Conducting an Experiment for Validating the Combined Model of
Immersion and Flow
Ehm Kannegieser
1
, Daniel Atorf
1
and Josua Meier
2
1
Fraunhofer Institute of Optronics, System Technologies and Image Exploitation,
Fraunhoferstraße 1, 76131 Karlsruhe, Germany
2
Karlsruhe Institute of Technology, 76131 Karlsruhe, Germany
Keywords: Flow, Immersion, Measurement, Experiment and Validation.
Abstract: Detecting high intrinsic motivation and Flow states is key for successful adaptation processes that may be
used to improve learning outcome in Simulations and Serious Games. Until now, the method of choice to
measure Flow, is the usage of questionnaires. Because of the shortcomings of this method, the ultimate goal
is, to establish an alternative measuring method through correlations of physiological sensor data. Beforehand,
the theoretical model of Flow is enhanced with the more fine-grained model of immersion plus the design and
implementation of an experiment to validate said model is introduced. In conclusion a perspective towards
preliminary test results and upcoming data analysis is given.
1 INTRODUCTION
Education is one of the biggest challenges of the 21
st
century. What better way to improve it than using the
vast amount of technology available to us today. A
successful tool aiding in education are “Serious
Games” (Girard, Écalle and Magnan, 2013), games
which do not put entertainment value in the front, but
rather focus on achieving learning experiences in
players. One of the biggest questions in the field of
Serious Game analysis is: How can this learning
effect be improved? Previous studies find that the
learning effect of Serious Games is linked to the fun
they provide to players (Deci and Ryan, 1985; Krapp,
2009). Thus, the question becomes, how can fun be
improved? And based on that, how can fun be
measured?
When looking at the raw definition of fun,
becoming voluntarily engrossed into an activity,
similarities can be found to the definition of Flow
given by Csikszentmihalyi (Csikszentmihalyi, 1991),
which describes the optimal enjoyment of an activity.
As such, Flow becomes an interesting measurement
when analysing the fun experienced during gameplay
(Beume et al., 2008). In order to better measure the
range of the immersive experience, the sub-optimal
state of experience, Immersion, is also looked at.
Flow and Immersion are described in more detail in
sections 2.1 and 2.2 respectively, but for general
purposes, they can be thought of as states of high
concentration on the game. By measuring these states,
Serious Game developers can make judgements about
how fun, and respectively, how much learning value
is provided by their game.
However, there are still problems when using this
approach. As they are subjective experiences,
measuring Flow and Immersion is difficult. The
current approach to measure them is based on
questionnaires (Nordin, Denisova and Cairns, 2014).
These questionnaires can either be used during the
game disrupting the player’s concentration – or
after the game, leading to imprecise results.
Additionally, questionnaires can only elicit subjective
measurements, further degrading the quality of the
data gathered.
For this reason, the development of a system for
automatic measurement of Immersion and Flow
becomes increasingly interesting. Instead of using
questionnaires filled out by participants, this system
uses the player’s physiological measurements to
determine their current Flow/Immersion state. In this
paper, a study towards the development of such an
automatic measurement system is presented. The
study attempts to link the experience of Flow and
Immersion in participants with reactions in their
physiology. First, a combined model of Flow and
Immersion is presented in section 3.1 in order to
252
Kannegieser, E., Atorf, D. and Meier, J.
Conducting an Experiment for Validating the Combined Model of Immersion and Flow.
DOI: 10.5220/0007688902520259
In Proceedings of the 11th International Conference on Computer Supported Education (CSEDU 2019), pages 252-259
ISBN: 978-989-758-367-4
Copyright
c
2019 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
measure the player experience in games in better
detail. The experiment structure is presented in
section 3.2. Section 4 presents how physiological
measurement data is used in evaluating the study, as
well as approaches for evaluating the model and the
physiological measurement data with Flow. In section
5, the validity of the study is analysed based on
internal and external factors. Finally, section 6
presents a conclusion and a future outlook.
2 RELATED WORK
2.1 Flow
The definition of Flow was first brought up by
Csikszentmihalyi and is commonly known as the state
of optimal experience of an activity and a state of
great concentration (Csikszentmihalyi, 1991). In the
context of sports, Flow might also be known as being
in the zone. Temporal and spatial dissociation are
the main symptoms of having reached the Flow state.
