Professional Documents
Culture Documents
Gamer Style: Performance Factors in Gamified Simulation
Gamer Style: Performance Factors in Gamified Simulation
Related works that both epileptic and healthy control subjects show task
Cortical mapping related increases in frontal midline theta activity with
Cortical mapping refers to the recording of activity in the associated decreases in alpha activity, further supporting a
surface areas (cortex) of the brain. Electrical activity occurs relationship between increased theta and decreased alpha
at different frequencies and is thought to indicate brain activity in mental effort. Due to a concern that youths spend
behaviors associated with the different lobes of the brain, long hours playing video games, one study looked at spatio-
frontal, parietal, temporal and occipital. The frontal cortex spectral mapping during extended video game play,
regulates high order processing and motor functions, the specifically exploring frontal midline theta activity
parietal lobe is involved with somatosensory functions, and associated with mental effort [15]. Length of game play was
the temporal and occipital lobes control audio and visual found to be correlated with increasing frontal theta wave
functions respectively. Alpha signals (7-13Hz) are typical activity, suggesting that extended game play requires more,
of a relaxed waking state [18]. Beta frequencies are quite rather than less, cognitive input. Furthermore, their results
fast at 13-25Hz and signal intense mental activity. They are suggest that theta activity is higher during strategy games
more prominent over frontal and parietal regions and there than competitive games. As serious games usually involve
is some evidence to suggest that beta activity mediates some form of strategy, it would be expected that serious
motor activity, with faster response times predicted by game play would result in increased theta activity.
higher beta activity [27]. The slow delta rhythms (0-4Hz) Accordingly, increases in frontal midline theta activity and
are often large in amplitude but tend to be related to deep associated decrease in alpha activity have been found in
sleep, especially rapid eye movement (REM) sleep. Theta serious game progression [7, 31]. These studies have all
activity (4-7Hz) is also associated with sleep, however concentrated on the relationship between theta and alpha
several studies have found increased frontal theta activity activity, but have not conducted full cortical mapping of all
correlated with working memory tasks [7, 13, 15, 24, 31]. frequencies during serious games.
Many of the studies employing EEG based assessment of Serious games and gamified simulations for healthcare
video game play in the healthcare field are based on a Serious games and gamified simulations are employed in a
seminal study exploring EEG mapping of cortical activity variety of fields, but appear to be particularly effective in
during a working memory task [13]. This study healthcare for both patients and practitioners. A 2013
convincingly demonstrated increases in theta activity across review of serious games for health found over 100 games
the frontal midline area of the brain. This is the area [34], some of which were found to influence positive health
involved in problem solving, planning and aspects of behavior (reviewed in [2]). A 2012 review of serious games
working memory including attention. Where theta activity for training medical professionals found 30 available
was increased, alpha signals decreased. This led the games, 17 of which were designed specifically for training
researchers to postulate that frontal midline theta activity and 13 commercial games that facilitated the development
was related to mental effort, and that attenuation of alpha of skills useful for medical staff [14]. Most of the games
signals freed cortical resources for increased task difficulty, designed specifically for training were developed for team
as described in earlier research [26]. Their findings are training in acute care and triage, while two were developed
further supported by their observation that alpha activity for specific surgeries (coronary bypass and knee
increases with practice, suggesting that fewer cortical restructure). Commercial games facilitated psychomotor
resources are required following knowledge acquisition. skills useful for performing laparoscopies. Since this
review, more serious training games have been developed
A review of functional magnetic resonance imaging studies for specific procedures (for example [19]) including two
reported that activation of the parietal lobes is consistently training simulators designed by CSIRO, one for
associated with memory retrieval [33]. This review colonoscopies and one for bronchoscopies.
complements previous EEG studies showing a role for the
METHODOLOGY
parietal lobe in event recollection [26]. Recently, theta
This study used questionnaires, game play characteristics
activity in the temporal lobe was demonstrated to be
and spectral analysis to explore features of players’
associated with autobiographical memories with detailed
performance in a gamified simulation that have the
visual imagery [12].
potential to impact training outcomes. Approval was
Video game assessment using EEG obtained from the CSIRO Animal Food and Health
As video games are thought to rely on working memory for Sciences Human Research Ethics Committee: Low Risk
the acquisition of knowledge to allow game advancement Review Panel (23/2013).
through level progression, frontal midline theta activity has
Participants
been explored in clinical and recreational populations.
