Content uploaded by Göran Falkman
Author content
All content in this area was uploaded by Göran Falkman on Mar 11, 2015
Content may be subject to copyright.
210
Proceedings of the 5th International Conference on Automotive User Interfaces and
Interactive Vehicular Applications (AutomotiveUI ‚13), October 28–30, 2013, Eindhoven, The Netherlands.
Presenting system uncertainty in automotive UIs for
supporting trust calibration in autonomous driving
Tove Helldin
University of Skövde
54128, Skövde,
Sweden
+46 500 44 83 83
tove.helldin@his.se
Göran Falkman
University of Skövde
54128, Skövde,
Sweden
+46 500 44 83 35
goran.falkman@his.se
Maria Riveiro
University of Skövde
54128, Skövde,
Sweden
+46 500 44 83 43
maria.riveiro@his.se
Staffan Davidsson
Volvo Car Corporation
405 31,Gothenburg,
Sweden
+46 31 59 98 24
sdavidss@volvocars.com
ABSTRACT
To investigate the impact of visualizing car uncertainty on drivers’
trust during an automated driving scenario, a simulator study was
conducted. A between-group design experiment with 59 Swedish
drivers was carried out where a continuous representation of the
uncertainty of the car’s ability to autonomously drive during snow
conditions was displayed to one of the groups, whereas omitted
for the control group. The results show that, on average, the group
of drivers who were provided with the uncertainty representation
took control of the car faster when needed, while they were, at the
same time, the ones who spent more time looking at other things
than on the road ahead. Thus, drivers provided with the
uncertainty information could, to a higher degree, perform tasks
other than driving without compromising with driving safety. The
analysis of trust shows that the participants who were provided
with the uncertainty information trusted the automated system less
than those who did not receive such information, which indicates
a more proper trust calibration than in the control group.
Categories and Subject Descriptors
H.1.2 [User/Machine Systems]: Human factors
General Terms
Design, Human Factors, Experimentation
Keywords
Uncertainty visualization, trust, automation, driving, acceptance.
1. INTRODUCTION
Technological advances have led to the development of numerous
driver assistance systems such as adaptive cruise control, lane
departure warning, collision avoidance, automatic parking and
driver drowsiness detection systems. Experiments with fully
autonomous cars have also been carried out, which provides us
with a glimpse of what the future might hold. The purposes of
developing such support systems are to make driving safer, easier,
more relaxing and more enjoyable. However, such goals can only
be achieved if the driver feels comfortable enough to hand over
control to the automation and if a good cooperation between the
driver and the automation can be achieved. Studies from other
domains, such as the aviation domain, have shown that the
anticipated positive effects of automation might be diminished
due to human-automation cooperation related problems, such as
automation misuse and disuse [1, 2], automation surprises and
mode confusion [3, 4], reduced situation awareness [5],
complacency as well as over reliance on the automation [6]. To
reduce the possible negative effects of automation, while at the
same time reinforce the positive ones and promote a safe and
appropriate usage of the automation, several researchers have
highlighted the importance of informing the human operators of
the strengths and limitations of the automated systems used, as
well as the continuous state of the automation (see for instance [7-
11]). For example, in the study by Stanton and McCaulder [12], it
became evident that the drivers had insufficient knowledge of the
limitations of the adaptive cruise control system, resulting in
collisions due to the drivers’ inappropriate levels of trust in the
automated system. This finding is in line with the research
reported by Dzindolet et al. [13]and McGuirl and Sarter [7] where
it was found that operators who were provided with continuous
feedback regarding the performance of the automated aid had
more appropriate trust in the aid than operators who were not
given such information.