The concept of Flow is based on the intrinsic and
extrinsic motivation model of behavioural
motivation, which describes that actions can either be
motivated based on external factors, such as money
or fear of reprimand, or internal factors when
performing an action for the joy of the action itself.
Based on this intrinsic motivation, Csikszentmihalyi
considers the optimal way to experience and enjoy an
activity as one that is intrinsically motivated and
fulfils a certain set of requirements. He first describes
this in a three-channel model, showing that a fine
balance between skill of the person and challenge of
the task at hand must be achieved in order to reach the
Flow state. What makes Flow special is that
compared to passive enjoyment of an activity, the
Flow state enables people to enjoy even traditionally
taxing actions such as demanding work.
All this makes Flow an interesting research topic
for video games. Unlike television or regular books,
games are meant to be enjoyed through active
participations. Furthermore, the Flow state is similar
to the effect experienced by many players, including
loss of a sense of time and spatial awareness. In order
to study the Flow state in games, Sweetser and Wyeth
map Flow to games in their Game Flow
Questionnaire, mapping the original components of
Csikszentmihalyi’s Flow model (Sweetser and
Wyeth, 2005). This Game Flow Questionnaire is later
adapted into the EGameFlow questionnaire for use
with Serious Games (Fu, Su and Yu, 2009).
Other variations of Flow questionnaires exist.
Rheinberg et al. design a Flow questionnaire to be
used iteratively and multiple times to elicit the Flow
state in rapid succession due to its low amount of
questions (Rheinberg, Vollmeyer and Engeser, 2003).
While it was originally for use with sports, its
questions are formulated in a way that it can also be
used in general purpose environments.
Flow is generally measured using questionnaires.
Attempts to link the Flow state with physiological
measurements have been made in the past but have
not yet reached a level that they can fully replace
questionnaire elicitation. For example, Cheng
measured Flow in relation to eye movements and
found that a lower amount of rest points may be an
indicator towards increased Flow (Cheng, 2014).
2.2 Immersion
When talking about Immersion, it is important to
realize that there are two concurrent definitions of
Immersion being used in parallel (Zhang, Perkis and
Arndth, 2017). The first definition of Immersion is
based on the term presence and refers to the feeling
of being physically inside a virtual world. The second
definition, which will be used throughout the rest of
this paper, is engagement-based Immersion and is
based on the effects of an activity on a person, similar
to Flow. As such, it has been called “the sub-optimal
experience of an activity” in reference to Flow as the
“optimal experience of an activity”.
Unlike with Flow, which is defined by the
definition given by Csikszentmihalyi
(Csikszentmihalyi, 1991), there are multiple different
approaches to defining and measuring Immersion.
Ermi and yrä define Immersion based on three
dimensions in the model presented in their paper
(Ermi and Mäyrä, 2005). The first dimension, sensory
Immersion, refers to the visual and auditory
presentation of the game as the player takes it in. The
idea is that a game with better visuals and audio will
be better at immersing players in a virtual world,
whereas poor visual and audio quality will distract
from the experience presented. The second
dimension, imaginative Immersion, is based on story
elements and the world of the game. Finally, the third
dimension, challenge-based Immersion, is based on
the player’s engagement with the game world. Its
definition is synonymous with the definition of Flow
given by Csikszentmihalyi in that it refers to a balance
of skill and challenge to reach a higher state of focus.
However, it does not include the state of apathy,
added by Csikszentmihalyi to refer to situations in
which both skill and challenge are low. The model
described by Ermy and yrä is special in that it
defines both a dimension based on player
Conducting an Experiment for Validating the Combined Model of Immersion and Flow
253
engagement, challenge-based Immersion, and two
dimension based on analysing the game, imaginative
Immersion and sensory Immersion.
The most exhaustive analysis of engagement-
based Immersion is given by Cairns et al. in their
series of papers (Cairns et al., 2006; Jennett et al.,
2008). They define Immersion as a hierarchal model
with three levels, each level representing a higher
level of Immersion. The first level, engagement,
refers to the basic idea of interacting with a game. The
second level, engrossment, is reached when players
become emotionally involved with the software
presented, either positively or negatively. In this state,
controlling the game starts feeling completely natural
and input devices such as the mouse or the keyboard
are no longer consciously part of the experience. The
final level, Total Immersion, is reached when players
completely become in sync with their avatars in the
game and lose their sense of both time and of their
surroundings completely.
