game from
for a 20% latitude due to the special waves mentioned earlier beginning to end (in absolute numbers). However, relatively
in this paper as well as an observable end-game effect [27]. speaking, we find more completionists within the
While looking deeper into the playing behavior of the perfectionistic behavior pattern (see table1).
Perfectionists (eye-balling the plots that compare score and
As suggested by literature, the game managed to incentivize
number of repetitions per wave per player (see:
a certain number of players to optimize their playing
http://tinyurl.com/appdx2& http://tinyurl.com/appdx3R) a
behavior. Even though the group of Perfectionists lacks in
second group of player type emerged, displaying the opposite
terms of quantity compared to the Rushers (consisting of only
playing behavior in regards to the repeat-option compared to
~5% to 7% of the total amount of players) it exceeds in terms
the Perfectionists (hereafter referred to as “Rushers”). We
of quality of performance (measured by looking at the
adapted our definition for the Perfectionists to afford for the
aggregated performance of the last repetition of each wave)
Rushers as follows: Players that have not used the repeat
as well as effort (measured by the aggregated number of
option in 80% or more of the waves they played. Here a 20%
repetitions per wave):
latitude was included affording for cases of unintentionally
hitting the repeat-button due to its suggestive design (see: TABLE II. COMPARISON OF PLAYING PERFORMANCE & EFFORT
http://tinyurl.com/appdx4).
In order to identify and analyze commitment in Perfectionists Rushers
relationship to playing behavior, we divided the players into average last repetitions average last repetitions
performance per wave performance per wave
groups according to exit peaks (waves after which a big group Pretest 99.9% 1.75 76.9% 1.43
of players left). With the 34 waves of the game the groups Field
99.0% 2.14 81.3% 1.74
were labelled “not interested” (left between wave 1-8), trial
“mildly interested” (left between wave 9-16), “interested” In a paired t-test the average last performance of the
(left between wave 17-26), “committed” (left between wave Perfectionists was significantly higher than that in the
27-34) and, as a subgroup of the “committed” group: Rushers in both the pretest (T = 3.83; p < .001) and the field
“completed” (finished the game: wave 34). A complete trial (T = -10.61; p < .001). The measured effort (repetitions
playthrough of the game takes an unexperienced player per wave) showed a significant difference between the groups
between 4-6 hours. in the field trial (T= -7.19; p < .001) but not in the pretest (T
IV. RESULTS/FINDINGS = 1.80; p = 0.085).

The findings of this study are based on the analysis of the V. DISCUSSION
aggregated, anonymized user data from a period of nine Considering that our study is a field trial, even though we
months (one batch of data from the pretest and one from the analized a considerably large set of data, the results can only
field trial). Of the total of 1660 downloads the game could be taken as indicators that should be isolated and tested in an
accumulate to the end of the trial, we received data of 1045 experimental setting in the laboratory to increase internal
players, of which 184 had to be discarded because of validation. Furthermore, since the transmission of our data
incomplete or inconsistent data-sets. The pretest accumulated depends on a reliable internet connection on the side of the
a total of 118 datasets and the field trial 743 players for the players, the actual percentages might vary (seeing that we lost
post-update version of the game. 179 data points from inconsistent data alone). While the
endgame effect we found in our plots is a researched
TABLE I. DISTRIBUTION OF PLAYING BEHAVIOR BY COMMITMENT
phenomenon in the field of economic science [27], there
Pretest (n=118) Field Trial (n=743) hasn’t been much research on the endgame effect in games.
Rushers: Perfection- Rushers: Perfection- This is something we want to look into more, because it isn’t
18 (15%) ists: 6 (5%) 142 (19%) ists: 55 (7%)
Exit-Groups clear if the adaptation can be made as linear as we did.