We argue that more research is needed to evaluate the
effectiveness of providing feedback on changes in the automated
system’s capability during autonomous driving. To the authors’
knowledge, no research has addressed how to convey the limits
and performance of automatically driven cars to their drivers as a
means to achieve appropriate trust. As such, the objective of this
study was to evaluate the effects of visualizing a continuous
representation of car uncertainty on the drivers’ trust in the
automatic support system used. First, we wanted to assess if such
visualization would make the drivers able to more appropriately
calibrate their trust in the system while at the same time making
them aware of the limitations of the system. Secondly, due to one
of the motivations for introducing automation in cars, namely to
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that
copies bear this notice and the full citation on the first page. Copyrights
for components of this work owned by others than ACM must be
honored. Abstracting with credit is permitted. To copy otherwise, or
republish, to post on servers or to redistribute to lists, requires prior
specific permission and/or a fee. Request permissions from
Permissions@acm.org.
AutomotiveUI '13, October 28 -30 2013, Eindhoven, Netherlands
Copyright 2013 ACM 978-1-4503-2478-6/13/10…$15.00.
http://dx.doi.org/10.1145/2516540.2516554
211
Proceedings of the 5th International Conference on Automotive User Interfaces and
Interactive Vehicular Applications (AutomotiveUI ‚13), October 28–30, 2013, Eindhoven, The Netherlands.
enable drivers to feel more relaxed while driving and even to
perform other things while travelling, we also wanted to
investigate if displaying the uncertainty representation would
result in a higher number of drivers performing other tasks than
driving during the test scenario.Finally, we also wanted to test if
the drivers being presented with the uncertainty information
would look away more from the road than the drivers not being
provided with this information, while at the same time being
better prepared to take control over the car when/if needed (i.e.
requiring less time to take manual control over the car).
The paper is structured as follows: section 2 provides information
regarding advances within the area of uncertainty and system
reliability visualization. Section 3 presents the study setup,
whereas section 4 reports on the study findings and section 5
presents a brief analysis of the results. Section 6discusses the
results obtained whereas the conclusions and ideas for future work
are found in section 7.
2. RELATED WORK
The visualization of uncertainty in the context of automatic
driving has been recently studied by Beller et al. [14]. The aim of
this study was to evaluate whether communicating when the car
was uncertain using a symbol (a face with an uncertain
expression) improved the driver-automation interaction. A driving
simulator experiment varying the level of uncertainty with 28
participants was conducted. The results show that the presentation
of uncertainty information increasedthe time to collision in cases
of automation failure, that situation awareness was improved and
that automation with the uncertainty symbol received increased
acceptance and higher trust ratings. These positive results
regarding the visualization of information related to smart systems
in cars seem to coincide with two previous studies, i.e., Verberne
et al. [15]and Seppelt and Lee [8].
Seppelt and Lee [8] investigated if a visual representation of the
adaptive cruise control (ACC) behavior promote appropriate
reliance and support effective transitions between manual and
ACC control. Twenty-four participants were recruited to drive in
two different situations, with different failure types. In traffic
conditions, the participants relied more appropriately on ACC
when the information about the ACC was present. Moreover, it
promoted faster and more consistent braking responses and show
additional positive effects in other traffic situations. The authors
suggest that providing drivers with continuous information about
the state of the automation is a promising alternative to providing
warnings.
The work presented by Verberne et al. [15]focuses on
investigating if representations of descriptors of three ACCs with
different automation levels that either shared their driving goals or
not affected trustworthiness and acceptability of those systems. A
driving experiment with 57 participants was carried out. The
results show that ACCs that took over driving tasks while
providing information were more trustworthy and acceptable than
ACCs that did not provide information.
Several relevant works regarding the influence of uncertainty
visualization on decision-making can be found in other research
areas, such as the military domain. For example, Finger and
Bisantz [16]studied the use of blended and degraded icons to
represent uncertainty regarding the identity of a radar contact as
hostile or friendly. The first part of the study showed that
participants could sort, order and rank five different sets of icons
conveying different levels of uncertainty. In the second part of the
study, three of the pairs of icons were used in an application in
which participants should identify the status of contacts as
friendly or hostile. Three conditions were studied: with degraded
icons and probabilities, with non-degraded icons and probabilities
and with degraded icons only. The results demonstrate that
participants using displays with only degraded icons performed
better on some measures and as well on other measures, than the
other tested conditions. Thus, the use of distorted or degraded
images may be a viable alternative to convey situational
uncertainty.