The weakness of the model presented by Cairns et
al. is that while it presents a way to measure
Immersion, it can make no statement about which
level of the Immersion hierarchy players are in at a
given moment. Cheng et al. improve this model by
adding dimensions to each levels of the hierarchy
(Cheng et al., 2015). The first level engagement is
broken down into the three dimensions attraction,
time investment and usability. Attraction refers to the
ability of the software to make users use it. Time
investment refers to the entry barrier of the first
Immersion level, which is spending time with the
application. Usability refers to the software being
usable, as non-usability would prevent user
engagement. The second level, engrossment, consists
of emotional attachment and decreased perceptions.
Emotional attachments can be either positive or
negative. Decreased perceptions refers to the loss of
sense of time and loss of spatial awareness. The
highest level, Total Immersion, is split into the two
dimensions presence and empathy. Presence refers to
the concept presented in the beginning of this section,
the feeling of being present in a virtual location
despite physically being present in the real world.
Empathy refers to the level of connection with the
player avatar and describes a state in which the player
can feel the emotions experienced by the avatar in the
game. They also present their findings in form of a
questionnaire for measuring Immersion.
2.3 Flow vs Immersion
When looking at the definition of engagement-based
Immersion as a hierarchical construct by Cairns et al.
and Cheng et al., a large amount of overlap can be
seen with the Flow definition presented by
Csikszentmihalyi in section 2.1. Both definitions
have requirements corresponding to the player feeling
in control and being presented with an adequate
challenge. Both Flow and the two higher levels of the
Cairns et al. Immersion model lead to an experience
of real-world disassociation, containing both a loss of
a sense of time and spatial awareness. The most
curious overlap is presented in the highest level of
Immersion, Total Immersion. In that state, players
appear to be cut off completely from the outside
world. This sounds similar to Flow, which has similar
symptoms. In fact, Georgiou and Kyza define the
empathy dimension of the extended model presented
by Cheng et al. in section 2.2 as Flow, considering it
as part of the Total Immersion state (Georgiou and
Kyza, 2017). The main difference between Flow and
Immersion is that Flow does not consider the player’s
emotional involvement in the game. A comparison
between the components of Flow and Immersion is
presented in table 1:
Table 1: Comparison between Flow and Immersion.
Flow
Immersion
Task
The Game
Concentration
Cognitive
Involvement
Skill/Challenge
Balance
Challenge
Sense of Control
Control
Clear Goals
Emotional
Involvement
Immediate Feedback
Reduced Sense of
Self and of Time
Real World
Dissociation
3 ONGOING RESEARCH
3.1 Proposed Model
One of the proposed ideas in this paper is a combined
model of Flow and Immersion (Kannegieser, Atorf
and Meier, 2018). For that purpose, the Flow model
as described by Csikszentmihalyi and the
engagement-based Immersion model described by
Cairns et al. are used. Cairns et al. have three layers
in their Immersion model, engagement, engrossment
and Total Immersion. The highest Immersion states,
engrossment and Total Immersion, share similar
phenomena, such as a loss of a sense of time and a
loss of spatial awareness. As such, a model is
CSEDU 2019 - 11th International Conference on Computer Supported Education
254
proposed, in which Immersion is a hierarchical
structure and Flow is a state at the top of the
hierarchy. The proposed model is shown qualitatively
in figure 1. This figure shows the relationship
between increased Flow and the Immersion levels
defined by Cairns et al. However, it must be noted that
there is no relationship between the skill/challenge
balance and Immersion.
Figure 1: Hierarchical Immersion model presented by
Cairns et al. and improved upon by Cheng et al. (left).
Proposed combined model of Flow and Immersion.
Qualitative view, the skill/challenge balance does not
influence Immersion.
3.2 Experiment
The experiment presented in this study has two goals.
First, the validation of the combined model of Flow
and Immersion. Second, gathering physiological data
that can be used to find a correlation between
physiological measurements and Flow/Immersion
states.
3.2.1 Physiological Measurements
Special care has to be taken when selecting
physiological metrics for use in a Flow experiment.
