Content uploaded by Keke Long
Author content
All content in this area was uploaded by Keke Long on Jan 29, 2023
Content may be subject to copyright.
Impacts of Riding Comfort on the Attitudes of Riders, Drivers, and Pedestrians Toward
Autonomous Shuttles
Keke Long1, Xiaowei Shi*2, Zhiwei Chen3, Yuan Wang4, Xiaopeng Li*1
1 Department of Civil and Environmental Engineering, University of Wisconsin-Madison
2 Department of Civil and Environmental Engineering, University of Michigan
3 Department of Civil, Architectural, & Environmental Engineering, Drexel University
4 Department of Civil and Environmental Engineering, University of South Florida
ABSTRACT
Automated vehicle (AV) shuttles are emerging mobility technologies that have been widely piloted and deployed.
Public attitude is critical to the deployment progress and the overall social benefits of automated vehicle (AV)
technologies. The AV shuttle demonstration was regarded as a good way for possible attitude improvements.
However, not all existing AV shuttle technologies are mature and reliable enough. Some frustrating uncomfort
issues in AV shuttle demonstrations may adversely affect public attitudes of AV technologies. Studying the impact
of the comfort of the demonstration on attitudinal change can help us provide guidance for future demonstrations.
Thanks to the AV shuttle project piloted at Dunedin, Florida, this paper interviewed 161 people before and after
taking an open road AV shuttle ride. In addition to the participant's demographic information, driving-related
information and psychological-related information, we collected the AV shuttles' operation status (e.g., abrupt
brake). A series of discrete outcome models were employed to estimate the factors influencing people's initial
opinions before the AV shuttle ride and people's attitudes change after the ride. The most important finding is that
an unsatisfactory riding experience could negatively affect participants' attitudes toward AV shuttles. Particularly,
the number of abrupt brakes raises more concerns when the participants are drivers who share the rights of way
with the AV shuttle compared to when the participants are pedestrians or AV shuttle riders. Moreover, the
participants' attitudes toward sharing public space with AV shuttles as pedestrians and human drivers exhibited
more negative changes compared with simply riding on AV shuttles. These results improve our understanding of
public attitudes toward the existing AV technologies and help identify new opportunities to improve AV technology
adoption.
Keywords: Automated vehicle, technology acceptance, field experiment, discrete statistical model, random
parameters.
2
1. INTRODUCTION
Automated vehicle (AV) shuttles have been made accessible to general public via a great number of pilots
and demonstrations across the world in recent years. AV shuttles hold promise to sustainably improve transportation
safety, mobility, and accessibility in the near future (NHTSA, 2016). Realizing these potential benefits of AV
shuttles hinges on people's attitudes toward this technology. Positive attitudes will help facilitate the adoption of
the new technology and realize its full potential benefits for society. Therefore, it is critical to understand people's
attitudes and their relationships with contributing factors for identifying means to nurture healthy and positive
attitudes toward the AV shuttle technology.
Recent studies found that people's attitudes toward AV technologies are influenced by their demographic
characteristics (Ahmed et al., 2020; Krueger et al., 2016; Salonen, 2018; Sheela & Mannering, 2020; Shi et al.,
2020), driving habits (Ahmed et al., 2020), and psychological characteristics (Eden et al., 2017; Liu et al., 2019;
Liu & Xu, 2020; Shi et al., 2020; Xu et al., 2018). Besides these intrinsic factors that are not easy to modify, there
is an increasing interest in identifying exogenous factors that are modifiable for possible attitude improvements.
One such low-hanging fruit factor is the AV ride experience, which has been shown to positively influence
participants' attitudes toward AVs (Kolodge et al., 2020; Pyrialakou et al., 2020; Xing et al., 2022). With these
instructive findings, AV shuttle manufacturers and operators invest intensively in AV demonstrations (Dan, 2020)
to shape people's attitudes and expectations to promote technology adoption via demonstrations.
Nevertheless, outcomes of real-world demonstrations were not always consistent with this optimistic
expectation. In contrast, it is often shown that a non-negligible portion of participants may negatively change their
attitudes after riding AV shuttles (Liu et al., 2019; Liu & Xu, 2020; Shi et al., 2020). Most studies attributed such
negative attitude change to factors other than riding comfort since they simply assume all rides are homogeneous
and their comfort levels are identical. However, with existing immature and unreliable AV technologies, different
rides may result in heterogeneous feelings of comfort from the participants. It was reported that unexpected AV
shuttle operations, such as abrupt brakes, may occasionally happen and cause riding discomfort during real-world
demonstrations (Templeton, 2020). Such rides with high discomfort, as opposed to the intention of using
demonstrations to encourage technology adoption, may in fact adversely affect public attitudes and discourage AV
shuttle ridership. Evidently, different rides may experience different levels of comfort, which may result in
heterogeneous attitude changes. For example, pleasure rides with few abrupt brakes may tend to improve
participants' attitudes toward the technology after riding, while uncomfortable rides with frequent abrupt brakes
may degrade their attitudes. However, as most existing studies did not recognize the heterogeneity of riding comfort
across different rides, the effect of heterogeneous riding comfort on public attitude change is yet to be quantified.
Without quantifying this relationship, people may mistakenly associate attitude degradation with other less relevant
factors and fail to devise appropriate measures for improving public attitude and accelerating technology adoption.
Table 1 Studies involving real-world ride experience to analyze changes in public attitudes toward AV and AV shuttles.
Study
Test ride
mode
Number of
participants
Influencing factors
Demographic
characteristics
Driving related
information
Psychological
characteristics
Comfort of the
AV ride
Xu et al., 2018
AV on
enclosed
test sites
300
√
Liu et al., 2019
300
√
Liu & Xu, 2020
300
√
Shi et al., 2020
166
√
√
Eden et al., 2017
AV shuttle
on public
roads
17
Moták et al., 2017
532
√
√
Madigan et al., 2017
315
√
√
Salonen, 2018
197
√
Nordhoff et al., 2018
384
√
Hilgarter & Granig,
2020
19
√
√
This study
161
√
√
√
√
3
Furthermore, most existing studies focus on the attitudes of riders. But AV shuttles are expected to share
the rights of way with other road users (human drivers and non-motorists such as pedestrians) in a mixed traffic
environment, where the attitudes of these road users of AV shuttles also play a critical role in AV deployment. A
few studies investigated pedestrians' attitudes after providing them with real-world interaction experiences with AV
shuttles in field experiments and found that pedestrians' attitudes toward AV shuttles were influenced by
demographic features (Deb et al., 2017; Hulse et al., 2018; Pyrialakou et al., 2020) and favorable interpretations of
the AV brand (Reig et al., 2018). However, no definitive conclusions have been drawn regarding whether
interactions with AV shuttles lead to the increased trust of AV shuttles from pedestrians. One exemplary program
is the three-year BikePGH program in Swiss. While several studies showed that interaction experience improved
pedestrians' attitudes towards AV shuttles in the BikePGH program (Das et al., 2020; Penmetsa et al., 2019; Rahman
& Dey, 2022), other studies did not report substantial improvements (Xing et al., 2022). More studies are thus
needed to reconcile these contradictory results. Regarding human drivers' attitudes of AVs, studies based on real-
world interaction experiences were not found by the research team.