Wang et al. [17]examined the effects of presenting the aid
reliability on trust and reliance on a combat identification (CID)
scenario. Twenty-four participants carried out a simulated CID
task, half of whom were told the reliability level. The results
show that response bias varied more appropriately with the aid
reliability when it was disclosed than when not, and that trust in
aid feedback correlated with belief in aid reliability. The authors
highlight that to engender appropriate reliance on CID systems,
users should be made aware of system reliability.
3. METHOD
3.1 Participants
A total of 59 participants (31 male, 28 female) between 28and 58
years old (4 between 21–30 years, 22between 31–40, 25between
41–50 and 8between 51–60) with an average age of 41,2 years
took part in the simulator experiment.The participants were
selected from a population of 488 Volvo employees, mostly non-
technical personnel of whom none is involved in the development
of functionality for autonomous driving or the implementation of
the driver’s information module (DIM). The only prerequisite for
taking part in the study was that the participant had a driver’s
license.
Each participant was randomly assigned to a display condition. A
balanced latin square design was used in order to minimize the
effects of participants driving early in the morning, directly after
lunch and late in the afternoon. This led to 30 participants (16
males and 14 females) driving with the added uncertainty
information and 29 participants in the control group.
3.2 The DIM
Two interfaces were designed –one with and one without the
uncertainty representation. Figure 1shows a sketch of the DIM
design including: the speedometer, the engine speed, the fuel
level, the outside temperature, the current time, the current gear
used, the placement of the steering wheel during the autonomous
drive as well as the ability of the automation to maneuver the car.
During the experiments performed together with the control
group, the information regarding the automation ability was taken
away. Figur e 2depicts two states of the ability of the automation,
from high ability (left figure) to low ability (right figure) as
indicated by the color/transparency of the 7 levels, as well as the
red arrow, representing the threshold where the ability of the
automation no longer can be guaranteed.
212
Proceedings of the 5th International Conference on Automotive User Interfaces and
Interactive Vehicular Applications (AutomotiveUI ‚13), October 28–30, 2013, Eindhoven, The Netherlands.
Figure 1 - Sketch of the DIM used during the experiments
(including the representation of uncertainty).
Figure 2 - Graphical representation of the ability of the car to
drive autonomously, ranging from 7 (very high ability, left
figure), to 1 (no ability, right figure). The red marking
indicates the threshold for when the performance of the
automated driving system no longer can be guaranteed.
The DIM was placed in the instrument cluster of the car in front
of the driver. The design of the interfaces was carried out in
collaboration with an expert HMI designer at Volvo cars. As such,
we argue that its design is similar to other interfaces used in
Volvo cars.
3.3 Procedure and questionnaire
The participants were first informed of the purpose and setup of
the study. Thereafter, all participants were allowed to drive the
car simulator in manual mode for about 3–5 minutes so as to get
acquainted with the simulator. Directly after the training session,
the participants were informed of the prerequisites of the test
session: that the car could drive autonomously,but that the
performance of the automatic driving system was coupled to the
weather conditions. The participants were also informed that they
could at any time take control over the car by
steering/braking/giving gas to the car in accordance with their own
assessment of the appropriateness of using the system. The DIM
was explained to both of the groups; however, the uncertainty
representation (see Figure 1) was presented and explained to only
one of the groups (hereafter “with uncertainty information
group”).Before the start of the test session, the participants were
informed that there were newspaper and sweets at their disposal
in the passenger seat if they so pleased. Thereafter, the 9-minute
test session started.
After the test session, the participants were asked to fill out a
questionnaire about their trust in the system, using a modified
version of the trust in automation scale [18]. The participants
answered seven questions such as “I am confident in the system”
and “I can trust the system” using a seven point Likert scale
ranging from 1 (fully disagree) to 7 (fully agree). The instructions
given to the drivers and the questionnaire can be found in the
appendix.