Certain types of measurement may hinder the Flow
experience and distract participants from becoming
immersed in the game. For this reason, metrics were
chosen that could be measured with a minimum of
intrusion and promised to yield relevant results.
The first physiological measurement type used is
Galvanic Skin Response (GSR). GSR measures skin
conductivity based on two electrodes placed on the
body. This metric was chosen due to being a useful
indicator in other psychological states, most notably
arousal (Mandryk and Atkins, 2007). Usually, GSR is
measured with electrodes placed on the hand, but
since hand movements would cause problems when
recording data, foot measurement provides an
alternate measuring possibility (Gravenhorst et al.,
2013). The skin conductance signal consists of two
different signals which are overlaid on top of each
other. One signal which changes quickly in response
to stimuli over seconds, and one signal that changes
slowly over minutes. The sensor used to record GSR
during the experiment is the Shimmer3 GSR+ unit. It
works using Bluetooth, which means participants do
not get obstructed by cables placed around their legs.
During test runs, participants have noted that they no
longer realized they were wearing the sensor,
suggesting it does not hinder the Flow experience.
The second measurement type used is an
electrocardiogram (ECG). Like GSR, ECG was used
successfully in previous studies regarding
physiological states, which made it interesting for the
study (Mandryk and Atkins, 2007). An ECG is used
to measure heart muscle activity from different angles
and can be used to extract heart rate and amplitude. It
is measured using electrodes place in the chest region.
For the experiment, five electrodes were used. The
sensor device used in the experiment is the Shimmer3
ECG unit. Like the GSR+ unit described above, it was
chosen due to its lack of cables, making the
measurement device less noticeable when
participants are wearing clothes.
The third measurement type is eye tracking.
During gameplay, player’s eye movements on the
screen are recorded. Eye movement is divided into
saccades, the movement, and fixation points, on
which the gaze is focused. Previous work in the
research has linked a lower number of fixation points
to higher Flow (Cheng, 2014), making this
measurement an interesting observation point. As
measurement is taken indirectly, it does also not
influence Immersion and Flow states. The camera
used in the experiment is the Gazepoint GP3 tracker.
The final measurement taken is web cam footage
of the player playing the game. Using this footage,
emotion recognition can be performed. The weakness
of this approach is that only emotions clearly
displayed on the face of the participant can be elicited
with great confidence. Other options for emotion
recognition, questionnaires and a facial EMG are
considered too distracting when playing games.
Electrodes on the face were found to be harder to
ignore than electrodes placed below clothes on the
chest. A full-HD camera records the centrally
positioned participant.
3.2.2 Experiment Structure
The experiment structure is based on a previous
experiment designed for measuring Flow in Serious
Games using physiological measurements (Atorf,
Hensler and Kannegieser, 2016).
The number of participants chosen for the
experiment is 40, as this number is similar to the
Conducting an Experiment for Validating the Combined Model of Immersion and Flow
255
number of participants used in other experiments in
this area (Cairns et al. 2006, Jennett et al., 2008). There
were no requirements for participants, as the
experiment is aiming for as close to a random selection
as possible.
The experiment consists of three phases. During
the setup phase, the game is selected, and the sensors
are placed on the participant. Game selection is free.
Participants can bring their own games or use a
distribution platform like Steam to install a game of
their choice. Free game selection was chosen to
improve the odds of players reaching higher Flow and
Immersion states, at the cost of game-specific analysis
options.
During the second phase of the experiment, the
gaming phase, participants play the game for 30
minutes. The duration was chosen based on test runs,
as 30 minutes were found to be enough to reach the
Flow / Total Immersion state. While the participant is
playing, the physiological measurements presented in
3.2.1 are recorded as well as gameplay footage.
Finally, during the assessment in the third phase,
participants watch a recording of their game session as
well as web cam footage of themselves. While
watching this footage, Flow and Immersion
questionnaires are filled out about how immersed the
participant was at the time of the recording. By making
participants fill out questionnaires while watching a
recording, more accurate data can be gathered without
interrupting the Flow/Immersion during the game
session itself to fill out questionnaires. A similar
approach using video footage is used in the study by
Rajava and Kivikonga (Ravaja and Kivikonga, 2008).
Three questionnaires are used in the experiment, with
one of them being split into two parts.