The above discussions identify two critical research gaps in the literature, which are important prerequisites
for deriving a further understanding of and developing effective strategies to promote public acceptance of AV
shuttles. First, existing studies have not fully considered the influence of the comfort of AV ride experience on
public attitudes toward AV shuttles. Second, to our best knowledge, no study provides a holistic investigation of
the public attitudes toward AV shuttles, including AV riders, human drivers, and other road users such as pedestrians
who share the rights of way with AV shuttles.
To fill these gaps, this study used the AV shuttles operating in Dunedin, FL, as a field experiment to
investigate public attitudes toward AV shuttles, including riders, pedestrians, and human drivers (we call them
drivers in the remainder of this paper for the convenience of the illustration). A survey was carried out to measure
the participants' attitudes of the AV shuttle before and after riding on the AV shuttle. Different from existing studies,
the survey collected information from the participants from three perspectives, asking them to answer the same set
of questions as riders, pedestrians, and human drivers before and after a ride. Apart from the information that was
commonly collected in existing studies (such as individual characteristics, and attitudes on riding comfort before
and after rides), the survey collected data related to the comfort of each participant's ride. The data allow the
investigation of exogenous factors influencing the participants' initial attitude and attitude change towards AV
shuttles with statistical models. The initial attitude was modeled to offer a baseline for understanding the
participants' attitudes before the field experiment. The attitude change was to unveil both intrinsic and exogenous
influencing factors that change public attitudes toward AV shuttles. To this end, a set of likelihood ratio tests was
conducted, revealing statistically different differences in participants' initial attitude and attitude change toward AV
between the three perspectives. That is, participants showed different initial attitudes and attitude changes toward
AV shuttles when answering the questions in the survey as riders, pedestrians, and drivers. Thus, models were
estimated separately from the perspectives of riders, pedestrians, and drivers. Specifically, ordered probit models
with random parameters are adopted to model the initial attitude toward AV shuttles, and multinomial logit models
are employed to explain the attitude change.
With these, this study makes two contributions to the literature. First, it studies how the comfort of the ride
experience affects public attitudes toward AV shuttles. Results show that an unsatisfactory riding experience could
negatively affect public attitudes toward AV shuttles. The unsatisfactory experiences were mostly caused by
existing immature AV technologies (abrupt brakes, driving slowly, being too conservative at intersections). These
results, thus, offer evidence to our argument that existing AV technologies are too immature to realize the expected
benefits of AV shuttle demonstrations. AV pilot program planners need to be aware of the necessity of test ride
comfort to minimize the possible negative outcomes (such as causing distrust) of their programs. Second, this study
investigates people's willingness to share the rights of way with AV shuttles as pedestrians and drivers. The analysis
revealed that the participants' attitudes toward sharing the rights of way with AV shuttles as pedestrians and human
drivers exhibited more negative changes than simply riding on AV shuttles. This result indicates that obtaining trust
from pedestrians and human drivers to share public roads with AV shuttles is probably a more critical challenge in
the wide deployment of AV shuttles in the real world. AV operators might consider pedestrians' and drivers'
concerns when promoting AV shuttles by reducing distrust and uncertainty when pedestrians and drivers interact
with AV shuttles. Overall, these results depict a complete picture of the public acceptance of existing AV technology.
4
They offer important insights that AV pilot program planners and managers can draw on when planning future AV
pilot programs. They are also useful for practitioners to design future AV deployment strategies in the real world.
The rest of this paper is organized as follows. Section 2 and Section 3 introduce the data collection and
modeling approaches. Section 4 discusses the model estimation results for initial public attitudes and attitude change
toward AV shuttle comfort. Section 5 concludes the work and points out future research directions.
2. DATA COLLECTION
The primary goal of this study is to unveil intrinsic and exogenous factors influencing the public attitude
toward AV shuttles and attitude change. To this end, we first conducted a survey collecting basic information and
people's attitude toward the AV shuttle before and after the ride. This section briefly introduces the data collection
process.
The attitude data used in this study are from the survey data collected in Dunedin from May to July 2022.
Figure 1 shows the AV shuttle operated by Pinellas Suncoast Transit Authority (PSTA). The AV shuttles are electric
vehicles with a capacity of 10 passengers (8 seats and 2 extended seats). There are no steering wheel or pedals in
AV shuttles. The AV shuttles will operate following a pre-determined route and react to the surrounding traffic
environment in real-time. A safe guard is always on board to turn the vehicle on and off and monitor the operation
of the AV shuttle. The maximum speed of the AV shuttles is 20 km/h. Figure 2 shows the route of the AV shuttles.
The circle route is around 1,600 meters, with eight stations along a two-lane road.
(a)
(b)
Figure 1 AV shuttle operated by PSTA (a. Exterior look. b. Interior view while driving.)
Figure 2 Route of the AV shuttle (Source: Google maps)
5
A questionnaire was designed to collect public attitudes of AV shuttles (see Appendix for details). The
questionnaire has three parts: 1) basic information about the participants (including demographics of the participants
(Q1-Q4), daily commute information (Q5, Q6), and previous experience with emerging vehicle technologies (Q7,
Q8)), 2) attitudes toward the safety of the AV technology before taking the AV ride (Q9-Q11), 3) attitudes toward
the comfort of the AV technology after taking the AV ride (Q12-Q14). In the second and third parts, participants
were provided a 5-point scale ranging from 'very uncomfortable' to 'very comfortable' to show their feeling of
comfort. There were three perspectives of attitude. The first perspective asked participants whether they felt
comfortable as AV shuttle riders. The second perspective asked participants whether they felt comfortable as
pedestrians walking on AV shuttle-passed roads. The third perspective asked participants whether they felt
comfortable as drivers sharing the right of way with AV shuttles.
All participants will be informed that the vehicle was a self-driving AV shuttle before answering the
questionnaires. The first and second parts are filled before participants get on the AV shuttle or during the first few
minutes of the ride. Participants filled out the third part of the questionnaire after riding in the AV shuttle for at least
10 minutes. Both paper-based and web-based questionnaires with the same questions were used.
Additionally, during the survey, information related to the ride experience (including weather, test time,
test date, and the number of emergency braking) was manually recorded by a well-trained graduate research assistant
who stayed in the AV shuttle.