3.4 Simulator
The experiments were carried out at the Human Machine
Interaction (HMI) laboratory at Volvo Car Corporation,
Gothenburg, Sweden. The laboratory contains several integrated
systems: a driving simulator and a fully functioning cockpit (see
Figures 3 and 4).
Figure 3 - Driving simulator, Volvo Car Cooperation HMI lab.
Figure 4 - Overview of HMI lab
The participants drove the car simulator through a snowed two-
lane country-side road, with a number of sharp turns, but with no
other traffic (see Figure 5). Due to the weather conditions, the
intensity of snowing varied from 0% to 100% where the
maximum amount of snowing is illustrated in Figure 6, and the
minimum amount (0%) corresponds to a clear sky with full
visibility. The snowing intensity varied according to Figure 7. The
direction and speed of the car were controlled by the automation.
The speed of the car was independent of the snowing intensity
(see Figure 7), but did change at times according to road
conditions (sharp turns and hilltops). When the visibility was the
worst, the car simulator could no longer maneuver the car. At this
moment, the driver had to act accordingly by taking control of the
car (either braking or steering). Following the scenario used in the
experiments, the driver had to take manual control over the car in
a slight curve. At this event, the automation stopped working
213
Proceedings of the 5th International Conference on Automotive User Interfaces and
Interactive Vehicular Applications (AutomotiveUI ‚13), October 28–30, 2013, Eindhoven, The Netherlands.
completely,meaning that no gas or steering was provided, without
giving any warning to the driver (apart from the graphical
representation in Figure 2 and switching to the manual DIM).
Figure 5 - The route used during the tests
Figure 6 - Picture displaying the maximum amount of snowing
Figure 7 - The snowing intensity during the experiments
3.5 Collected data
Logs from each simulator session were recorded. The quantitative
data thus collected corresponds to values from steering angle,
brake, acceleration, look away time and weather conditions.
Cameras were used to record all the sessions.
In addition to the quantitative data, qualitative data was collected
through observing the participants. The data collected included to
which extent the driver had his/her hands on the steering wheel, if
the participant stayed on the road after take-over and if the
participant read the newspapers or ate the sweets provided. Time
to take-over (TTO) was calculated by analyzing the logged data,
measuring the time between the snowing intensity reaching its
maximum (1.0) and a significant change in either braking or
steering (indicating that the driver had taken control over the
vehicle).
4. RESULTS
Regarding time to take-over, the group provided with the
uncertainty representation needed 1.9 seconds to take control of
the car on average while the control group needed 3.2 seconds.
The individual results are shown in Figure 8below.
Figure 8 - Individual results of time to take over
The differences between the two groups are summarized in
Figur e 9below.
Figure 9–Summary of time to take over
The results were submitted to a one-way ANOVA analysis. The
analysis showed that, with a 95% certainty,there is a statistically
significant difference between the two groups [F(1, 57)]= 5.62,
p=0.02].
Regarding looking away from the road, the group provided with
the uncertainty representation, on average spent 18% of the
driving time looking at other things but the road, while the control
group was looking at other things 12% of the time on average. The
individual results are shown in Figure 10 below.
214
Proceedings of the 5th International Conference on Automotive User Interfaces and
Interactive Vehicular Applications (AutomotiveUI ‚13), October 28–30, 2013, Eindhoven, The Netherlands.
Figure 10 –Individual results of proportion of look away time
The differences between the two groups are summarized in
Figur e 11 below.
Figure 11 –Summary of proportion of look aw ay time
A one-way ANOVA analysis showed that, with a 95% certainty
there is a statistically significant difference between the two
groups [F(1, 57)]= 4.81 , p= 0.03].