Not Looking at the percentages we were intrigued by the much
8 (44%) 2 (33%) 99 (70%) 31 (56%)
interested higher number of Rushers identified in the dataset in
Mildly comparison to the Perfectionists. An explanation might be
6 (33%) 2 (33%) 18 (13%) 11 (20%)
interested
Interested 3 (16%) 2 (17%) 14 (10%) 4 (7%) found by the suspected average age of the players (the game
Committed 1 (6%) 1 (17%) 11 (8%) 9 (16%) is designed to target children and youngsters). However,
(Completed) 0 (0%) 0 (0%) 11 (8%) 9 (16%) since the gathered data is anonymized (for reasons of data
privacy), the factor of age is one that should be tested for in
When analyzing the data-sets of the pretest, we found that
player behavior additional to personality traits.
a total of 5% of the players behaved according to the
Perfectionist pattern while 15% of the players could be VI. CONCLUSION / OUTLOOK
distinguished as Rushers. In the actual field trial both
numbers increased. Now 7% of the players could be With the aim of making a change in the domain of waste
identified as Perfectionists and 19% as Rushers. Even though sorting, it is most beneficial to the environment if players are
the pretest delivered a smaller set of data compared to the incentivized to aim for perfect performance. In our research
field trial, the same player behavior could be identified in we wanted to evaluate how effective our game design
both versions. The percentages of both the Perfectionists as decisions have been in terms of incentivizing perfect play.

Authorized licensed use limited to: Universidad de Sonora. Downloaded on February 08,2024 at 21:51:58 UTC from IEEE Xplore. Restrictions apply.
While having managed to incentivize a consistent number of [9] J. R. (Carnegie M. U. Anderson, Cognitive Psychology and Its
players to aim for perfect play as well as long-term Implications, 8th ed., vol. 53, no. 9. Worth Publishers, 2015.
commitment, the number of perfectionists we managed to [10] N. Yee, “PRESENCE: Teleoperators and Virtual Environments,
incentivize with the current gameplay did not meet our 15, 309-329.,” Online, vol. 15, no. 3, pp. 309–329, 2006.
expectations (compared to numbers from similar studies). In
[11] J. Tuunanen and J. Hamari, “Meta-synthesis of player
the process we identified a secondary (larger) group that
typologies,” Proc. Nord. Digra 2012 Conf. Local Glob. - Games
seemed to be more incentivized to interact with the secondary
Cult. Soc., 2012.
content of the game (explorative design elements like
minigames and side-quests all relating to the topic of waste [12] R. Bartle, “Richard A. Bartle: Players Who Suit MUDs,” MUSE
management). This group showed the opposite of the web site, April, 1996.
intended behavior, not responding to our repetition [13] K. P. Kallio, F. Mayra, and K. Kaipainen, “At Least Nine Ways
incentivizing design elements at all, however getting further to Play: Approaching Gamer Mentalities,” Games Cult., vol. 6,
in the levels. After conducting some informal interviews with no. 4, pp. 327–353, 2011.
playtesters that matched the perfectionist behavior, certain [14] P. Zackariasson, N. Wåhlin, and T. L. Wilson, “Virtual Identities
statements indicated that the strive for perfection quickly
and Market Segmentation in Marketing in and Through
depleted the general motivation to follow through with the
Massively Multiplayer Online Games (MMOGs),” Serv. Mark.
whole game. The behavior appeared to be more compulsion
Q., vol. 31, no. 3, pp. 275–295, 2010.
than motivation -based. We thus came to believe that aspects
of personality could be a strong influence on gameplay [15] S. Deterding, D. Dixon, R. Khaled, and L. Nacke, “From game
behavior. With this indicator in mind it can be helpful for the design elements to gamefulness,” Proc. 15th Int. Acad. MindTrek
community to know that both behaviors might emerge in Conf. Envisioning Futur. Media Environ. - MindTrek ’11, pp. 9–
similar proportions in similar kinds of games which would 11, 2011.
allow for game designers of future learning-oriented serious [16] J. Jansz and M. Tanis, “Appeal of playing online First Person
games to aim their nudges more precisely at the respective Shooter Games.,” Cyberpsychol. Behav., vol. 10, no. 1, pp. 133–
player group. In a planned laboratorial experiment, we want 136, 2007.
to try to quantify the role of personality in playing behavior. [17] D. Williams, N. Ducheneaut, N. Yee, and E. Nickell, “The Social
Also, it is planned to compare the benefits of a breadth of
Life of Guilds in World of Warcraft,” Games Cult., vol. 1, no. 4,
knowledge approach (seeing more of the games learning
pp. 338–361, 2006.
content while not interacting with it in depth) to the depth of
knowledge approach we aimed for by measuring and [18] R. O. Frost, P. Marten, C. Lahart, and R. Rosenblate, “The
comparing the learning effect with an A/B test. dimensions of perfectionism,” Cognit. Ther. Res., vol. 14, no. 5,
pp. 449–468, Oct. 1990.