The first questionnaire used is the Immersion
questionnaire presented by Cheng et al. based on their
improvement upon the hierarchical model presented by
Cairns et al. (Cheng et al., 2015; Cairns et al. 2006).
The questionnaire was chosen, as it can be used to
measure the likeliness to be in each of the individual
Immersion levels, making it useful to compare Flow
with Total Immersion to test the theorized model
presented in 3.1. In order to track the participant’s
movement within the different Immersion levels, the
Immersion questionnaire is asked every three minutes
during the recording. However, test runs proved that
the Immersion questionnaire was too long at 17 bullet
points. Due to this, the quality of the responses given
by participants deteriorated. The questionnaire was
split into one immersive tendency part with ten
questions that is asked at the beginning of the
assessment phase, and one iterative part with seven
questions that is asked every three minutes. The
questions were chosen based on their contribution to
determining which state players are in, and the
questions with the most contribution for each
dimension were put into the iterative questionnaire.
The Flow questionnaire used in the experiment is
the Short Flow Scale Questionnaire developed by
Rheinberg et al. (Rheinberg et al., 2003) It was
chosen due to its low amount of ten questions. While
it was originally designed for use with sports, its
method of measuring Flow for activities is formulated
in a general-purpose sense. During the experiment,
social factors of online games are not taken into
account, meaning the social component the
GameFlow questionnaire adds to the original Flow
model can be ignored for this setup. The Flow Short
Scale questionnaire is asked every six minutes, along
with every second elicitation of the Immersion
questionnaire.
Finally, the third questionnaire used in the
assessment phase is the Game Experience
Questionnaire (IJsselsteijn, de Kort and Poels, 2013).
It is used as a support questionnaire alongside the
other two questionnaires. It is not focused on Flow or
Immersion but asks a wide range of questions about
how the participant felt during the game session. By
gathering more general info about the player
experience and linking it with the Flow/Immersion
data measured, new insights can be provided in what
experiences facilitate a higher Flow or Immersion
state. The questionnaire is asked once after the video
playback of the game footage has concluded.
An overview of the different phases is presented
in figure 2.
Figure 2: Phases of the experiment.
4 RESULTS
4.1 Preparing the Physiological Data
When working with GSR values from participants,
the first step is normalizing the data in a way results
from different participants can be compared reliably.
Normalization is performed by treating values of the
skin conductance measurement as a percentage of the
CSEDU 2019 - 11th International Conference on Computer Supported Education
256
participant's minimum and maximum value (Mandryk,
2008; Lykken and Venables, 1971). The normalization
equation follows to:
After this normalization, the calculation becomes a
regular correlation analysis.
Evaluations regarding ECG focus on calculating
heart rate changes. For this reason, both heart rate
changes and heart rate standard deviation are
calculated from the signal. Heart rate is calculated
using the time between two consecutive maximums in
the signal. This can be achieved by comparing sum-
differentiated peaks and threshold detection. Based on
(Pan and Tompkins, 1985).
Based on previous studies, the most interesting
aspect of eye tracking is the number of fixation points.
The input data are gaze points of the player, in order to
remove noise and get the final amount of fixation
points, a spatial threshold is used as outlined in the
thesis presented by Olsson (Olsson, 2007).
Emotion recognition data is based on the web cam
footage recorded during the gaming session. First, for
every second of the video, a cropped image of the
participant’s face is generated. This cropped image is
scaled to 256x256 and used as input for a pretrained
Convolutional Neural Network that recognizes
emotions in cropped face images (Levi and Hassner,
2015). As the web cam footage features players
frontally, this CNN delivers good results for the data
presented. The output of the net are probabilities for the
seven states anger, disgust, fear, happiness, neutral,
sadness and surprisedness.
4.2 Model Analysis Approach
The original assumption of the Flow/Immersion model
presented in this paper is that Flow and Immersion are
distinct concepts, and that Flow is considered the final
state of the Immersion hierarchy. From this definition,
it is assumed that Flow and the Total Immersion state
strongly correlate. This result can be proven by
correlation analysis between the Flow questionnaire
results and the Immersion questionnaire results. For
this correlation analysis, Spearman rank correlation is
used. When correlating questionnaire results, one
problem that comes up during correlation analysis is
that, they can only produce discrete results. This makes
Pearson analysis impossible if questionnaire results do
not follow a normal distribution (Sullivan and Artino,
2013). For this reason, Spearman analysis is chosen for
the correlation analysis. For each Flow questionnaire,
a normalized Flow value describing the likelihood for
the participant to be in Flow is calculated. This Flow
value is compared to the values of the three Immersion
levels calculated from the corresponding Immersion
questionnaire.