As mentioned in the introduction, the questionnaire answering data has the potential to capture the
influencing factors of participants' attitudes before the field experiment. The information related to the ride
experience has the potential to unveil previously overlooked influencing factors that change public attitudes toward
AV shuttles.
3. METHODS
This section presents the statistical evaluation criteria, and methodological approaches for initial attitudes
and for attitude change to investigate factors influencing the public attitude toward AV shuttles and attitude change.
3.1. Statistical evaluation criteria
Since our questionnaire involved participants' attitudes before and after the ride experience and attitudes
from three perspectives (riders, pedestrians, and drivers), we first had to confirm whether different models are
necessary for those different attitudes. This study conducted a likelihood ratio test (Washington et al., 2020) to
assess whether statistically significant differences exist between riders, pedestrians, and drivers in terms of their
initial attributes and attribute changes toward AV ride comfort. The likelihood ratio test is a statistical test of the
goodness of fit between models. Separate models of two subgroups are compared to the model of the whole group
to see if the two separate models of two subgroups fit the dataset significantly better. If so, there are significant
differences between the two subgroups, and different models are necessary. Estimation results from three models
are used: a model for all participants, a model for group A (e.g., the attitude before the test ride), and a model for
group B (e.g., the attitude after the test ride). With these model estimates, a likelihood ratio test was conducted as
follows:
(
1
)
where is the log-likelihood at the convergence of a model using the data of all participants, and
are the log-likelihood at the convergence of models using the data of group A and group B. The resulting
is distributed, with degrees of freedom equaling to the difference between the numbers of parameters in the
models. Chi-Square value is applied to determine whether the null hypothesis stating that the parameters in two
models are equal can be rejected (Chen & Li, 2021).
3.2. Methodological approach for initial attitudes
The participants' initial attitude of the comfort of the AV shuttles is studied first. Given the ordered nature
of the available responses to this question, an ordered probability modeling approach is adopted for the analysis.
Traditional ordered probability models are specified by defining an unobserved variable, , for each observation
as the linear function,
6
(
2
)
where is a vector of explanatory variables determining the discrete answers for participant , is the vector of
estimable parameters, is the disturbance term. Note that only one participant chooses "very uncomfortable". To
ensure an adequate sample size, the two negative altitude choices ("very uncomfortable" and "uncomfortable")
options are summarized. The non-numerical ordered initial attitudes, (B represents before the test ride), are
converted to integers without loss of generality (i.e., 1 = "very uncomfortable" and "uncomfortable", 2 = "neutral",
3 = "comfortable", 4 = "very uncomfortable").
(
3
)
where to are estimable thresholds that define . They are estimated jointly with the model parameters .
With this, if is assumed to be normally distributed across observations with a mean equal to 0 and variance equal
to 1, the ordered selection probability is
(
4
)
where is the standardized cumulative normal distribution.
To capture the unobserved heterogeneity of explanatory variables, a random parameters approach
(Washington et al., 2020) is integrated into the model. Thus, the parameters can vary across observations. This will
help avoid potential bias and erroneous statistical inferences. The estimable parameters are written as,
(
5
)
where is a vector of estimable parameters that potentially varies across observations , is the vector of mean
parameter estimates across all observations, and is a vector of randomly distributed terms. Estimation of the
random parameters ordered probit is undertaken by the simulated maximum likelihood approach.
3.3. Methodological approach for attitude change
Although previous research found that the initial attitudes might cause the anchoring effects that will affect
the participants' final attitudes (Sheela & Mannering, 2020), the likelihood ratio test verifies that participants'
attitudes were unstable between their initial assessment of AV safety and their final assessment. With this result, a
series of models are estimated to understand which factors determine the likelihood of participants shifting from
their initial attitudes about AV safety. The ordered probit estimation results provide insights into the factors
influencing initial attitudes on AV safety. Here we focus on attitude change from different initial attitudes.
After excluding the incomplete questionnaire answers without feedback for specific questions (e.g., people
without a driver's license could not perceive the comfort level of being a driver), there were a total of 25 kinds of
attitude shifting. However, due to the limited number of observations, the number of several attitude shifting was
too small. For example, among the attitudes from the perspective of pedestrians, only three people changed from
'uncomfortable' to 'normal', and only one changed from 'uncomfortable' to very 'uncomfortable'. Continuously using
the statistical models to estimate the influence factors for these two kinds of attitude change may lose significance.
Therefore, only the attitude change is studied in the rest of this paper. Three kinds of attitude changes are studied:
negative change, non-change, and positive change. Among three kinds of attitudes, if the post-ride attitude changes
to a more negative attitude, the post-ride option is categorized as negative; if the attitude does not change, the post-
ride option is categorized as no-change; if the post-ride attitude changes to a more positive attitude, the post-ride
option is categorized as a positive change.
We applied the multinomial logit model for attitude change. The probability of attitude change is:
(
6
)
7
(
7
)
(
8
)
where , and are the probabilities that participants have a negative change, non-change, and positive
change after the test ride. , and are corresponding indirect utility functions.
4. RESULTS AND DISCUSSION
This section discusses the results of statistical models. Section 4.1 presents general descriptive statistics of
the survey data. Section 4.2 presents results from the likelihood ratio test, which shows the necessity of separately
modeling riders, pedestrians, and drivers and pre-ride and post-ride attitudes. Sections 4.3 and 4.4 show the
evaluation results of initial attitudes and attitude change, respectively.
4.1. Descriptive statistics
Data from 161 participants were obtained. The age and gender distributions of the participants are generally
consistent with those from the local population obtained from the American Community Survey 5-Year Data
(shown in Figure 3), which shows the participants were representative of the general population. To ensure accuracy,
teenagers under ten did not participate in the questionnaire.
Figure 3 Comparison of gender and age distribution of participants and the local population
Participants have different attitudes from the three perspectives. Figure 4 illustrates people's initial attitude
of comfort from three different roles. The percentage of participants who choose 'Very uncomfortable' or
'uncomfortable' as pedestrians and drivers exceeds the percentage of participants who choose 'Very uncomfortable'
or 'uncomfortable' as AV shuttle riders. This may be because pedestrians and drivers are more vulnerable to a crash
with AV shuttles.
8
Figure 4 People's initial attitude of comfort from three different roles.
Figure 5 shows people's attitudes of comfort from three different roles after test rides. The percentage of
participants who choose 'Very uncomfortable' or 'uncomfortable' as AV shuttle riders and drivers exceeds the
percentage of participants who choose 'Very uncomfortable' or 'uncomfortable' as pedestrians. This may be because
AV shuttles show sensitive collision avoidance behavior in operation. This collision avoidance behavior enhances
the comfort of pedestrians. However, too-sensitive collision avoidance causes abrupt brakes. The sharp braking can
make the AV shuttle riders feel uncomfortable and increase the danger of rear-end collision to vehicles following
the AV shuttle.