REFERENCES [19] K. L. Bevins and C. D. Howard, “What we know about
gamification so far,” vol. 6, no. 1, pp. 58–84, 2018.
[1] K. Werbach and D. Hunter, For the win : how game thinking can [20] L. S. Lisitsyna, A. A. Pershin, and M. A. Kazakov, “Game
revolutionize your business. Wharton, 2012. Mechanics Used for Achieving Better Results of Massive Online
[2] J. Hamari and J. Koivisto, “‘Working out for likes’: An empirical Courses,” Springer, Cham, 2015, pp. 183–192.
study on social influence in exercise gamification,” Comput. [21] J. A. Rose, J. M. O’Meara, T. C. Gerhardt, and M. Williams,
Human Behav., vol. 50, pp. 333–347, 2015. “Gamification: using elements of video games to improve
[3] R. N. Landers and A. K. Landers, “An Empirical Test of the engagement in an undergraduate physics class,” Phys. Educ., vol.
Theory of Gamified Learning: The Effect of Leaderboards on 51, no. 5, p. 055007, Sep. 2016.
Time-on-Task and Academic Performance,” Simul. Gaming, vol. [22] J. Gaston and S. Cooper, “To Three or not to Three: Improving
45, no. 6, pp. 769–785, 2014. Human Computation Game Onboarding with a Three-Star
[4] Y. Liu, T. Alexandrova, and T. Nakajima, “Gamifying intelligent System,” Proc. 2017 CHI Conf. Hum. Factors Comput. Syst. -
environments,” Ubi-MUI ’11 Proc. 2011 Int. ACM Work. CHI ’17, pp. 5034–5039, 2017.
Ubiquitous meta user interfaces, no. June 2015, p. 7, 2011. [23] “Foldit.” University of Washington Departments of Computer
[5] P. Denny, “The effect of virtual achievements on student Science & Engineering and Biochemistry, 2008.
engagement,” Proc. SIGCHI Conf. Hum. Factors Comput. Syst. - [24] E. D. Mekler, F. Brühlmann, K. Opwis, and A. N. Tuch, “Do
CHI ’13, p. 763, 2013. points, levels and leaderboards harm intrinsic motivation?,” Proc.
[6] G. Barata, S. Gama, J. Jorge, and D. Gonçalves, “Improving First Int. Conf. Gameful Des. Res. Appl. - Gamification ’13, pp.
participation and learning with gamification,” Proc. First Int. 66–73, 2013.
Conf. Gameful Des. Res. Appl. - Gamification ’13, pp. 10–17, [25] J. Hamari, “Computers in Human Behavior Do badges increase
2013. user activity ? A field experiment on the effects of gamification,”
[7] N. Buclet and O. Godard, Municipal Waste Management in Comput. Human Behav., pp. 1–10, 2015.
Europe : a Comparative Study in Building Regimes. Springer [26] S. Deterding, “Gamification: designing for motivation,”
Netherlands, 2000. Interactions, vol. 19, pp. 14–17, 2012.
[8] E. L. Deci, R. Koestner, and R. M. Ryan, “Extrinsic rewards and [27] R. Selten and R. Stoecker, “End behavior in sequences of finite
intrinsic motivation in education: Reconsidered once again,” Rev. Prisoner’s Dilemma supergames A learning theory approach,” J.
Educ. Res., vol. 71, no. 1, pp. 1–27, 2001. Econ. Behav. Organ., vol. 7, no. 1, pp. 47–70, 1986.

Authorized licensed use limited to: Universidad de Sonora. Downloaded on February 08,2024 at 21:51:58 UTC from IEEE Xplore. Restrictions apply.

You might also like