Willingness of Distracted Smartphone Users on the Move to be Interrupted in Potentially Dangerous Situations

Abstract

The use of smart devices has become an integrated part of our everyday life. Communication is now possible everywhere at any time. The distraction caused by these devices, however, can lead to potentially dangerous situations. To mitigate these situations various researchers have proposed and developed solutions to analyze the environment and to alert the user if a situation is determined dangerous. While seeking technical solutions the concerns of the users are usually not addressed. With our study we put the belongings of the user into focus and investigated the acceptance and other issues of such type of applications. We found that environment awareness for users receiving warnings over those who do not can be significantly improvement. We found that 56% of the participants who tested a safety application and received warnings would use it while only 46% would use it a-priori without experiencing it. We can also confirm that females are more willing to use a safety application in comparison to men.

Full text

Willingness of Distracted Smartphone Users
on the Move to be Interrupted
in Potentially Dangerous Situations
Sandra Beuck, Alexander Scheurer, Matthias W¨
olfel
School of Digital Media
Furtwangen University
Furtwangen, Germany
Email: matthias.woelfel@hs-furtwangen.de
Abstract—The use of smart devices has become an integrated
part of our everyday life. Communication is now possible
everywhere at any time. The distraction caused by these devices,
however, can lead to potentially dangerous situations. To mitigate
these situations various researchers have proposed and developed
solutions to analyze the environment and to alert the user if
a situation is determined dangerous. While seeking technical
solutions the concerns of the users are usually not addressed.
With our study we put the belongings of the user into focus
and investigated the acceptance and other issues of such type
of applications. We found that environment awareness for users
receiving warnings over those who do not can be significantly
improvement. We found that 56% of the participants who tested
a safety application and received warnings would use it while only
46% would use it a-priori without experiencing it. We can also
confirm that females are more willing to use a safety application
in comparison to men.
I. INTRODUCTION
The use of smart devices has become an integrated part
of our everyday life. The number of smartphone users has
increased in recent years and is likely to keep increasing in
the future [1]. Communication is now possible everywhere
at any time. The distraction by those devices, however, can
lead to potentially dangerous situations. Studies by Nasar,
Hecht and Werner [2], Neider et al. [3] and Tapiro et al. [4]
have already analyzed different forms of user’s smartphone
distraction while walking, as well as risks resulting from this
trend. To reduce dangerous situations caused by smart devices
various researchers (reviewed in Previous Work) have proposed
and developed solutions to analyze the environment and to
alert the user if a situation is labeled potentially dangerous.
While such systems seem to be useful and might lead to a
reduction in accidents, it is not obvious if such applications
would be used—let alone be installed—by the users. This
effect has been seen for example with safety belts as well
as bike helmets. While the reasons for resistance might be in
discomfort and rumpled hair, in the case of safety apps the
reason might be in a reduced battery duration or unwanted
disruptions.
Our study aims to evaluate the user acceptance for safety
apps sensing the environment, the way the information has
to be presented, if it is perceived beneficial and whether it
would be used. We can confirm that such a system shows
significant improvement in environment awareness for users
receiving warnings over those who do not. In addition, we have
observed that the interest in safety apps is raised by more than
10%, for those who experienced such a system in a real-world
scenario.
II. PREVIOUS WO RK
This sections reviews previous works on user behavior,
technical solutions and type of warnings.
A. User Behavior
Nasar and Troyer [5] analyzed the contents of the National
Electronic Injury Surveillance Systems-Database (NEISS) [6],
which is a database that records information about injuries
from 100 American hospitals categorized by the involvement
of consumer products. They found an increasing number of
pedestrian injuries due to mobile phone usage in public spaces
from 2004 to 2010.
The user’s perspective is shown by a study released by the
American Academy of Orthopedic Surgeons about distracted
walking [7]. 2000 adults were asked about their smartphone
habits, specifically about distracted walking, distracted walk-
ing incidents and how often they observe this in others. 28%
use their mobile device at least sometimes while walking and
26% have already experienced a distracted walking incident.
They define a distracted walking incident as at least bumping
into something or somebody. The number of observed dis-
tracted walking is at 84% (at least sometimes) and distracted
walking incident at 38%.
Nasar, Hecht and Werner [2] as well as Neider et al. [3]
analyzed distraction for pedestrians while talking on the phone.