In addition to the proportion of the total time spent on looking at
other things than the road ahead, the number of times drivers
looked away for more than 2 seconds were counted, since this is
regarded as the limit for how long a driver can look away while
maintaining awareness of the situation ahead [19]. The group
provided with the uncertainty representation looked away for
more than 2 seconds 8.0 times on average, while the control group
looked away 5.0 times on average (see Figure 12 below).
Figure 12 –Summary of look away periods > 2 seconds
A one-way ANOVA analysis (α = 0.05) showed that there is no
statistically significant difference between the two groups with
respect to looking away from the road ahead for longer periods of
time [F(1, 57)]= 2.54, p=0.12].
Trust was assessed using the scale for trust in automated systems
developed by Jian et al. [18]. Participants of both groups answered
the questions after the driving exercise, using a seven point Likert
scale ranging from 1 (fully disagree) to 7 (fully agree). The
questions are listed in the appendix. The results are shown in
figures 13-14. The mean of the scores was used as an overall trust
score (as presented by Beggiato and Krems [10]). The average
trust value for the control group was 5.30, while the group with
uncertainty representation shows an average trustworthiness of
4.89.
Reliability was measured using Cronbach’s alpha values. The
values obtained, 0.87 (with uncertainty representation) and 0.85
(control group) show a good internal consistency (0.8 ≤ α < 0.9).
Figure 13 –Whisker plot representation of the answers to the
trust questionnaire for both groups (with and without
certainty representation). Values are between 1 (min, fully
disagree) and 7 (max, fully agree).
215
Proceedings of the 5th International Conference on Automotive User Interfaces and
Interactive Vehicular Applications (AutomotiveUI ‚13), October 28–30, 2013, Eindhoven, The Netherlands.
Figure 14 –Answers to trust questions. The control group
scores higher in all the questions but the first one (“I
understand how the system works –its goals, actions and
output”)
The analysis of the participants’ responses regarding system’s
trustworthiness (see questions in the appendix) show that, on
average, the control group perceives the car without information
about uncertainty more trustworthy (mean=5.30 vs. mean =4.89).
5. ANALYSIS
The results show that presenting (un)certainty information results
in better prepared drivers in take-over situations. The difference in
look away times between the two groups manifested itself in that
of the 33 drivers that stayed on the road after take-over, 20 (61%)
were drivers provided with the uncertainty information.
Furthermore, the results show that drivers presented with
(un)certainty information look away from the road to a higher
degree. Although looking away more in terms of total time
compared to the control group, the drivers who were presented
with the uncertainty information did not look away for longer
periods of time more often.
Lastly, the collected qualitative data indicate that drivers who
were presented with the uncertainty information would perform
other tasks than driving during the test scenario. Of the 15 drivers
that read the newspapers, 9 (60%) were from the test group. Of
the 21 drivers that ate of the sweets, 11 (52%) were from the test
group. More important, of the drivers that read the papers and
drove off the road at take-over, only 20% were from the test
group. Of the 28 drivers that, to a lesser or greater extent, kept
their hand on the steering wheel during the test, 12 (43%) were
drivers provided with the uncertainty information. More
important, of the drivers that kept their hands on the wheel and
drove off the road at take-over, only 20% were from the test
group.
To summarize, the results show that drivers provided with the
uncertainty information performed better in take-over situations
and they were are also more comfortable with performing other
tasks while driving, as compared to drivers without this
information.
6. DISCUSSION
Results from the study show that the drivers who were informed
of the car uncertainty were better prepared in take-over situations.
Also, these drivers had better calibrated their trust in the
automatic driving system, whereas the control group reported on
higher trust ratings despite the needed manual take-over in the
scenario used in the training sessions. These findings are in line
with the work presented by McGuril and Sarter [7] where the
participants who were informed of the system confidence were
better able to more appropriately calibrate their trust in the
decision aid.