Both the attitude before and after the test ride show people are more conservative about driving near AV
shuttles rather than walking around AV shuttles. However, previous survey-based studies had different conclusions
about the safety perception of walking near an AV compared to driving near an AV (Pyrialakou et al., 2020). One
possible reason for this discrepancy is that real AV ride experience is necessary. When the observers saw the real
AV, the attitude answers would naturally receive the effect of the slow driving speed of the AV. Future research
could also further investigate what AV shuttles' characteristics cause people to feel uncomfortable from pedestrians'
and drivers' perspectives.
The above qualitative comparisons of people's attitudes of comfort show a clear difference among the three
perspectives. In the next sections, the detailed factors of the attitude of comfort are quantitatively analyzed in
separate statistic models of the three perspectives.
Figure 5 People's attitude of comfort from three different roles after test rides.
4.2. Likelihood ratio test result
The likelihood ratio test allows us to determine whether different models are needed for attitudes before
and after the test ride and the attitudes from three perspectives. The likelihood ratio test results are summarized in
Table 2. Two separate attitudes, Attitude 1 and Attitude 2, are evaluated to see if different statistical models are
required. Comparing the riders' pre-ride attitudes with the pedestrians' pre-ride attitudes, we found a statistic of
9.740 with 4 degrees of freedom. This indicates that we are more than 95.5% confident to reject the hypothesis that
riders' pre-ride attitudes model and pedestrians' pre-ride attitude model are the same. Similarly, comparing the riders'
pre-ride attitudes with riders' post-ride attitudes gives a statistic of 23.868 with 6 degrees of freedom. This
suggests we are more than 98.7% confident to reject the hypothesis that riders' pre-ride attitudes model and
9
pedestrians' pre-ride attitude model are the same. These results provide substantial evidence that the riders' pre-ride
attitude model is not the same as the post-ride attitude model. And the pre-ride attitude is not stable from the
perspective of riders, pedestrians, and drivers. Based on these results, different models could be adopted for pre-
ride attitude and post-ride attitude, and also different models for different perspectives.
Table 2 Likelihood ratio test results between different attitudes ( values with degrees of freedom in parenthesis and
confidence level in brackets)
Attitudes 2
Attitudes 1
Riders' pre-ride attitudes
Pedestrians' pre-ride attitudes
Drivers' pre-ride attitudes
Riders' pre-ride attitudes
-
9.740 (4) [>95.5%]
24.868 (6) [>99.9%]
Pedestrians' pre-ride attitudes
9.740 (4) [>95.5%]
-
12.825 (5) [>97.5%]
Drivers' pre-ride attitudes
24.868 (6) [>99.9%]
12.825 (5) [>97.5%]
-
Riders' post-ride attitudes
23.868 (6) [>98.7%]
-
-
Pedestrians' after-ride attitudes
-
10.740 (4) [>96.5%]
-
Drivers' post-ride attitudes
-
-
12.225 (5) [>97.5%]
4.3. Model estimation results of initial attitudes
The estimation results of the ordered probit model for the initial attitude of comfort in AV shuttles and the
corresponding marginal effects are shown in Table 3. Five variables were found to significantly influence initial
attitudes toward AV shuttles.
Participants over the age of 60 are more likely to have a negative attitude of the AV shuttle. The marginal
effect shows that participants over the age of 60 are 7.1% less likely to have a "Very comfortable" attitude than
younger people. This result matches those from earlier studies (Madigan et al., 2017). However, Nordhoff et al.
(2018) pointed out that seniors have a higher intention to use AV shuttles. This disagreement may come from other
hidden factors, including the route of the AV shuttle, the schedule, etc., and deserves further study.
Participants with high income (annual personal income higher than 100k) are inclined to have a negative
attitude of the AV shuttle. Participants with high incomes have a 6.4% higher probability of having negative
attitudes than others with lower incomes. Numerous studies have shown that respondents' economics influence their
attitudes toward using AV shuttles, but their findings are completely opposite. Previous research found that people
with high incomes are more negative toward automated techniques (Shi et al., 2020) and are less concerned about
AV crashes and equipment/system failure (Ahmed et al., 2020). A cross-country survey also found that respondents
from countries with higher GDP per capita were more critical of using AV shuttles. However, other research
indicates that wealthy groups are more likely to adopt AVs (Sheela & Mannering, 2020) or AV shuttles
(Cunningham et al., 2019). Determining the impact of personal wealth on attitudes will help target users and set
ticket prices.
Participants with long commute times (commute time is longer than 30 minutes) are inclined to have a
negative attitude of the AV shuttle. Participants with long commute times have a 22.4% higher probability of having
negative attitudes than shorter commute times. This factor was not included in previous studies. A previous study
found that the higher a person's monthly fuel cost (Shi et al., 2020), the more negative the attitude toward AV.
Although there is no strict relationship between commuting time and fuel cost, these similar findings suggest that
AV may be more suitable to be accepted as a short-distance travel mode in the future.
Participants who drive alone in commuting and are highly interested in new technologies are more apparent
to have a positive attitude of the AV shuttle. There are two possible reasons for this result. First, driving alone is
boring, and easy to get distracted, while AV technology can avoid the safety risks caused by driver distraction.
Second, people who drive by themselves are very familiar with driving, thus they have confidence in AV technology.
People who do not drive alone (share ride, transit, bicycle) lack driving experience and therefore do not understand
and trust new technologies about driving.
Participants who are highly interested in new technologies have a 9.6% higher probability of having a "Very
comfortable" attitude than people less interested in new technologies.
Two of the five variables were found to produce normally distributed random parameters with statistically
significant standard deviations indicating significant unobserved heterogeneity among the participants. The long
10
commute time indicator (1 if the participant's commute time is longer than 30 minutes, 0 otherwise) has an estimated
mean parameter of -1.137 and a standard deviation of 0.988. This demonstrates that the effect of this variable
increases the likelihood of having a 'normal' attitude by roughly 55.96%. This heterogeneity may stem from different
expectations of the AV shuttle's effectiveness in fastening commuting. Part of the participants with long commute
times regards AV shuttles as a complementary traffic mode of the current transportation system that can quickly
transport people from homes to stations. Other groups of people may not see AV shuttles as comfortable as they do
not compare favorably with existing buses. Therefore, the attitudes of people with long commutes toward AV shuttle
are worth studying in depth to investigate whether AV shuttle is likely to be accepted as a new solution to long
commutes. The other random parameter is the highly interested in new technologies indicator (1 if the participant
is highly interested in new technologies, 0 otherwise), it is found with a mean of 0.564 and a standard deviation of
0.632, indicating that the effect of this variable increased the likelihood of having a 'normal' attitude for roughly
11.70%. This indicates that even among people who are willing to try new technologies, the attitude of AV is not
the same.