Their studies focused on different forms of distraction while
walking or crossing a street. They compared talking on the
phone to listening to music and to not using a mobile device
at all. Both studies show an increased level of distraction
for listening to music while talking on the phone offered the
highest level of distraction.
A recent study by Tapiro et al. [4] investigated this topic
further by looking into distraction while crossing a street for
2017 International Conference on Cyberworlds
978-0-7695-6215-5/17 $31.00 © 2017 IEEE
DOI 10.1109/CW.2017.60
9
2017 International Conference on Cyberworlds
978-0-7695-6215-5/17 $31.00 © 2017 IEEE
DOI 10.1109/CW.2017.60
9
2017 International Conference on Cyberworlds
978-1-5386-2089-2/17 $31.00 © 2017 IEEE
DOI 10.1109/CW.2017.60
9

different age groups ranging from 7 to 29 years old. They
investigated four conditions: undistracted,naturalistic,visual
search, and arithmetic task.Undistracted means to complete
the crossing task without distraction and no mobile device.
For all other conditions participants had a phone conversation
with a research assistant, giving them a task. A naturalistic
phone conversation can be described as casual small talk.
In the visual search task, participants were asked to locate
items in their surroundings. For the arithmetic task participants
had to solve arithmetic questions, suitable for their age. The
participants were judged by their decision on how to safely
cross the road. The results show that younger participants were
more likely to be distracted and cross the street in a dangerous
situation. The visual search task shows the highest level of
distraction, followed by arithmetic task.
Following up on the visual task, Schwebel et al. [8] ana-
lyzed listening to music, texting and talking on the phone,
also while crossing the street. They rated the participant’s
crossing behavior by their speed, time to spare and general
safety aspects. They found texting to have a high level of
distraction.
Haga et al. [9] compared smartphone usage with the task
of texting, watching a video and playing a game. Their
participants had to walk a 3x3m perimeter with a straight line
to walk on. They measured how accurately they step on the
line, as well as how timely they react to visual and auditory
cues around them. They found that playing a game was the
most distracting usage, followed by texting.
B. Technical Solutions
The reviewed studies clearly outlined that various forms of
smartphone usage is distracting and can lead users and others
into dangerous situations. Because user’s mindset and behavior
is likely not changing researchers proposed technical solutions.
To improve pedestrian’s safety they proposed to sense the
environment, classify the situation and to warn the distracted
smartphone user about potential dangerous situations in order
to avoid collisions.
A quite rigorous solution is HeadsUp by Zhou [10]. It
simply prevents smartphone usage while walking at all. It
blocks the screen and displays a warning message if a user
is walking and using the smartphone at the same time.
Other applications do not block smartphone usage while
walking in general, but intervene in those cases where algo-
rithms detect potential dangerous situations. Such applications
include InfraSee [11], UltraSee [12] and CrashAlert [13].
Those approaches all used additional sensors, not yet inte-
grated into the smartphone, for instance depth cameras to
recognize upcoming obstacles.
Automatic Accident Detection and Alarm System (Au-
toADAS) [14] is relying only on sensor data (camera, gyro-
scope and accelerometer) provided by ordinary smartphones
to warn the user in case of upcoming dangers via sound, text
message, pop-up or vibration.
C. Warnings
The work CrashAlert [13] includes a brief evaluation of
user’s preferred kind of warnings: It analyzed different meth-
ods of displaying a small slice of the devices camera’s image
on the user’s screen while walking. Thus, the user does not
need to move their head up to see upcoming static or dynamic
obstacles. The authors used three different forms to display the
camera’s image: color image shows a slice of the camera’s
image, depth image is a visualization of depth for a fixed
distance of 5 meters, and the masked image combines both,
showing the user the closest obstacles by abstracting the image
with a mask. All display methods presented a red alert when
the obstacle was less than 2 meters away. The study showed
that the participants found the colored and masked images
more detailed but also a lot harder to recognize compared
to the depth images, thus they needed more attention for the
former.
III. TES T APP LI CATI ON
The aim of our study was to determine the user’s preferences
for warning systems as well as the user’s willingness to
utilize such systems. We developed a study under conditions
which matched real life conditions as closely as possible. The
participants were, therefore, put into a situation in which they
got distracted by a smartphone while walking a certain route.