Participants (N=15) were recruited from staff at the
Mental effort related EEG activity has been used to
Australian eHealth Research Centre through email
elucidate whether video games trigger abnormal neural
invitation. Inclusion criteria included minimum age of 22
responses in adolescents subject to seizures compared to
(the corpus callosum, the tract joining the hemispheres, and
healthy aged matched participants [24]. Here it was found
the last tract to fully develop in the human brain, is mature
2015
Gamification #chi4good, CHI 2016, San Jose, CA, USA
in the majority of people by age 22 [20]); limited (RB1 and RB2). The target bronchi, next in the traversal
knowledge of lung anatomy and bronchoscopies; and order, is underlined. Visited bronchi, those the user has
normal or corrected-to-normal vision. Males were already traversed are also indicated for the user. In
overrepresented (n=12, 80%), which is consistent with the transitioning to level two (Figure 1C), names of individual
proportion of men (82%) who work as respiratory and sleep bronchi are removed, including target and visited bronchi,
medicine clinicians (thoracic departments) in Australia and the trainee is guided only by the overarching
[16]. Most participants were in their 30s or 40s (M=38.4, navigational information. In level three (Figure 1D), no
range 22-50, one participant did not disclose age), with post information is provided and the trainee must navigate the
graduate qualifications (n=11, 73%). Participants answered bronchial tree in the correct order without being guided by
a brief questionnaire on self-ratings of information directional or ordinal cues.
technology (IT) skills and knowledge of anatomy (score out
To advance the simulated bronchoscope further into the
of 10 where 1=no knowledge and 10=familiar working
lung the user moved the mouse pointer onto and left-clicked
knowledge); whether they wore glasses; and had used
target bronchi. Bronchoscope retraction was achieved by
virtual simulators or bronchial simulators (yes/no).
pressing the space bar.
Equipment
EPOC EMOTIV EEG headset
Bronchoscopy simulator While traditional EEG devices are complex, expensive and
The gamified simulation used was a bronchoscopy require specialist technicians, the Emotiv headset is a
simulator. A bronchoscopy is a highly invasive medical relatively cheap, compact device that is easy to set up and
procedure in which a thin tube tipped with a light source comparatively comfortable. The Emotiv headset was
and camera is passed through the mouth and trachea to designed as a brain-computer interface for gaming, however
perform an endoscopic examination of the lungs it has been employed for a number of research activities
Bronchoscopies are used to directly observe the branches including the assessment of psychological and cognitive
and nodes (bronchi) of the lung for a number of pathologies states (reviewed in [22]).
and conditions. Due to the complicated morphology of
bronchial branching, extensive training is required to ensure Participants used the simulator running on a desktop PC
the entirety of the lung anatomy is traversed. Such training activated via a mouse and keyboard while wearing the
is beneficial for patient care and reduces the time spent Emotiv wireless EEG headset (http://emotiv.com/eeg/). The
performing this invasive procedure. Historically, trainees headset recorded 14 channels at 128Hz across frontal (left
have learned this procedure through books, training videos AF3, F3, F7, FC5; right AF4, F4, F8, FC6), temporal (left
or observing experts undertake a procedure. With rapidly T7; right T8), parietal (left P7; right P8) and occipital (left
changing technologies, these methods are time consuming O1; right O2) lobes, with bi-polar reference nodes (CMS,
and do not instill high confidence levels in the trainees. DRI) using the international 10-20 system (Figure 1).