Table 3 Ordered probit model for people's initial attitudes toward comfort as riders.
Variable Description
Mean
Estimated
Parameter
t Statistic
Marginal Effects
Very
uncom-
fortable/
Uncom-
fortable
Normal
Comfor
-table
Very
comfor-
table
Constant (Standard deviation of
parameter distribution)
1.503
(0.169)
6.84
(1.88)
Senior indicator (1 if the participant is
over 60-year-old, 0 otherwise)
0.255
-0.503
-2.36
0.071
0.126
-0.125
-0.071
High-income indicator (1 if the
participant has an annual personal
income higher than 100k, 0 otherwise)
0.292
-0.483
-2.26
0.064
0.126
-0.122
-0.068
Long commute time indicator (1 if
participant's commute time is longer than
30 minutes, 0 otherwise) (Standard
deviation of parameter distribution)
0.161
-1.137
(0.988)
-4.27
(3.80)
0.224
0.196
-0.308
-0.112
Drive alone indicator (1 if the main
commute mode of the participant is
driving alone, 0 otherwise)
0.503
0.465
2.40
-0.059
-0.128
0.116
0.072
Highly interested in new technologies
indicator (1 if the participant is highly
interested in new technologies, 0
otherwise) (Standard deviation of
parameter distribution)
0.435
0.564
(0.632)
2.97
(4.36)
-0.061
-0.158
0.122
0.096
Threshold 1
1.446
7.53
Threshold 2
2.939
12.30
Number of observations
161
Log-likelihood at zero
-199.607
Log-likelihood at convergence
-184.200
McFadden
0.077
The estimation results of the ordered probit model for people's initial attitude of comfort as pedestrians are
shown in Table 4. Similar to the modeling result of people's initial attitude of comfort in AV shuttle, short commute
time and being highly interested in new technologies are still significant characteristics that drive people to have
initial positive attitudes as pedestrians.
The results also suggest that the participants with long commute times, automated riding experience, and
high interest in new technologies had considerable unobserved heterogeneity in their initial attitudes toward AV
11
shuttles. Two reasons might explain why automated riding experience has considerable unobserved heterogeneity
in initial attitudes toward AV shuttles, especially in our sample. The first reason is the various expectations among
participants. If their previous automated riding experience met their expectations, their attitude of comfort towards
AV shuttles would be positive. If the previous experience did not meet expectations, it is vice versa. Further
investigation of how people expected the AV shuttle would be needed to verify this conjecture. Another possibility
is that participants had previously experienced automated driving at different technical levels. We did not set a strict
definition of automated riding experience, which means those participants who had only experienced vehicles with
low automation levels that are currently commercially available (e.g., automatic cruise, lane keeping function) are
regarded as having automated riding experience. Thus, the AV shuttle, as an L3 level of automation technology, is
regarded more advanced than the vehicles with low automation levels and more comfortable. Other participants
may have experienced or know about better autonomous driving (e.g., commercial self-driving vehicles such as
Tesla that can plan their paths). Therefore, the AV shuttle, which follows a fixed track, seems less comfortable.
This result indicates that the comfort of the AV shuttle riding experience has an important impact on the public and
deserves in-depth analysis. If those currently available AV technologies cannot meet the user expectations, this may
leave people with a negative initial impression of other AV technologies in the follow-up process.
Table 4 Ordered probit model for people's initial attitudes toward comfort as pedestrians.
Variable Description
Mean
Estimated
Parameter
t Statistic
Marginal Effects
Very
uncom-
fortable/
Uncom-
fortable
Normal
Comfor-
table
Very
comfor-
table
Constant (Standard deviation of
parameter distribution)
1.018
(0.469)
6.02
(4.85)
Long commute time indicator (1 if
participant's commute time is longer than
30 minutes, 0 otherwise) (Standard
deviation of parameter distribution)
0.168
-0.924
(0.527)
-3.42
(2.10)
0.203
0.123
-0.234
-0.092
Automated riding experience indicator (1
if the participant has automated riding
experience before test ride, 0 otherwise)
(Standard deviation of parameter
distribution)
0.768
0.536
(0.411)
2.32
(1.95)
-0.142
0.111
0.103
-0.072
Highly interested in new technologies
indicator (1 if the participant is highly
interested in new technologies, 0
otherwise) (Standard deviation of
parameter distribution)
0.445
0.821
(0.908)
4.28
(5.63)
-0.127
-0.192
0.182
0.137
Threshold 1
1.382
8.48
Threshold 2
2.731
12.53
Number of observations
155
Log-likelihood at zero
-205.542
Log-likelihood at convergence
-191.137
McFadden
0.070
The estimation results of the ordered probit model for people's initial attitude of comfort as drivers are
shown in Table 5. People with short commute times and who are highly interested in new technologies always have
a more positive attitude toward AV shuttles than others. Note that the commute time indicator is statistically
significant in all three initial models, indicating that people with long commute times are less likely to embrace
automated driving technology in the future. This may be because of the low operating speed of AV shuttles.
Females and participants over the age of 60 have a more negative attitude of comfort as drivers. Previous
studies also reported more conservative attitudes among females and seniors (Hulse et al., 2018). These attitudes
12
might be affected by the respondent's confidence in their own driving ability, as recent studies suggest that females
are less likely than males to be confident drivers (Bergdahl, 2005).
The statistically significant random parameters were only found in the senior indicator. This implies that
attitudes toward AV shuttles vary greatly among seniors. Some seniors who can drive themselves may view AV
shuttles as less comfortable and convenient than private cars, while some seniors who cannot drive themselves may
welcome AV shuttles as their travel opportunities increase.
Table 5 Ordered probit model for people's initial attitudes toward comfort as drivers.
Variable Description
Mean
Estimated
Parameter
t Statistic
Marginal Effects
Very
uncom-
fortable/
Uncom-
fortable
Normal
Comfor-
table
Very
comfor-
table
Constant
1.061
4.65
Senior indicator (1 if the participant is
over 60-year-old, 0 otherwise) (Standard
deviation of parameter distribution)
0.273
-0.523
(0.778)
-2.36
(3.59)
0.147
0.032
-0.080
-0.099
Female indicator (1 if the participant is
female, 0 otherwise)
0.561
-0.344
-1.73
0.094
0.036
-0.051
-0.080
Long commute time indicator (1 if
participant's commute time is longer than
30 minutes, 0 otherwise)
0.165
-0.818
-2.94
0.254
0.015
-0.134
-0.134
Highly interested in new technologies
indicator (1 if the participant is highly
interested in new technologies, 0
otherwise)
0.453
0.567
2.92
-0.158
-0.060
0.083
0.135
Threshold 1
1.040
7.59
Threshold 2
1.941
10.12
Number of observations
139
Log-likelihood at zero
-188.994
Log-likelihood at convergence
-176.838
McFadden
0.064
Comparing people's initial attitudes from three perspectives, the results suggested that the influencing
factors differed across three perspectives. Some factors were found to be only significant in particular perspectives.