The route was prepared with different types of obstacles.
The study by Haga [9] indicates that games are causing
the strongest distraction for walking smartphone users in
comparison to other applications. In addition, it is easy to
control the overall experience of the participant in a game-
like scenario and thus create an ongoing and controlled level of
distraction, as such we decided to use a game-like application
for our study.
To fulfill our needs it was necessary to develop our own
specialized prototype. Displaying warnings, fine-tuning its
difficulty and capturing data would have been difficult with an
existing smartphone application. Using a self-developed game
also rules out user familiarity, which would very much conflict
with the expected level of distraction and comparability of
achieved scores.
The developed prototype consists of two parts:
•The game which keeps the participants busy and displays
the warning screens (see Figure 1)
•The remote-control which sends commands to the game
and was controlled by the study director (see Figure 2).
Both parts of the prototype are developed in Unity [15].
The source codes of both applications are available under the
following references: [16], [17]. The commands and messages
are transferred via representational state transfer (REST) [18]
over a mobile data connection. The delay between sending
a command on the remote and the execution on the game
application was almost instantaneous in pre-tests. During the
user study these results were reproduced by the test team by
directly testing the functionality on location and doing dry runs
of the obstacle course. The participants’ obvious reactions to
101010

Fig. 1: Screenshot of the game: right—game interface, left—
an in-game warning (The warning reads: “Attention! Obstacle!
[pictogram] Heads-up!”)
the warnings in relation to the warning being sent by the study
director indicated no significant delay. This is backed up by the
participants confirming in casual chat after the test that they
indeed reacted in response to a warning in specific moments.
A. The Game
The game is a simple balancing exercise, the task is to keep
a red ball in the middle of the screen. The optimal position
is indicated by a circled gradient in the background of the
game, as well as a visible score (see Figure 1). Because the
participants could simply hold the phone steady to balance
the ball it was necessary to add random impulses in random
directions that must be actively compensated for in order
to attain a decent score and keep the participants’ attention.
If the ball touches the sides, points are deducted which is
accompanied by a red flashing border. The random impulses
combined with the point deduction urges the participant to
focus their attention on the game and thus creates a constant
level of distraction.
As we have reviewed in the section about previous work,
sensing dangerous situations in the environment using smart-
phones can be considered as work in progress. Therefore we
decided to simulate this behavior by the well-known Wizard
of Oz method [19]. The events are triggered by an unseen
operator and led the participants to believe that the computer
system acts autonomously.
If the game receives the command to show a warning, a
full screen warning is shown for 5 seconds as in Figure 1
and the phone vibrates for 0.5 seconds, a tap on the screen
hides the warning message again. The written words and the
pictogram are necessary since the participants are not told of
the possibility that they are being warned about obstacles.
If for example the warnings were just the vibration, the
participant might be confused and probably would not relate
the vibration to any sort of warning. This means that at least
for the first warning event the participant learns about the
warning system, and the subsequent warnings can then be
considered as a learned experience.
Fig. 2: Screenshot of the remote-control application
The game logs the current score every second, as well as
any message or command it receives from the remote.
B. The Remote-Control
The remote is used by the study director to send commands
to the game application, the obvious main function is to send
the command to display warnings. Every command produces
a log entry on the game application.
IV. USE R STU DY
To test the users within potentially dangerous real-life
situations (events) without endangering the participants, we
carefully designed an obstacle course in a fully protected
and controlled environment on campus. The participants were
filmed during the test and events were logged. This provided
the material for later analysis which was completed by a
questionnaire filled out by the participants after walking the
obstacle course.
A. Participants
We focused our study mostly on young adults, since re-
search shows that they are the demographic group with the
highest risk [20]. In total, 67 randomly chosen participants
took part in our study and were rewarded with five Euros for
participation. A detailed overview is provided in Table I. All
participants personally owned a smartphone and were used to
handle it.
Participants were separated into two main groups and one
control group:
•Warning: The participants were asked to walk through
the obstacle course while playing the game. During the
walk they were given on screen warnings of upcoming
obstacles.
Warning No Warning Control Sum
Female 8 12 6 26
Male 15 12 14 41
Total 23 24 20 67
TABLE I: Overview of user groups and distributions.