Procedure
CSIRO’s low fidelity, ‘part task’ bronchoscopy simulator
Prospective participants attended a brief overview of the
(Figure 1) was designed for novice trainees when learning
project where consent forms, participant information sheets
the Ikeda nomenclature [17] for lung bronchial tree naming
and interview questionnaires were completed. Eligible
and navigation to enhance knowledge of lung anatomy and
participants were appointed scheduled times for
improve efficacy of bronchoscopic navigation [9]. The
participation in the study, consisting of one Tuesday and
simulator includes a number of game elements including
one Friday session (four days apart). Due to the number of
increasingly difficult levels, task achievement to advance
participants, experiments spanned three weeks.
through levels, return to start for error. Game metrics
include level times and number of errors. To ensure the user The first session, on Tuesday, comprised fitting the Emotiv
experience mimics a real bronchoscopy, visual game EEG headset and describing the controls of the
elements are kept to a minimum (see [29]). Bronchoscopy simulator. Spectral data was recorded from
the start of the game play until all three levels were
Three simulator levels provided users with increasing levels
completed with three consecutive error-free traversals or 20
of difficulty by reducing the number of navigational cues
failed attempts were made in any of the three levels. The
visible on the screen. Figure 1A shows the start point for
second session, on the following Friday, consisted of a
the simulation, a low fidelity depiction of a bronchoscope
refresher of the controls and then a retrial of the highest
camera after passing the trachea. Figure 1B depicts the
level completed in their first session and again spectral data
bronchoscope image after entering (traversing) the right
was recorded. Participants had the option of removal of the
main bronchus, with the apical (RB1) and posterior (RB2),
Emotiv EEG headset every 15 minutes if the headset felt
sublobar lung segments visible. Cues visible during
uncomfortable. In order to explore features of performance,
traversal when the simulator is in level one (the least
high performing participants (those who achieved level
difficult mode) (Figure 1B) include overarching
three in session 1) were separated from low performing
navigational information provided around the diameter of
participants (those who achieved level two in Session 1).
the trachea, bronchi names floating in the bronchi lumen
2016
Gamification #chi4good, CHI 2016, San Jose, CA, USA
Figure 1. The bronchoscopy simulator facilitates learning lung anatomy and nomenclature in bronchoscopy trainees, providing less
navigational information with each level. The simulator is using a computer with keyboard and mouse while wearing Emotiv
headsets, that recorded EEG activity from 14 points across frontal (left AF3, F3, F7, FC5; right AF4, F4, F8, FC6), temporal (left
T7; right T8), parietal (left P7; right P8) and occipital (left O1; right O2) lobes, with two bi-polar reference nodes (CMS, DRI).
2017
Gamification #chi4good, CHI 2016, San Jose, CA, USA
2018
Gamification #chi4good, CHI 2016, San Jose, CA, USA
2019
Gamification #chi4good, CHI 2016, San Jose, CA, USA
Frontal midline Several studies have correlated frontal midline theta Is level complexity reflected in
theta Session 1 activity with task complexity [13, 15, 24, 31]. frontal midline theta activity?
Posterior theta The parietal and temporal lobes are associated with Do groups differ in memory
in both groups recall of events [33] and autobiographical memories recall?
[12], respectively.
Fronto-parietal Beta activity across the fronto-parietal regions is Does beta activity vary across
beta across thought to facilitate motor behavior, and is sessions?
sessions correlated with decreased response times [27].
Table 3: Summary of comparisons undertaken in the EEG spectral analysis (see Fig. 1 for electrode composition).
This shows that there are hemispheric differences in game While this trend did not reach significance, it reflects the
play, but only for Group L3. No significant complexity of level two compared to level one and level
interhemispheric differences were found in Group L2 three. This trend is supported by our findings in the game
(Wald Chi square (1)=0.400, p=0.527, β=0.245, CI 95%=(- characteristics where an increased number of errors
0.513,1.002)). (traversals), and hence longest level completion times, were
recorded for level two.
To determine if this activation pattern was consistent across
sessions for Group L3 participants we compared
hemispheric activation in Sessions 1 and 2.We found a
significant interaction between Session and Hemisphere for
alpha activity (Wald Chi square (1)=7.255, p=0.007, β=-
1.182, CI 95% (-2.043,-0.322)), where alpha activity was
most attenuated in the right (rather than the left) hemisphere
in Session 2 (compared to Session 1). This is consistent
with an increased requirement for visuo-spatial cortical
resources to navigate the level on spatial cues alone in the
absence of nomenclature guides.