Females showed a significantly negative attitude towards AV shuttles as drivers. However, in the attitudes
of riders and pedestrians, gender did not significantly affect attitudes. This difference may be caused by women's
lack of confidence in their driving skills rather than their attitude toward AV shuttles. The comfort attitudes might
be affected by the respondent's confidence in their driving ability. Females are less likely than males to be confident
drivers reported by a previous study (Bergdahl, 2005).
The automated riding experience only made participants feel more comfortable from the passenger's
perspective but did not affect rider or driver attitudes, which suggests that test rides on AV shuttles can largely
dispel pedestrian anxiety. This may be because people are the most vulnerable road users. Therefore, people feel
the least comfort from the pedestrian's perspective when they have no experience with AV shuttles. This suggests
the importance of focusing on pedestrian-oriented publicity to dispel pedestrians' worries.
Two factors only affected riding the AV shuttle, but not the perspective of sharing the road with the AV
shuttle. This may be determined by the traffic mode of the shuttle. People with high incomes do not want to ride in
AV shuttles. This may be because high-income people, who can afford private cars, think that shuttles are not as
comfortable as driving and thus do not want to take shuttles, but there is no negative attitude towards AV technology.
Therefore, people with high incomes are not repulsed by AV technology. They are just not potential users of shuttles.
Future research could explore their acceptance of other forms of AV, such as AV taxis and private AV shuttles. The
13
other factor, driving alone, only positively impacts the attitude of AV shuttle riders. This is possible because people
who often drive alone may regard driving as a burden and prefer AV shuttles as an additional way to travel.
In summary, the difference in people's attitudes toward AV shuttles from three perspectives may come from
personal characteristics and transportation modes. On the one hand, respondents' judgment of their driving ability
affects their attitudes as drivers. Respondents who have experienced the ability of AV technology to avoid collisions
with people have a more positive attitude toward AV shuttles. Another aspect is the respondents' acceptance of AV
shuttles as a new form of transportation. When respondents do not like shuttles as a traffic mode (participants with
high income or who do not have to drive alone may not enjoy public transportation), they will be reluctant to act as
riders but may be willing to try other autonomous technology.
4.4. Model estimation results of attitude change
The multinomial logit model is used to estimate participants' attitude change toward the comfort of the AV
shuttle after a test ride, and the model estimation results are shown in Table 6. Seven variables are found to affect
participants' attitude change on AV safety significantly. Therefore, the estimated utility functions are:
(
9
)
The lack of constant in the no-change function establishes it as a 0 baseline. The negative constant in
and shows people are more willing to keep their minds.
Although People with high incomes or long commute times have negative attitudes towards AV shuttles as
riders, they are more likely to change their minds positively than others. The elasticity analysis shows that
participants with an annual personal income higher than 100k are 54.95% more likely to change their minds
positively. People with commute times longer than 30 minutes are 10.02% more likely to change their minds
positively. This result shows that the AV shuttle operation companies might involve more people trying its service
so that people who were originally optimistic about AV shuttle may change their attitude after the ride experience.
The number of abrupt brakes is an important factor that leads to negative change. A 1% increase in the
number of abrupt brakes results in a 1.96% decrease in the probability of not changing the attitude. The abrupt
brakes recorded in this study were sharp enough to be perceived by passengers. During the operation, the abrupt
brakes are caused by several factors: 1) surrounding obstacles (e.g., overtaking vehicles, pedestrians on motor roads,
grass or tree branches extending into the roads); 2) interrupted communication. Vehicles in these scenarios tend to
adopt the behavior of slowing down sharply to zero and then starting to drive again. Even though all passengers
were required to wear seat belts and were informed of the risk of sharp braking in advance, the abrupt brakes of the
AV shuttle were unexpected and much more than common human-driven buses. This clearly had a significant
influence on the rider's attitude of comfort.
Table 6 Multinomial Logit Estimation Results for Change of Attitude toward the comfort of the AV shuttle as riders
(parameters defined for [NC] Negative change; [NO] Non-change function; [PC] Positive change).
Variable Description
Estimated
Parameter
t statistic
Negative change constant [NC]
-2.521
-4.82
Positive change constant [PC]
-0.912
-3.26
Variables that do not vary across alternate outcomes
High-income indicator (1 if the participant has annual personal income higher than
100k, 0 otherwise) [NC]
-1.907
-3.14
Number of abrupt brakes [NC]
1.985
5.60
High-income indicator (1 if the participant has annual personal income higher than
100k, 0 otherwise) [NO]
-0.987
-2.34
Long commute time indicator (1 if participant's commute time is longer than 30
minutes, 0 otherwise) [PC]
1.614
3.29
Number of observations
161
Log-likelihood at convergence [LL(β)]
-138.372
14
The multinomial logit model results of participants' attitude change toward comfort as pedestrians are
shown in Table 7.
In the model of participants' attitude change toward comfort as pedestrians, people also prefer to keep their
minds. Similar to the model of participants' attitude change toward the comfort of the AV shuttle, people with high
incomes and long commute times are more seemingly to change their minds positively, and the number of abrupt
brakes leads to negative change. Besides, people with automated riding experience are less likely to negatively
change their attitude of comfort as pedestrians. This might be because they had preconceived notions about how the
AV shuttle would function, and the test ride experience fulfilled their expectations.
Table 7 Multinomial Logit Estimation Results for Change of Attitude toward the comfort as Pedestrians (parameters defined
for [NC] Negative change; [NO] Non-change function; [PC] Positive change).
Variable Description
Estimated
Parameter
t statistic
Negative change constant [NC]
-2.331
-4.36
Positive change constant [PC]
-0.931
-2.07
Variables that do not vary across alternate outcomes
High-income indicator (1 if the participant has annual personal income higher than
100k, 0 otherwise) [NC]
-1.388
-2.07
Number of abrupt brakes [NC]
0.771
2.41
Automated riding experience indicator (1 if the participant has automated riding
experience before the test ride, 0 otherwise) [NC]
2.100
2.91
High-income indicator (1 if the participant has annual personal income higher than
100k, 0 otherwise) [NO]
-1.002
-2.41
Automated riding experience indicator (1 if the participant has automated riding
experience before the test ride, 0 otherwise) [NO]
1.451
2.41
Long commute time indicator (1 if participant's commute time is longer than 30
minutes, 0 otherwise) [PC]
1.356
2.76
Number of observations
152
Log-likelihood at convergence [LL(β)]
-131.11
Estimation results for change of attitude toward the comfort as drivers are presented in Table 8. Participants
also like maintaining their thoughts from the driver's perspective. Similar to previous results, people with long
commute times are more likely to change their minds positively, and the number of abrupt brakes leads to negative
change. Moreover, females are less likely to change their minds negatively.