Even though the drivers with certainty information were better
prepared to take control of the car while, on average, spending
more time doing other activities, the results show that the trust
scores for this group were worse than the group without aid. The
findings reported in this paper are in contrast to the ones reported
by Beller et al. [14]and Seppelt and Lee [8], that recommend that
providing drivers with continuous information about automation is
preferable to providing warnings, and that information about
automation increase trust and acceptance. A possible explanation
for interpreting our results can be found in Dzindolet et al. [13],
where the role of trust in automation reliance is studied. Their
findings suggest that participants initially considered automated
decision aidstrustworthy and reliable, but, after observing the
automated aid make errors, participants distrusted even reliable
aids, unless an explanation was provided regarding why the aid
might err. Knowing why the aid might err increased trust in the
decision aid and increased automation reliance, even when the
trust was unwarranted. Thus, it should be further investigated if
the visual representation of the car uncertainty used in our study
should be complemented with additional information regarding
why the level of uncertainty was high.
Another representation of the car’s ability to autonomously drive
could have generated different results than the ones obtained.
According to Seppelt and Lee [8], not just any representation of
continuous information will enhance driving performance. In the
experiment presented in [8], it was concluded that the use of color
dilution to represent sensor degradation in a rain condition was
not an effective cue. Further, to combine a graphical
representation of uncertainty together with a haptic and/or sound
etc. could result in better performance regarding time to take over
times and longer look away times.
Individual differences in trust in automation and automation
reliance should be further explored. Lee and Moray [20,21]found
strong individual differences in automation use –some
participants were prone to use manual control, others were prone
to use automation. As such, future work should include a further
analysis of the results in relation to the participants’ estimated
locus of control and driving styles.
Several limitations of the study should be mentioned. The driving
scenario and exercise might be considered very simple, but it was
designed to analyze the effect of uncertainty visualization on trust
and how the drivers would take over the control of the car when
automation could no longer guarantee a safe drive, thus, we tried
to minimized other experimental variables that could affect the
driving task and affect our analysis on trust and automation
reliance. It might be that the simplicity of the scenario made some
of the participants in the test group neglect the uncertainty
representation and concentrated on the weather conditions instead.
Moreover, the uncertainty of the automation could have been
216
Proceedings of the 5th International Conference on Automotive User Interfaces and
Interactive Vehicular Applications (AutomotiveUI ‚13), October 28–30, 2013, Eindhoven, The Netherlands.
associated with additional parameters than the weather, such as
other contextual information, e.g. the traffic situation.
Regarding the validity of the study here presented we would like
to highlight that during the design of the experiment we ruled out
extraneous variables that could affect our study on trust issues and
automation making as simple scenario as possible (e.g. avoiding
dense traffic or overtaking situations) guaranteeing, thus, internal
validity. An experiment has external validity (generalizability) if
the results are not unique to a particular set of circumstances, but
generalizable. We are confident that the large number of
participants in this study, as well as the selection criteria applied,
make the results presented generalizable.
7. CONCLUSIONS
This paper has reported on an empirical study performed together
with 59 drivers in a simulator experiment, were the effects of
displaying continuous support system uncertainty during an
automated driving scenario was evaluated. The results indicate
that drivers who were informed of the car uncertainty were better
prepared to switch to manual control when required than the
control group. Further, the control group showed tendencies of
automation bias, resulting in inappropriate calibrations of trust,
which is also in line with research presented by [13], where it was
concluded that people generally have positive expectations of
unfamiliar automated decision aids.
Future work will include additional data collection regarding the
participants in the study so as to associate the results with
information regarding the drivers’ driving style and their
perceived subjective locus of control. Future work will also
explore other driver-automation forms of collaboration. As
discussed by Inagaki [22], for the human-automation
collaboration to progress, the automation might need to
implement some control actions when it determines that the
human is in a condition where he/she is unable to give directives
to the automation, resulting in automated technologies that are
able to understand the human’s psychological and physiological
conditions, intentions and actions in relation to the situation. In a
driving scenario, such cooperation could be based on the
automated car’s understanding of the current status of the driver
(alert, sleeping, texting etc.) and adapt the level of automation and
the frequency of warnings accordingly.