Within group differences in frontal midline theta activity
The frontal midline area of the brain is involved in problem
solving and planning, and particularly planning of motor
activity. It is also the area of the brain that directs attention
to relevant information while suppressing inappropriate
responses. We described earlier previous research that
found increasing frontal midline theta activity associated
with increased task complexity [13], and this finding has
been replicated in recreational video games [15, 24] and
simulation tasks [31]. We were therefore interested to see if
activity increased in this area of the brain during navigation
of the bronchoscopy simulator. The frontal midline area
corresponds to electrodes F3 and F4 (Figure 1, Table 3).
We compared activity in this area for each group to gauge
increasing task complexity across the levels of the game in
Session 1. For Group L2, no differences in frontal midline Figure 2: A trend toward an increase in frontal midline theta
theta activity were found between level one and level two. activity in both left (F3) and right (F4) hemispheres (reflecting
For Group L3, a trend toward increased frontal midline increased complexity) was found in level two for Group L3
theta activity was observed during level two of the three (those who completed all three levels in Session 1).
levels played (Figure 2).
2020
Gamification #chi4good, CHI 2016, San Jose, CA, USA
Groups differ in posterior theta activity Frontal-parietal beta activity increased in Group L3
The learning of tasks relies on working memory, which One last area of interest for us was the fronto-parietal
involves a network of brain activity including the frontal region. The frontal lobe houses the motor cortex while the
midline areas previously mentioned. This network extends parietal lobe houses the somatosensory cortex. This region
into parietal and temporo-parietal cortices [11]. We would therefore co-ordinates motor responses to perceived stimuli.
therefore expect to see some increased activation in the Beta activity across the fronto-parietal regions is thought to
parietal lobes during game play. In this study we found facilitate motor behavior, and as noted earlier, is correlated
increased theta activity in the parietal lobes in both groups. with decreased response times (faster response times are
What was most intriguing was a significant interaction predicted by higher beta activity [27]). As Group L3 had
between Session and Level achieved (Wald Chi square consistently fast level traversal and completion times, and a
=9.296, p=0.002, β=4.164 CI 95% (1.487,6.842)) (Figure high level of accuracy in Session 2 (Table 2), we were
3). For participants in Group L2, average theta activity in interested to see if there were neural correlates for this
the parietal lobe for Session 1 was significantly higher than behavior. For Group L3 a significant interaction between
in Session 2 (Wald Chi square = 4.823 p=0.028, β=2.266 CI Session and Frequency was found for both frontal (Wald
95% (0.244, 4.823)). Conversely, for participants in Group Chi square=5.429, p=0.020, β=-0.669, CI 95% (-1.233,-
L3, average theta activity in the parietal lobe for Session 2 0.106)) and parietal (Wald chi square = 11.328 p=0.001,
was significantly higher than in Session 1 (Wald Chi square β=-2.006, CI 95% = (-3.178, -0.833)) lobes, where beta
= 4.500 p=0.034, β=-1.898, CI 95% (-3.652,-0.144)). activation in Session 2 was significantly higher than in
Activity in the parietal lobe is particularly associated with Session 1. This finding, together with increased right
episodic memory retrieval, including recall of personal hemisphere activity (see above) is in accordance with
events and experiences [33]. This activity could also be previous studies [27]. Group L2 also showed a significant
explained by the role of the parietal lobe in visuomotor interaction between level and frequency for beta activation
transformations for actions of the hand [10]. (Wald chi square = 3.969, p=0.046, β=1.501, CI 95% =
(0.024, 2.977)), with increased beta activity in Session 1.
During our analysis it was revealed that in addition to
However, the increase was limited to the parietal lobes, and
increased theta activity in the parietal lobe in Session 1,
therefore did not involve the motor cortex. This shows that
those in Group L2 also had significantly higher theta
beta activity does vary across sessions and may underpin
activity in the temporal lobe in Session 1 than in Session 2
speed and accuracy.