Table 8 Multinomial Logit Estimation Results for Change of Attitude toward the comfort as drivers (parameters defined for
[NC] Negative change; [NO] Non-change function; [PC] Positive change).
Variable Description
Estimated
Parameter
t statistic
Negative change constant [NC]
-0.642
-1.35
Positive change constant [PC]
-0.112
-0.41
Variables that do not vary across alternate outcomes
Number of abrupt brakes [NC]
0.635
2.17
Female indicator (1 if the participant is female, 0 otherwise) [NC]
-0.831
-1.82
Highly interested in new technologies indicator (1 if the participant is highly interested
in new technologies, 0 otherwise) [NO]
0.708
1.93
Long commute time indicator (1 if participant's commute time is longer than 30
minutes, 0 otherwise) [PC]
1.385
2.74
Number of observations
135
Log-likelihood at convergence [LL(β)]
-133.54
Table 9 presents elasticity comparisons across the three perspectives. Among the six variables in the table
that were significant across all age groups, there are some noteworthy differences. An instance is the elasticity of
15
the number of abrupt brakes. The number of abrupt brakes has the greatest impact on the driver's attitude. The
number of abrupt brakes raised the likelihood of negative change in attitude by 27.14% from the perspective of AV
shuttle riders and 25.35% from the perspective of pedestrians, but 86.87% from the perspective of drivers. This
difference may be explained by the operation process. There is ample empirical evidence during the survey showing
AV shuttle might impede other vehicles. Most extremely, one rear-ends accident occurred during the survey right
after abrupt brakes of the AV shuttle. Although no injury or death was caused, all four participants who experienced
the accident negatively changed their attitudes in both rider perception and driver perception. Because of the small
number of survey participants who experienced the accident, the accident is not included as an influencing factor
in the models. In other situations, even if the abrupt brakes did not cause an accident, they showed negative effects
besides uncomfortable. For instance, being overtaken by other vehicles is one of the major causes of abrupt brake.
The AV shuttle is unable to accurately judge and predict the safety of this situation, so it will stop sharply and wait
until the overtaking vehicle is far away before resuming driving. This kind of abrupt brake let the riders aware of
the dissatisfaction of the following vehicles because of the low cruising speed. This shows that people are more
worried about the disruption of traffic order and the instability of the vehicle embodied in the abrupt brakes than
the discomfort caused by the abrupt brakes to passengers. In addition, high-income people are more likely to change
their attitudes, especially as passengers. This may be because high-income people do not rely on shuttles for travel,
and once the AV shuttles are not comfortable enough, they will choose other traffic modes.
Table 9 Comparison of elasticities for variables in the models of three different perspectives (elasticities are specific to [NC]
Negative change; [NO] Non-change function; [PC] Positive change).
From the Perspective of AV
Shuttle riders
Elasticity
From the Perspective of
Pedestrians
Elasticity
From the Perspective of
Drivers
Elasticity
Variables significant in all three models
Number of abrupt brakes
[NC]
27.14
Number of abrupt brakes
[NC]
25.35
Number of abrupt
brakes [NC]
86.87
Long commute time
indicator [PC]
10.02
Long commute time
indicator [PC]
9.21
Long commute time
indicator [PC]
9.05
Variables significant in two models
High-income indicator
[NC]
-47.39
High-income indicator
[NC]
-46.17
High-income indicator
[NO]
-183.9
High-income indicator
[NO]
-17.94
Variables significant in one model
Automated riding
experience indicator [NC]
11.41
Automated riding
experience indicator [NO]
3.40
Female indicator [NC]
20.98
Highly interested in
new technologies
indicator [NO]
3.29
5. CONCLUSIONS AND FUTURE RESEARCH
Unlike private AVs that focus on tapping potential customers, AV shuttle is more concerned with the public
trust from various groups in society. Understanding the factors that influence public attitudes toward AV shuttles is
critical to increasing confidence in the technology and aiding its future deployment. This research collected
information from 161 participants who experienced AV shuttle services, including personal characteristics, attitudes
toward AV shuttles before a test ride, and attitudes toward AV shuttles after a test ride. Statistical models were used
to analyze the factors influencing participants' initial attitudes and factors influencing attitude change. All significant
variables found at least in one of the initial attitudes and attitude change models are summarized in Table 10. Our
main findings are as follows:
16
1. Abrupt brakes of AV shuttle during the demonstration ride deteriorate the attitudes about the comfort
of AV shuttles, especially from the driver's perspective.
2. Participants feel less comfortable sharing the rights of way with AV shuttles as drivers than taking the
AV shuttle or sharing the rights of way with AV shuttles as pedestrians, for both the initial attitude and
the change of attitude after taking the test ride of AV shuttles.
3. Participants with long commute times or high income tend to be conservative on AV shuttle comfort,
but their attitudes significantly changed toward the positive side after a successful ride.
4. The seniors and females are more conservative about sharing the rights of way with AV shuttles, but
successful rides also allow them to change their minds positively.
Table 10 Summary of initial attitude and attitude change models' findings.
Variable Description
Initial Attitude of Comfort
Positive Change of the Comfort Attitude
From the
perspective
of AV
shuttle riders
From the
perspective
of
pedestrians
From the
perspective
of drivers
From the
perspective
of AV
shuttle riders
From the
perspective
of
pedestrians
From the
perspective
of drivers
Senior indicator (1 if the
participant is over age 50, 0
otherwise)
-
n
-
n
n
n
Female indicator (1 if the
participant is female, 0
otherwise)
n
n
-
n
n
+
High-income indicator (1 if
the participant has annual
personal income higher than
200k, 0 otherwise)
-
n
n
+
+
n
Long commute time indicator
(1 if participant's commute
time is longer than 30 minutes,
0 otherwise)
-
-
-
+
+
+
Drive alone indicator (1 if the
main commute mode of the
participant is driving alone, 0
otherwise)
+
n
n
n
n
n
Automated riding experience
indicator (1 if the participant
has automated riding
experience before the test ride,
0 otherwise)
n
+
n
n
-
n
Highly interested in new
technologies indicator (1 if the
participant is highly interested
in new technologies, 0
otherwise)
+
+
+
n
n
-
Number of abrupt brakes
n
n
n
-
-
-
Note: In the initial attitude model, "+" indicates the variable with a positive effect on the initial attitude, "-" with a negative effect on the
initial attitude; In the attitude change model, "+" indicates the variable contributing to positive attitude change, "-" maintaining the same
attitude or worse; "n" indicates no significant effects.