4.3 Physiological Analysis Approach
The first test when checking how physiological
measurements can be used to determine Flow and
Immersion is to look at the direct correlation between
the measurements taken and the values calculated by
the Immersion and Flow questionnaires. The metrics
used in this correlation are the ones described in section
4.1. Based on the raw correlation results, further steps
can be taken to measure Immersion and Flow using
physiological metrics.
First, these metrics are directly correlated to the
Flow and Immersion questionnaire results using the
Spearman correlation method explained in section 4.2.
These coefficients may still be too low to use them as
standalone indicators for measuring Flow, however,
their existence may help towards finding future
connections.
The first advanced technique used to try and link
Flow with physiological measurements is fuzzy logic.
Mandryk and Atkins took a similar approach in their
measurement of arousal and valence using GSR, ECG
and EMG values (Mandryk and Atkins, 2007). Using
fuzzy logic rules, new values are created using the
measured metrics. For example, excitement is defined
as high GSR AND high HR, meaning the minimum of
both values. Using this approach, a set of rules is
defined and correlated with Immersion and Flow. The
main difficulty with this technique is coming up with
useful rules based on the raw correlation results. If a
correlation between physical measurements and
Immersion exists, it may be harder to find than having
an expert define a set of rules.
The next idea is to build a classifier via the data that
was gathered in the experiment. Using deep learning, a
relation can be found between its input features and the
two classes called Flow/non-Flow, which are separated
by a threshold value based on the questionnaire value.
For this purpose, physiological data is generated for
every second of the experiment that was measured.
Flow and Immersion are extrapolated over the intervals
they cover, three minutes and six minutes, respectively.
The next step becomes selecting the input features used
in the net. For GSR, these features are the mean, the
standard deviation, the maximum, the minimum, the
maximum ratio, the minimum ratio, the mean of
negative values and the ratio of negative values. These
values are calculated for the first and second order
Conducting an Experiment for Validating the Combined Model of Immersion and Flow
257
derivatives as well. For heart rate, the mean and
standard deviation are calculated for the signal itself
and its first and second order derivatives. Eye tracking
is represented with its amount of fixation points, as
well as a heat map of fixation points and how often they
have been visited. The final set of features analyzed is
the cropped face image of the participant’s web cam
recording.
5 CRITICAL ANALYSIS
5.1 Internal Validity
There are several factors that need to be taken into
consideration when regarding the internal validity of
the study performed. As the study was short in length,
with randomly selected participants, effects that
usually occur in studies running over a longer period
do not appear. Maturation, the changes in participants
over time, Mortality, participants dropping out during
the study, and repeated testing, influences when
repeating the test multiple times, can be ruled out as
threats to internal validity. As the study took place over
four months, history might provide a threat to internal
validity. External influence that changed over time is
the weather, which was hot in summer, but less so in
autumn. Measures were taken to reduce the influence
outside weather has on the experiment via heat
regulation of the room the experiment was performed
in. Instrumentation was kept consistent during the
experiment. Fixes to the experiment setup were made
based on test runs performed before the real study took
place.
Participants were not filtered and are self-selected.
As such, there might be a bias towards people who
enjoy playing video games, and who have time to
participate in the study. Selection interaction between
participants informing other potential participants of
the experiment has happened, but as all participants
received the same information before the experiment
started, this does not influence the result of the study.
As participant selection was not chosen based on the
observed variables, Immersion and Flow, Statistical
Regression does not become a problem either.
As participants have no way of knowing their own
physiological measurements during the experiment,
they are unable to fill out questionnaires with a
meaningful bias either in favor of the hypothesis or
against it. Participants are not informed of the goal of
the study to further link Flow and Immersion, so they
do not introduce any bias when filling out both Immer-
sion and Flow questionnaires at the same time.