(Wald chi square = 4.515, p=0.034, β=3.835, CI 95%
(0.298,7.372)). Temporal lobe theta activation has been Summary of spectral profiles
shown to be associated with recall of autobiographical For Group L3, theta activation was higher in Session 1 than
memories [12]. This could indicate the groups do differ in in Session 2, and frontal midline theta activity was highest
memory recall. during level two in Session 1. During the Session 2, Group
L3 had greater activation in the right hemisphere, and
These findings are particularly interesting. Taken together increased beta activity in the frontal and parietal lobes. The
one speculative explanation could be that during Session 1, spectral profile of Group L3 suggests that they relied most
participants in Group L2 drew on previous autobiographical on visuo-spatial cues, and expended the most mental effort
schema to assist in learning the material in the gamified on level two in Session 1. The increased activity in Session
simulation, and hence parietal and temporal lobes were 2 suggests they drew on memory to coordinate a high level
activated; whereas participants in Group L3 relied more of accuracy. For Group L2, theta and beta activation was
heavily on recall in Session 2, potentially of the material higher in Session 1 in temporal and parietal lobes, which
gleaned in Session 1. These variations in the neural may suggest an attempt to draw on existing schema. While
approach to Session 1 could explain the performance more studies would be required to elucidate less speculative
differences between the groups. interpretations, these spectral profiles clearly demonstrate
group differences in processing game features.
2021
Gamification #chi4good, CHI 2016, San Jose, CA, USA
Figure 3: Spectral profile measured in log power for those participants who reached level two (Group L2) and those who reached
level three (Group L3) for Sessions 1 and 2. The profiles differed within groups between sessions and between groups. Asterisks
represent significantly higher activity within groups between sessions. A significant interaction between level and session was found
where Group L2 had greater activation in Session 1 compared to Session 2, and Group L3 had greater activation in Session 2
compared to Session 1 (0–4Hz delta, 4–7Hz theta, 7–13Hz alpha, 13–25Hz beta).
DISCUSSION research supports recent evidence of age-related slowing of
In this study we found that age influenced player reaction times within a video game scenario [32], and is in
performance, with younger participants demonstrating accord with traditional studies of age related declines in
greater knowledge acquisition through level progression task performance (for example [6]). As such, our findings
and completion assessment. On average, older participants have significant real-world implications. A study of the
made more errors, were consistently slower at individual cardiothoracic workforce in Australia found that only 1.9%
traversals and were significantly slower at whole level of cardiothoracic surgeons was aged less than 35 years [1].
completion than younger participants. Our EEG analysis This places trainee surgeons in the age overlap between
revealed differing spectral profiles that may account for Group L2 and Group L3, where the greatest variation in
variations in player performance. knowledge acquisition through gamified simulation is likely
One interpretation of spectral profiles is that performance in to be observed.
Group L3 was characterized by a reliance on visuo-spatial In 2015, the age of 35 is pivotal in our digital era as it
cues and potentially recall of Session 1 material to facilitate distinguishes digital natives, those born after 1980, from
completion of Session 2. In contrast Group L2 may have digital immigrants [25]. Digital natives have been raised,
attempted to draw on existing autobiographical schema to and are now being born, in a sea of technology, surrounded
facilitate game progress. Differences, however, could also by all the gadgets of the digital age. Their education is
be due to procedural or recognition memory. provided in a completely different language, and as a result
To our knowledge, this is the only assessment of knowledge they think differently, and process information in ways that
acquisition and spectral mapping during gamified are functionally different to digital immigrants [25]. We
simulation play involving a bronchoscopy simulator, and found it interesting, but not altogether surprising, that the
the first study to demonstrate age related differences in age difference we noted in level achievement hinged on this
medical training game characteristics and performance. Our digital divide. While there were a couple of digital
immigrants in Group L3, the group was predominantly
2022
Gamification #chi4good, CHI 2016, San Jose, CA, USA
composed of digital natives, whereas all Group L2 level three is completed with ease, and that this information
participants were digital immigrants in the age range 35–50. is retained at least in the short term. Therefore, this minor
The three participants who did not complete level two were change to the game design may have the potential to
at the upper end of this range. As serious games provide a significantly improve game performance for digital
digital form of knowledge acquisition, in the ‘language’ of immigrants.