111111

E1: Pedestrian E2: Bike E3: Boxes E4: Car
Fig. 3: Events on the obstacle course
•NoWarning: This group is identical to the Warning group,
except they did not receive any warnings.
•Control: The participants were asked to answer a ques-
tionnaire only without walking or playing.
B. Obstacle Course
The length of the route was about 400 meters and it took
each participant between three to five minutes to complete
it. Four different events were set up to simulate potentially
dangerous situations.
C. Events
The first two events, see E1 and E2 in Figure 3 and 4, on the
way to the turning point were a cyclist and a pedestrian, trying
to collide with the participant. They switched their position
after every participant to eliminate a possible bias introduced
by the first event (by getting to know the first warning).
The actor assigned for the particular event was instructed to
almost walk/drive into the participant but to avoid a collision at
the last possible but safe moment. The third and fourth events,
see E3 and E4 in in Figure 3 and 4, had fixed positions and
where placed on the way back from the turning point. Event
E3 introduced boxes on the ground, which were placed after
the participant passed for the first time. The fourth event was
a car backing out of a parking lot into the path the participant
would take, but stopping well before he would pass behind it.
The pathway as well as the events were designed to resemble
everyday scenarios that may have already been experienced by
some of the participants themselves. To cover a broad variety
of potential situations the events featured different aspects and
characteristics, fast (bike), slow (pedestrian, car) and static
obstacles (boxes) or loud (car) and quiet (bike, pedestrian,
boxes) events.
D. Test Team
A team consisting of ten people was in charge of the test on
site: one person each was in charge of one of the four events.
Three people followed the participant at distance while she
or he was walking through the route (study director, camera
operator and one supporter for documentation). Two people
directed the participants to answer the survey after passing
the route and one person acquired new participants by asking
students on campus.
E. Safety Measures
The study was designed such that no participant was ever
exposed to a dangerous situation by taking part in this test.
Every member of the test-team was instructed to be vigilant
of the situation and to intervene if a real possible dangerous
event could occur, namely any situation which was not part
of our test (like other cars). All team members were remitted
to avoid touching the participant let alone causing a collision
with them.
F. Test Procedure
The test was structured in the following way:
1) The participant is welcomed to the test and asked what
they already know about what is going to happen to
ensure they don’t know anything about the process.
2) The participant is introduced to the test and is told that
the aim of the study is to evaluate smartphone gaming
and to get as many points as possible relative to the
time they take for the given route while walking at their
regular speed.
Start | Finish
Turning
Point
E4
E1/2
E1/2
E3
Fig. 4: Map of the obstacle route [21]
121212

3) They are introduced to the route so they can follow the
path by themselves. They are not told about any of the
events.
4) The smartphone is handed to the participant and they are
able to familiarize themselves with the game for about
one minute.
5) The study director instructs the participant to start and
sends the start command via the remote. The participant
starts walking.
6) The Warning group receive warnings by triggered events
from the study director just before an event occurs. The
NoWarning group runs into events without warning.
7) Just after an event came up, the study director sends the
command to log the event.
8) At the end of the test, the study director sends the end
command which ends the game for the participant.
9) The participant is asked to answer the questionnaire.
G. Questionnaire
The questionnaire was divided into different topics:
•Demographic questions
•Questions about smartphone usage
•Questions about smartphone distraction
•Questions about the provided prototype
•Questions about the properties of the provided prototype
•Questions about laws and penalties for smartphone re-
lated distracted walking
The questionnaire was set up to evaluate its outcome by
mostly using the Likert Scale [22]. These questions were
separated into four possible answers: strongly agree (4), agree
(3), disagree (2) or strongly disagree (1). Additionally, there
were multiple choice and free text answers.
The most relevant questions asked in the questionnaire were:
1) Have you ever had any incidents while using your
smartphone while walking?
2) Do you use your smartphone while walking outside?
3) Would you consider installing an app which warns you
of upcoming dangerous situations?
4) How would you classify the following characteristics
for such an app regarding: reliability, unobtrusiveness,
privacy, energy efficiency and if it needs to be easy to
use.
5) What kind of warning categories do you find useful?
Life-threatening situations, dangerous situations, possi-
bly dangerous situations, areas with high traffic, rush
hours.
6) How should a potential warning look? Pop-up, overlay,
display brightness, screen turning black, auditory signal,
vibration, permanent overlay, none, other.