The transition of control is also a topic which needs further
structuring and investigation [23]. According to Flemisch et al.
[11], there are many questions that need to be investigated
regarding the proper balance between the driver and the
automated systems, especial about the authority of the assistance
and automated systems in emergency situations. How to design
such driver-automation handovers must also be further explored.
APPENDIX
A1: Listed below are the questions for measuring trust answered
after the driving exercise:
•Q1: I understand how the system works –its goals,
actions and output
•Q2: I would like to use the system if it was available in
my own car
•Q3: I think that the actions of the system will have a
positive effect on my own driving
•Q4: I put my faith in the system
•Q5: I think that the system provides safety during
driving
•Q6: I think that the system is reliable
•Q7: I can trust the system
A2: Below follows the instructions given to the drivers provided
with the uncertainty information (the same instructions were given
to the control group, with the exception of the information
regarding the uncertainty representation):
•You will first practice to drive the car manually in the
simulator for about 5 minutes. Thereafter, the test
session will begin, which runs for about 10 minutes and
is performed fully autonomously, that is, the automation
maneuvers the car as good as it can, based on, amongst
other variables, the prevailing sight conditions.
•In the instrument cluster for autonomous driving the
following is displayed: the speed, the engine speed, the
current gear used, the outside temperature, the current
time, the fuel level, the position of the steering wheel
during the autonomous drive, and how confident the car
is about its ability to drive autonomously. The red arrow
indicates the limit for when the automation no longer
can maneuver the car.
•To start the test session, put the gear in driving mode.
The car starts to drive autonomously when you press the
gas pedal, thereafter you can let go of the pedal. You
can take control over the car again if you so please at
any time by steering/braking/giving gas at any time.
ACKNOWLEDGMENTS
This research has been supported by the Swedish Knowledge
Foundation under grant 2010/0320 (UMIF), Vinnova through the
National Aviation Engineering Research Program (NFFP5-2009-
01315) and the University of Skövde. We would like to thank
Reetta Hallila, Emil Kullander and Sicheng Chen (Volvo Car
Corporation) for making the study possible and enjoyable! We
also direct our thanks to the participants in the study.
REFERENCES
[1] R. Parasuraman and V. Riley, "Humans and automation: Use,
misuse, disuse, abuse," Human Factors, vol. 39, pp. 230–253,
1997.
[2] J. Lee and K. See, "Trust in automation: designing for
appropriate reliance," Human Factors: The Journal of the Human
Factors and Ergonomics Society, vol. 46, pp. 50–80, 2004.
[3] N. B. Sarter, D. D. Woods, and C. E. Billings, "Automation
surprises," Handbook of human factors and ergonomics, vol. 2,
pp. 1926–1943, 1997.
217
Proceedings of the 5th International Conference on Automotive User Interfaces and
Interactive Vehicular Applications (AutomotiveUI ‚13), October 28–30, 2013, Eindhoven, The Netherlands.
[4] C. Billings, Aviation Automation: The Search for a Human-
Centered Approach. Mahwah, New Jersey: Lawrence Erlbaum
Associates, 1997.
[5] M. R. Endsley, "Automation and situation awareness,"
Automation and Human Performance: Theory and Applications,
Parasuraman, R. and Mouloua, M., Eds. Mahwah, NJ: Erlbaum. ,
pp. 163–181, 1996.
[6] R. Parasuraman and D. H. Manzey, "Complacency and Bias in
Human Use of Automation: An Attentional Integration," Human
Factors: The Journal of the Human Factors and Ergonomics
Society, vol. 52, pp. 381–410, 2010.
[7] J. McGuirl and N. Sarter, "Supporting Trust Calibration and
the Effective Use of Decision Aids by Presenting Dynamic
System Confidence Information," The Journal of the Human
Factors and Ergonomics Society, vol. 48, pp. 656–665, 2006.