digital natives, it is possible that digital immigrants will
Training regimes
struggle to acquire knowledge through this mechanism. The findings from this study indicate that simply extending
In support of this assertion, completion assessment [23], timeframes for training regimes would not be helpful for
(whether a player successfully completes the game) is often those players who had trouble transitioning to level two. In
used as a simple indicator that the player sufficiently this study there were no time limits, the only limitation was
understands the subject matter. In our study, the eight 20 attempts at each level. Therefore, more training sessions,
participants in Group L2 did not complete the game in the rather than longer training sessions, would be most
two sessions provided, and three of these did not even beneficial. Alternatively, relaxing allowable error, by
complete level two, indicating that learning of the game increasing number of attempts, may facilitate learning.
material did not occur during the study paradigm. Future work
While digital immigrants may have difficulty learning The spectral analysis conducted in this study demonstrated
through serious games, preliminary evidence suggests that that Group L2 had a different spectral profile to Group L3.
serious games do enhance training outcomes for digital Whether this profile is related to the divide between digital
natives. A study of three serious games conducted with immigrants and natives would need to be explored in
college students (born after 1980) demonstrated mutually exclusive groups. Neural differences in
significantly better test scores for students who learned information processing are already being postulated for
using serious games compared to those who learned such functions as attention and memory [25], which both
through traditional instruction [4]. play a role in learning. Extending studies to knowledge
acquisition during serious games has the potential to
We have not found any studies that compare serious game provide greater insights into the applicability of serious
or simulation performance in digital natives and digital games in industries dominated by mature aged workers.
immigrants. Our findings suggest that knowledge
acquisition through existing serious games (and particularly CONCLUSION
those designed by digital natives) could be more suited to In this study, we found there were distinct features affecting
digital natives. One of the key findings in a review of players’ performance in a gamified simulation that have the
performance assessment in serious games is the importance potential to impact training outcomes, particularly for older
of scheduling of contents provided to the player [2]. participants. Our hypothesis that there would be a different
Therefore, the design of serious games needs to take into spectral profile depending on player performance, as
account personalization of player profiles, including age, measured through level achievement, was confirmed and
and structure the knowledge space accordingly. we found that this difference was evident in game play
characteristics, including number of errors and time to
Adaptations to improve knowledge acquisition complete traversals and therefore levels. We speculated that
Game design the effect of these factors on performance could be
All participants were able to navigate level one of the mitigated by minor step-wise changes to the game design
bronchoscopy simulator with ease, as would be expected and extending training regimes in order to produce the best
given the number of navigational and nomenclature possible outcomes for all trainees. Our novel findings are
guidance cues. More than 50% of the attempted traversals relevant for disciplines considering training through
were successful (error-free) and, on average, it took less gamified simulation and serious games. Inclusion of
than 15 minutes for participants to complete the level with features to accommodate age related profiles could allow all
three consecutive error-free traversals. The difficulty for players to execute serious games in gamer style.
Group L2 arose in the transition to level two. Here the
successful traversal rate dropped to less than 5%. There
were simply insufficient guidance cues to navigate the
bronchial tree in the correct order without error. For ACKNOWLEDGMENTS
participants who obtain low success rates in the We would like to thank all the participants who generously
bronchoscopy simulator, additional steps would be gave their time to this project. Special thanks to John
beneficial. Instead of removing all nomenclature for level Gardner for contributing to study design and Shlomo
two, removing nomenclature from half the bronchial lumen Berkovsky and CHI reviewers for insightful comments on
as an intermediary step, or a third at a time for additional the manuscript.
steps, may improve performance in this transition. Results
from Group L3 suggest that once level two is mastered,
2023
Gamification #chi4good, CHI 2016, San Jose, CA, USA
2024
Gamification #chi4good, CHI 2016, San Jose, CA, USA
2025