V. RE SU LTS
The bulk of the study’s data was gathered from three
sources. First, the study’s participants were filmed while pass-
ing the obstacle course. Second, the game itself stored log data.
The log data files contain the information which describe how
many points were gathered in relation to the time a participant
needed to complete the obstacle course. Afterwards, each
participant was asked to answer a questionnaire.
To evaluate the video data, every noticeable movement of
the participants head which can be interpreted as a way of
looking up/around, was counted and brought into relation to
the other two data sources. Note that we found a hand selection
of look-ups to be more reliable in comparison to gaze tracking
and decided on the former for this reason. If an event in a video
were not clearly recognizable for a moment (e.g. bad lighting
conditions) the participants average look-ups per second were
used for evaluation.
The following floating point numbers represent Likert
Scale’s average of the given answers. This means all values
higher than 2.5 show an agreement by over 50% of the
participants.
A. Acceptance
Table II summarizes the answers to the question: Would
you consider installing an app which warns you of upcoming
dangerous situations? The participant’s statements for this
question shows a general attitude towards the use of safety
applications. The average score 2.6 states that approximately
only half of the participants would install and use such an
application. It is interesting to note that this is particular
surprising because nearly all the participants (3.7) agreed that
it could be dangerous to use a smartphone while walking
(Cohen’s d≈1.23).
That means even though the smartphone users are aware of
the danger, some are still using the smartphone while walking
(all participants use their smartphone at least ‘rarely’ while
walking outside) but would not be willing to install a safety
application. But why is that? If we compare to other safety
systems such as bike helmets or safety belts they clearly reduce
the comfort. This, however, is not the case for the app and thus
is more comparable to other safety systems such as airbags.
The energy efficiency could be a reasonable cause. An average
Likert scale of 3.3 for energy efficiency indeed indicates that
people are concerned about a reduction in usage duration.
Warning NoWarning Control Sum
Female 3.1 (1.3) 2.4 (1.2) 2.7 (0.8) 2.7 (1.2)
Male 2.6 (0.8) 2.4 (1.3) 2.4 (1.2) 2.5 (1.1)
Total 2.9 (1.0) 2.4 (1.3) 2.5 (1.1) 2.6 (1.1)
TABLE II: Acceptance of the Applications (values in brackets
represent the standard deviations).
had an incident while using smartphone
str. disagree disagree agree str. agree
Female 2.5 (1.3) 2.4 (0.6) 2.9 (1.4) 2.8 (1.3)
Male 2.2 (1.5) 2.0 (1.0) 2.7 (0.9) 2.8 (1.0)
Total 2.3 (1.3) 2.2 (0.9) 2.8 (1.1) 2.8 (1.1)
TABLE III: Acceptance of a safety application in relation to
potential incidents already happened to the participants using
a smartphone while walking (values in brackets represent the
standard deviations).
131313

Tactile Acoustic Visual
Pop-Up Overlay Brightness Turning Black
Activate 75% 51% 36% 22% 9% 10%
Deactivate 16% 31% 33% 30% 42% 66%
Difference 58% 19% 3% -7% -33% -55%
TABLE IV: Type of Warning.
Life-Threatening Dangerous Potentially Dangerous High Traffic Area Dangerous Hour
Female 3.92 3.81 3.12 3.19 2.81
Male 3.85 3.73 3.10 3.12 2.49
Total 3.88 3.76 3.10 3.15 2.61
Warning 3.96 3.83 3.17 3.30 2.61
No Warning 3.83 3.83 3.04 2.92 2.75
Control 3.85 3.60 3.10 3.25 2.45
TABLE V: Acceptance of warning categories for a potential safety application sorted by groups.
Reliability Easy to Use Energy Efficiency Privacy Unobtrusiveness
Female 3.9 3.5 3.3 3.4 3.0
Male 3.8 3.4 3.3 3.2 2.7
Total 3.8 3.4 3.3 3.3 2.8
TABLE VI: Characteristics of a safety app.