[8] B. D. Seppelt and J. D. Lee, "Making adaptive cruise control
(ACC) limits visible," International Journal of Human-Comuter
Studies vol. 65, pp. 192–205 2007.
[9] B. Rajaonah, F. Anceaux, and F. Vienne, "Trust and the use of
adaptive cruise control: a study of a cut-in situation," Cognition,
Technology & Work, vol. 8, pp. 146–155, 2006.
[10] M. Beggiato and J. F. Krems, "The evolution of mental
model, trust and acceptance of adaptive cruise control in relation
to initial information," Transportation Research Part F: Traffic
Psychology and Behaviour, vol. 18, pp. 47–57, 2013.
[11] F. Flemisch, J. Kelsch, C. Löper, A. Schieben, and J.
Schindler, "Automation spectrum, inner/outer compatibility and
other potentially useful human factors concepts for assistance and
automation," in Human Factors for Assistance and Automation,
D. d. Waard, F. O. Flemisch, B. Lorenz, H. Oberheid, and K. A.
Brookhuis, Eds., ed Maastricht, the Netherlands: Shaker
Publishing, pp. 1–16, 2008.
[12] N. A. Stanton, M. Young, and B. McCaulder, "Drive-by-
wire: The case of driver workload and reclaiming control with
adaptive cruise control," Safety Science, vol. 27, pp. 149–159,
1997.
[13] M. Dzindolet, S. Peterson, R. Pomranky, L. Pierce, and H.
Beck, "The role of trust in automation reliance," International
Journal of Human-Computer Studies, vol. 58, pp. 697–718, 2003.
[14] J. Beller, M. Heesen, and M. Vollrath, "Improving the
Driver–Automation Interaction: An Approach Using Automation
Uncertainty," Human Factors: The Journal of the Human Factors
and Ergonomics Society, 2013.
[15] F. M. F. Verberne, J. Ham, and C. J. H. Midden, "Trust in
smart systems: Sharing driving goals and giving information to
increase trustworthiness and acceptability of smart systems in
cars," Human Factors: The Journal of the Human Factors and
Ergonomics Society, vol. 54, pp. 799–810, 2012.
[16] R. Finger and A. M. Bisantz, "Utilizing graphical formats to
convey uncertainty in a decision making task," Theoretical Issues
in Ergonomics Science, vol. 3, pp. 1–25 1997.
[17] L. Wang, G. A. Jamieson, and J. G. Hollands, "Trust and
Reliance on an Automated Combat Identification System,"
Human Factors: The Journal of the Human Factors and
Ergonomics Society, vol. 51, pp. 281–291, 2009
[18] J. Y. Jian, A. M. Bisantz, and C. G. Drury, "Foundations for
an empirically determined scale of trust in automated systems,"
International Journal of Cognitive Ergonomics, vol. 4, pp. 53–71,
2000.
[19] S. G. Klauer, T. A. Dingus, V. L. Neale, J. Sudweeks, and D.
Ramsey, "The impact of driver inattention on near-crash/crash
risk: An analysis using the 100-car naturalistic driving study
data," U.S. Department of Transportation Technical Report, 2006.
[20] J. Lee and N. Moray, "Trust, control strategies and allocation
of function in human-machine systems.," Ergonomics, vol. 35, pp.
1243–1270, 1992.
[21] J. Lee and N. Moray, "Trust, self-confidence, and operators'
adaptation to automation," International Journal of Human-
Computer Studies, vol. 40, pp. 153–184, 1994.
[22] T. Inagaki, "Smart collaboration between humans and
machines based on mutual understanding," Annual Reviews in
Control, vol. 32, pp. 253–261, 2008.
[23] A. Schieben, T. Gerald, F. Köster, and F. Flemisch, "How to
interact with a highly automated vechible. Generic interaction
design schemes and test results of a usability assessment " in
Human Centred Automation, D. de Waard, N. Gérard, L.
Omnasch, R. Wiczoreck, and D. H. Manzey, Eds., ed: Shaker
Publishing, pp. 251–267, 2011.