5.2 External Validity
External validity refers to how generalizable the results
found in the study are. The experiment presented in this
paper puts a strong focus on making its result
transferable to a large number of situations. Games are
self-selected by participants, erasing the problem of
results presented in this paper being valid only for a
game or a genre of games. The games chosen by the
participants were spread across different game genres.
Results also transfer well to the general population, as
participants were selected randomly. However, as they
are self-selected, they may introduce a bias towards
people interested in games.
The effects of the experiment environment were
attempted to be kept to a minimum. Of course, the
experiment computer differs from the personal
computer participants use at home, but having
participants bring along their own gaming hardware
would have increased the experiment duration and
effort unreasonably. External influencing factors, such
as noise or irritating lighting were avoided. Participants
wear either full size headphones or in-ear headphones,
based on their preferences. This helps block out noise,
so results in the Immersion experiment may not
transfer well to situations in which these quiet, non-
disturbing conditions cannot be achieved.
6 CONCLUSIONS
In the previous chapters, a study for linking
physiological measurements and Flow/Immersion is
presented. Preliminary results suggest that those
physiological measurements show no direct correlation
to Immersion levels or Flow. A more complex system
for evaluating these physiological measurements is
needed in order to gather meaningful results. The deep
learning approach presented in chapter four promises
to deliver useful results based on the input features
selected.
In section 3.1, the theory that Flow and Immersion
are linked is presented, which states that Flow is
observed along with Total Immersion, making it the
optimal experience of an activity compared to the sub-
optimal experience of an activity provided by the
hierarchical Immersion model. Preliminary results
support this theory. However, more in-depth
correlation analysis is needed to make further
statements. Future focus of this research will be put on
analysing both the model and the physiological data
that was gathered with the help of the deep learning
strategy presented in chapter 4.
CSEDU 2019 - 11th International Conference on Computer Supported Education
258
REFERENCES
Atorf, D., L. Hensler, and E. Kannegieser (2016). “Towards
a concept on measuring the Flow state during gameplay
of serious games”. In: European Conference on Games
Based Learning (ECGBL). ECGBL 2016. Paisley,
Scotland, pp. 955959. isbn: 978-1-911218-09-8. url:
http://publica.fraunhofer.de/documents/N-438328.html.
Beume, N. et al. (June 2008). “Measuring Flow as concept
for detecting game fun in the Pac-Man game”. In: 2008
IEEE Congress on Evolutionary Computation (IEEE
World Congress on Computational Intelligence), pp.
34483455. doi: 10.1109/CEC.2008.4631264.
Cheng, M.-T., H.-C. She, and L.A. Annetta (June 2015).
“Game Immersion Experience: Its Hierarchical Structure
and Impact on Game-based Science Learning”. In: J.
Comp. Assist. Learn. 31.3, pp. 232253. issn: 0266-4909.
doi: 10.1111/jcal.12066. url: http:
//dx.doi.org/10.1111/jcal.12066.
Csikszentmihalyi, Mihaly (Mar. 1991). Flow: The
Psychology of Optimal Experience. New York, NY:
Harper Perennial. isbn: 0060920432. url:
http://www.amazon.com/gp/product/0060920432/ref=si
3_rdr_bb_product/104-4616565-4570345.
Deci, Edward and Richard Ryan (Jan. 1985). Intrinsic
Motivation and Self-Determination in Human Behavior.
Vol. 3.
Ermi, L., and Mäyrä, F. (2005). Fundamental components of
the gameplay experience: Analysing Immersion.
Fu, Fong-Ling, Rong-Chang Su, and Sheng-Chin Yu (2009).
“EGameFlow: A scale to measure learnersenjoyment of
e-learning games”. In: Computers and Education 52.1,
pp. 101112. issn: 0360-1315. doi:
https://doi.org/10.1016/j.compedu.2008.07.004. url:
http://www.sciencedirect.com/science/article/pii/S03601
31508001024.
Georgiou, Yiannis and Eleni A. Kyza (Feb. 2017). The
Development and Validation of the ARI Questionnaire”.
In: Int. J. Hum-Comput. Stud. 98.C, pp. 2437. issn:
1071-5819. doi: 10.1016/j.ijhcs.2016.09.014. url:
https://doi.org/10.1016/j.ijhcs.2016.09.014.
Girard, C., Jean Écalle, and Annie Magnan (2013). “Serious
games as new educational tools: how effective are they?