Comparing the acceptance for the different groups it is
immediately apparent that the group of participants with warn-
ings had somewhat higher values (2.78) as compared to the
two other groups NoWarning (2.42) and Control (2.45). The
similarity of the latter two (Cohen’s d≈0.03) demonstrates that
our test setup did not influence the attitude towards a safety
system for smartphones. Using Cohen’s [23] interpretation of
the effect size, between the Warning and the other two groups,
d≈0.17, the differences can be rated as small. In absolute
numbers, however, this means an increase of approximately
10% from 46% for those not experiencing the warnings to
56% for those who did.
In comparison to the use of other activities on the smart-
phone which are done at least weakly both values 46% as
well as 56% are well above average: [24] Only search engines
are used (50%) and social networks are visited (47%) alike.
Even email (31%), watching videos (22%, look up maps and
directions (20%) are already way less used. And the two so
called ‘killer applications’ listen to music (16%) and playing
games (8%) are not even used by every fifth smartphone user.
This demonstrates that such kind of security applications have
the potential to be one of the most frequently used applications
on smartphones.
The values regarding the acceptance of a safety application
show a small discrepancy (Cohen’s d≈0.20) between men
and women. One reason for this could be the male tendency
to make risky choices [25]. It could also be explained by
women’s self-efficacy issues in testing situations [26]. Another
reason for the increased acceptance among females could be
that women tend to be more cautious overall, for example
women wear cycling helmets more often than men [27]. This
tendency seems to hold in our study as well.
It is also notable that participants who already experienced
incidents (question: Have you ever had any incidents while
using your smartphone while walking? answers: strongly agree
(4) and agree (3)) as a result of being distracted by their
smartphone (compare Table III, had on average a moderate
higher acceptance (Cohen’s d≈0.51) to use the investigated
safety application in comparison to the group of participants
who disagreed (2) or strongly disagreed (1) to this question.
The reasons why they would consider installing an appli-
cation which warns them of upcoming dangerous situations
were diverse:
•Accept help when it is offered.
•It is helpful when you are distracted by your smartphone.
•I don’t want to die just because I was careless.
•Everyone is distracted from time to time so it would be
helpful.
•In the test I was warned of obstacles which I had not
noticed.
B. Type of Warning
The use or acceptance of safety applications strongly de-
pends on the type of warning. Table IV summarizes the type
of notification the users would activate or deactivate. We found
that the results didn’t change much over the different inves-
tigated groups and thus did not present it in the table. While
75% would activate tactile (vibration) warnings, only 51% of
the participant would activate acoustic warnings. Surprisingly
all types of visual warnings would not be activated by the
majority of users. This effect becomes in particular visible by
comparing the types of warning which the users would activate
to those who would be deactivated. In general it can be said
that visual warnings would be deactivated by at least the same
141414

amount of people than would activate visual warnings. In our
opinion this is an interesting finding, because to the best of
our knowledge, all developed systems so far have used some
kind of visual feedback, some in combination with acoustic
feedback. Therefore, the acceptance of such systems could
possibly improved by preventing visual (at least on the main
display) and by employing tactile warnings.
C. Potential Dangers
We asked participants for potential dangerous conditions
they wanted to be warned about: life-threatening situations
(3.88) had the highest values followed by dangerous situations
(3.67), possibly dangerous situations (3.10), dangerous areas
(e.g. high traffic) (3.15), dangerous times (e.g. rush hour)
(2.61) see Table V. These values show the participants’ attitude
towards different forms of risk management. They don’t want
to get warnings in general, the warnings need to be classified
exactly. Especially the term possibly dangerous situations
caused the participants to accept the warning form less than
an explicit warning like dangerous areas, although a possibly
dangerous situation could be more relevant regarding the
hazard level. As the low value for dangerous times shows,
users prefer time relevant and explicit notifications in their
individual situations not for dangerous times in general.
D. Characteristics of a Safety Application
We also asked them to select their desired characteristics for
such a system. As expected users attached the most importance
to reliability followed by simple use,privacy,energy efficiency
and unobtrusiveness. Reliability of course has to be the main
characteristic for a safety system, especially if a person’s life
could be on the line (see Table VI).
E. Situation Awareness
By comparing the average look-ups in Table VII between
the two user groups Warning (47.2) and NoWarning (40.2)
the results might be stumbling because intuitively one would
expect that persons which rely on the warning messages (trust
in the reliability of the system) would look-up less or at least
not more than those without warnings. Besides the difference
between those two groups also male (47.9) and female (38.7)
show a big gap. By compensating for the inequality between
the male and female distribution between the two groups we
get an sex-compensated look-up for Warning of 41.7. A num-
ber which is still larger as the comparison group. Calculating
Cohen’s d which is d≈0.18 for the non-compensated numbers
reads now d≈0.09. And thus those differences cannon be
considered as relevant. It can be concluded that the safety
system had no negative influence on the user’s situation
awareness. It should be noted that the system had been used
only for a very short time and thus a long-term effect on this
user behavior cannot be determined by our experiments.
VI. CONCLUSION
We have investigated the willingness of smartphone users to
use safety applications and to be warned by them in potentially
Warning NoWarning Sum
Female 44 (19) 36 (18) 39 (18)
Male 50 (21) 45 (17) 47 (19)
Total 47 (17) 40 (17) 44 (19)
TABLE VII: Average times participants looked up (values in
brackets represent the standard deviations).
dangerous situations. We found that approximately half of the
suspects would be willing to use such an application. Two
main factors who are influencing their decision are:
•Previous experience of a real-life incidents caused by
distractions from smartphone usage.
•The experience of a reliable safety system.
While half the users might seem to be low, in comparison
to other apps it becomes apparent that it is indeed a pretty
high number and that it has the potential to become the next
‘killer’ application. While most developed systems in the labs
use some kind of visual feedback we found that tactile before
acoustics would be the preferred choice of the users—also in
combination. How this feedback can be provided and clearly
distinguishable from other signals are interesting research
questions. Also if it is necessary to give the direction of the
possible danger of if it is enough to warn without direction is
an open question.
We also want to point out that fully trusting and relying on
safety systems might bring new dangers: People think they do
not longer have to be careful about the surrounding because
they belief the system is ‘watching for them’. However, the
sensors or pattern recognition applications could malfunction-
ing or the system might be turned off. Todays systems are
not very sophisticated and maybe not convincing in terms of
accuracy, but in the long run we belief that todays limitations
can be overcome. Wouldn’t any prevented accident, harmful
or not, be a very good reason to further develop and use such
systems?
REFERENCES
[1] eMarketer, “Number of smartphone users worldwide from 2014 to
2019: cited by statista.com,” 2015. [Online]. Available: http://www.
statista.com/statistics/330695/number-of-smartphone-users-worldwide/
[2] J. Nasar, P. Hecht, and R. Wener, “Mobile telephones, distracted atten-
tion, and pedestrian safety,” Accident; analysis and prevention, vol. 40,
no. 1, pp. 69–75, 2008.
[3] M. B. Neider, J. S. McCarley, J. A. Crowell, H. Kaczmarski, and A. F.
Kramer, “Pedestrians, vehicles, and cell phones,” Accident; analysis and
prevention, vol. 42, no. 2, pp. 589–594, 2010.
[4] H. Tapiro, T. Oron-Gilad, and Y. Parmet, “Cell phone conversations and
child pedestrian’s crossing behavior: A simulator study,” Safety Science,
vol. 89, pp. 36–44, 2016.
[5] J. L. Nasar and D. Troyer, “Pedestrian injuries due to mobile phone use
in public places,” Accident; analysis and prevention, vol. 57, pp. 91–95,
2013.
[6] U.S. Consumer Product Safety Commission, “National electronic
injury surveillance system (neiss),” 2015. [Online]. Available: http:
//www.cpsc.gov/en/Research--Statistics/NEISS-Injury-Data/
[7] Ipsos Public Affairs, “Distracted walking study: Topline summary find-
ings.” [Online]. Available: http://www.anationinmotion.org/wp-content/
uploads/2015/12/AAOS-Distracted-Walking-Topline-11-30- 15.pdf
151515

[8] D. C. Schwebel, D. Stavrinos, K. W. Byington, T. Davis, E. E. O’Neal,
and D. de Jong, “Distraction and pedestrian safety: How talking on
the phone, texting, and listening to music impact crossing the street,”
Accident; analysis and prevention, vol. 45, pp. 266–271, 2012.
[9] S. Haga, A. Sano, Y. Sekine, H. Sato, S. Yamaguchi, and K. Masuda,
“Effects of using a smart phone on pedestrians’ attention and walking,”
Procedia Manufacturing, vol. 3, pp. 2574–2580, 2015.