A meta-analysis of recent studies”. In: J. Comp. Assisted
Learning 29, pp. 207219.
Gravenhorst, Franz et al. (Sept. 2013). Towards a Mobile
Galvanic Skin Response Measurement System for
Mentally Disordered Patients”. pp. 432435.
IJsselsteijn, W. A., de Kort, Y. A. W., and Poels, K. (2013).
The Game Experience Questionnaire. Eindhoven:
Technische Universiteit Eindhoven.
Jennett, Charlene et al. (Sept. 2008). “Measuring and
Defining the Experience of Immersion in Games”. In:
Int. J. Hum-Comput. Stud. 66.9, pp. 641661. issn: 1071-
5819. doi: 10. 1016/j.ijhcs.2008.04.004. url:
http://dx.doi.org/10.1016/j.ijhcs.2008.04.004.
Kannegieser, Ehm, Atorf, Daniel, and Meier, Josua (2018).
“Surveying games with a combined model of Immersion
and Flow”. In: MCCSIS 2018 Multi Conference on
Computer Science and Information Systems, Game and
Entertainment Technologies 2018.
Krapp, Andreas, Ulrich Schiefele, and Inge Schreyer (2009).
Metaanalyse des Zusammenhangs von Interesse und
schulischer Leistung. postprint.
Levi, Gil and Tal Hassner (2015). “Emotion Recognition in
the Wild via Convolutional Neural Networks and
Mapped Binary Patterns”. In: Proceedings of the 2015
ACM on International Conference on Multimodal
Interaction. ICMI ’15. Seattle, Washington, USA: ACM,
pp. 503510. isbn: 978-1-4503-3912-4. doi:
10.1145/2818346.2830587. url: http:
//doi.acm.org/10.1145/2818346.2830587.
Mandryk, Regan L. and M. Stella Atkins (Apr. 2007). “A
Fuzzy Physiological Approach for Continuously
Modeling Emotion During Interaction with Play
Technologies”. In: Int. J. Hum.-Comput. Stud. 65.4, pp.
329347. issn: 1071-5819. doi:
10.1016/j.ijhcs.2006.11.011. url:
http://dx.doi.org/10.1016/j.ijhcs.2006.11.011
Nordin, A., Denisova, A. and Cairns, P. (2014). Too many
questionnaires: measuring player experience whilst
playing digital games. Seventh York Doctoral
Symposium on Computer Science & Electronics, pp.69-
75.
Olsson, Pontus (2007). Real-time and Online Filters for Eye
Tracking”.
Pan, J. and W. J. Tompkins (Mar. 1985). “A Real-Time QRS
Detection Algorithm”. In: IEEE Transactions on
Biomedical Engineering BME-32.3, pp. 230236. issn:
0018-9294. doi: 10.1109/TBME.1985.325532.
Ravaja, Niklas and J Matias Kivikangas (May 2008).
Psychophysiology of digital game playing: Effects of
competition versus collaboration in the laboratory and in
real life. pp. 432–435.
Rheinberg, F., R. Vollmeyer, and S. Engeser (2003). “Die
Erfassung des Flow-Erlebens”. In: Diagnostik von
Motivation und Selbstkonzept. Göttingen: Hogrefe, pp.
261279.
Sullivan, Gail M. and Anthony R. Artino (Dec. 2013).
“Analyzing and Interpreting Data From Likert-Type
Scales”. In: J Grad Med Educ 5.4. 24454995[pmid], pp.
541542. issn: 1949-8349. doi: 10.4300/JGME- 5- 4- 18.
url:http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3886
444/.
Sweetser, Penelope and Peta Wyeth (July 2005).
“GameFlow: A Model for Evaluating Player Enjoyment
in Games”. In: Comput. Entertain. 3.3, pp. 33. issn:
1544-3574. doi: 10.1145/1077246.1077253. url:
http://doi.acm.org/10.1145/1077246.1077253.
Zhang, C., A. Perkis, and S. Arndt (May 2017). “Spatial
Immersion versus emotional Immersion, which is more
immersive?” In: 2017 Ninth International Conference on
Quality of Multimedia Experience (QoMEX), pp. 16.
doi: 10.1109/QoMEX.2017.7965655
Conducting an Experiment for Validating the Combined Model of Immersion and Flow
259