[10] Z. Zhou, “Headsup: Keeping pedestrian phone addicts from dangers
using mobile phone sensors,” International Journal of Distributed Sensor
Networks, vol. 2015, pp. 1–9, 2015.
[11] X. Liu, J. Wen, J. Cao, and S. Tang, “Infrasee: An unobtrusive alertness
system for pedestrian mobile phone users,” IEEE Transactions on Mobile
Computing, p. 1, 2016.
[12] J. Wen, J. Cao, and X. Liu, “We help you watch your steps: Unobtrusive
alertness system for pedestrian mobile phone users,” in 2015 IEEE
International Conference on Pervasive Computing and Communications
(PerCom), 2015, pp. 105–113.
[13] J. D. Hincapi´
e-Ramos and P. Irani, “Crashalert: enhancing peripheral
alertness for eyes-busy mobile interaction while walking,” in the SIGCHI
Conference, W. E. Mackay, S. Brewster, and S. Bødker, Eds., 2013, p.
3385.
[14] Z. Wei, S.-W. Lo, Y. Liang, T. Li, J. Shen, and R. H. Deng, “Automatic
accident detection and alarm system,” in the 23rd ACM international
conference, X. Zhou, A. F. Smeaton, Q. Tian, D. C. Bulterman, H. T.
Shen, K. Mayer-Patel, and S. Yan, Eds., 2015, pp. 781–784.
[15] Unity Technologies, “Unity - game engine.” [Online]. Available:
http://unity3d.com/
[16] A. Scheurer and S. Beuck, “Twiddlingphone,” 2016. [Online]. Available:
http://dx.doi.org/10.5281/zenodo.56473
[17] ——, “Twiddlingremote,” 2016. [Online]. Available: http://dx.doi.org/
10.5281/zenodo.56472
[18] R. T. Fielding, “Architectural styles and the design of network-based
software architectures,” Ph.D. dissertation, University of California,
Irvine, 2000-01-01.
[19] B. Martin and B. M. Hanington, Universal methods of design: 100
ways to research complex problems develop innovative ideas and design
effective solutions. Beverly, MA: Rockport Publishers, 2012. [Online].
Available: http://proquest.tech.safaribooksonline.de/9781592537563
[20] W. Niew¨
ohner, S. Ritter, D. Wickenkamp, A. Ancona, I. Briki,
K. Koch, M. Markmann, D. M¨
uller, and M. Niew¨
ohner, “Fußg¨
anger
und ihr nutzungsverhalten mit dem handy/smartphone in europ¨
aischen
hauptst¨
adten: Verkehrsbeobachtung.” [Online]. Available: http://www.
dekra-roadsafety.com/de/fussgaenger-ablenkung-smartphones/
[21] OpenStreetMap, “Furtwangen university, furtwangen im schwarzwald,
germany.” [Online]. Available: http://www.openstreetmap.org/#map=19/
48.05136/8.20872
[22] R. Likert, “A technique for the measurement of attitudes,” Ph. D.,
Columbia University, New York, NY, 1932.
[23] J. Cohen, Statistical Power Analysis for the Behavioral Sciences. Erl-
baum, Hillsdale, NJ: Routledge; 2 edition, 2012.
[24] Consumer Barometer, “What online activities do people do
on their smartphones at least weekly?” [Online]. Avail-
able: https://www.consumerbarometer.com/en/graph-builder/?question=
M7b1&filter=country:germany
[25] D. Herrero-Fern ´
andez, P. Mac´
ıa-Guerrero, L. Silvano-Chaparro,
L. Merino, and E. C. Jenchura, “Risky behavior in young adult pedestri-
ans: Personality determinants, correlates with risk perception, and gender
differences,” Transportation Research Part F: Traffic Psychology and
Behaviour, vol. 36, pp. 14–24, 2016.
[26] D. C. Zhang, S. Highhouse, and T. B. Rada, “Explaining sex differences
on the cognitive reflection test,” Personality and Individual Differences,
vol. 101, pp. 425–427, 2016.
[27] Institut f¨
ur Demoskopie Allensbach, “Fahrradfahren: Lieber ohne Helm,”
Allensbach, Germany. [Online]. Available: http://www.ifd-allensbach.
de/uploads/tx reportsndocs/PD 2013 08.pdf
161616