Our results have several implications on the policymaking of future AV technology:
1. Standardized tests on the operation quality of AV shuttles may be necessary before AV shuttles are
permitted for public roads. Experimental results reveal that the poor operation comfort of AV shuttles
significantly undermines people's attitudes toward AV shuttles. Therefore, AV testing standards
established by policymakers could include the comfort evaluation of the rider. The comfort of riders
17
could be evaluated using the human body vibration test (Cardinale & Pope, 2003). For example, the
severity, frequency, and duration of AV riders' exposure to unsafe vibration could be evaluated and
limited to avoid the negative effects of immature technology.
2. Vulnerable road users' acceptance of AV shuttles needed to be considered in relevant AV planning
decisions. Experiments have shown that people perceive sharing the road with AV shuttles even more
negatively than riding the AV shuttle. To address these concerns, AV manufacturers could consider
human factors when designing AV shuttles and consider the behavior of other road users
comprehensively. AV testing standards might not only focus on the competency of AV shuttles to
successfully achieve driving tasks under every possible environment and traffic condition but also
address whether AV shuttles' behavior has negative impacts on surrounding road users. For example, a
slow operating speed may induce other drivers' lane-changing tendency, and sudden speed changes may
deteriorate the trust of surrounding road users on AV shuttles. Existing AV testing studies often neglect
the perspective of other road users (Nowakowski et al., 2015). Moreover, insurance companies may
also consider safety when AV shuttles interact with other road users when insuring AV shuttles.
3. This work pinpoints groups with positive initial attitudes toward AV shuttles. This will help AV shuttle
operating agencies identify targeted customers with positive initial attitudes toward AV shuttles and
select pilot sites for AV shuttles. For example, our research shows participants with long commute
times or high income are more likely to improve their attitude towards AV shuttle via test rides, which
could help select AV shuttle demonstration sites; Our research also indicates that participants with high
interest in new technologies have a more positive attitude towards AV shuttles. Thus, AV shuttle
operators can place their advertisements near technology-related news, videos, or self-published content.
In the future, the potential user targeting of AV shuttles can also help city planners to plan AV shuttle
routes rationally and improve the utilization of AV shuttles.
Since significant differences can be observed in speed, operating hours, and fares between the studied AV
shuttle and future regularly operated AV shuttles, future research can be conducted in several directions. First, AV
shuttles operating in many cities are reported to have a low-speed limit (Kolodge et al., 2020). The over-slow
operating speed may prevent AV shuttle from a practical solution to daily transportation and lead to more overtaking,
thus further disrupting the traffic. Future research could assess how AV driving modes affect other vehicles and
look into ways to lessen their negative effects while retaining the safety and comfort of AV shuttles. Second, the
AV shuttles investigated in this study are not allowed to travel in the rain or snow. However, prior research (Ahmed
et al., 2020) has demonstrated that equipment or systems failure in poor weather (storm, high wind, snow, rain, etc.)
is a significant public concern. Transit agencies could focus on rigorous assessment and validation of AV shuttles
in poor weather conditions to allay public fears. The next step of research might focus on generating extreme
conditions for AV training to improve their stability in extreme weather. The third point is that existing AV
demonstrations do not charge riders. This free-ride model is not sustainable, so the company will need to look at
pricing reasonableness in the future to maximize revenue. The effect of AV fares on mode selection and
transportation equity could be the subject of furture research.
REFERENCES
Ahmed, S. S., Pantangi, S. S., Eker, U., Fountas, G., Still, S. E., & Anastasopoulos, P. C. (2020). Analysis of
safety benefits and security concerns from the use of autonomous vehicles: A grouped random parameters
bivariate probit approach with heterogeneity in means. Analytic Methods in Accident Research, 28, 100134.
https://doi.org/10.1016/j.amar.2020.100134
Bergdahl, J. (2005). Sex differences in attitudes toward driving: A survey. The Social Science Journal, 42(4),
595–601. https://doi.org/10.1016/J.SOSCIJ.2005.09.006
Cardinale, M., & Pope, M. H. (2003). The effects of whole body vibration on humans: dangerous or
advantageous? Acta Physiologica Hungarica, 90(3), 195–206.
Chen, Z., & Li, X. (2021). Unobserved heterogeneity in transportation equity analysis: Evidence from a bike-
sharing system in southern Tampa. Journal of Transport Geography, 91(July), 6.
https://doi.org/10.1016/j.jtrangeo.2021.102956
18
Cunningham, M. L., Regan, M. A., Ledger, S. A., & Bennett, J. M. (2019). To buy or not to buy? Predicting
willingness to pay for automated vehicles based on public opinion. Transportation Research Part F: Traffic
Psychology and Behaviour, 65, 418–438. https://doi.org/10.1016/j.trf.2019.08.012
Dan, C. (2020). Easy Mile Autonomous Shuttle Employs V2X Communication.
https://www.designnews.com/automotive-engineering/easy-mile-autonomous-shuttle-employs-v2x-
communication
Das, S., Dutta, A., & Fitzpatrick, K. (2020). Technological perception on autonomous vehicles: perspectives of
the non-motorists. Technology Analysis and Strategic Management, 32(11), 1335–1352.
https://doi.org/10.1080/09537325.2020.1768235
Deb, S., Strawderman, L., Carruth, D. W., DuBien, J., Smith, B., & Garrison, T. M. (2017). Development and
validation of a questionnaire to assess pedestrian receptivity toward fully autonomous vehicles.
Transportation Research Part C: Emerging Technologies, 84, 178–195.
https://doi.org/10.1016/j.trc.2017.08.029
Eden, G., Nanchen, B., Ramseyer, R., & Evéquoz, F. (2017). Expectation and experience: passenger acceptance
of autonomous public transportation vehicles. Lecture Notes in Computer Science (Including Subseries
Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 10516 LNCS, 360–363.
https://doi.org/10.1007/978-3-319-68059-0_30
Hilgarter, K., & Granig, P. (2020). Public perception of autonomous vehicles: A qualitative study based on
interviews after riding an autonomous shuttle. Transportation Research Part F: Traffic Psychology and
Behaviour, 72, 226–243. https://doi.org/10.1016/j.trf.2020.05.012
Hulse, L. M., Xie, H., & Galea, E. R. (2018). Perceptions of autonomous vehicles: Relationships with road users,
risk, gender and age. Safety Science, 102(April 2017), 1–13. https://doi.org/10.1016/j.ssci.2017.10.001
Kolodge, K., Cicotte, S., & Peng, H. (2020). Mcity Driverless Shuttle: What We Learned About Consumer
Acceptance of Automated Vehicles. October, 1–11.
Krueger, R., Rashidi, T. H., & Rose, J. M. (2016). Preferences for shared autonomous vehicles. Transportation
Research Part C: Emerging Technologies, 69, 343–355. https://doi.org/10.1016/j.trc.2016.06.015
Liu, P., & Xu, Z. (2020). Public attitude toward self-driving vehicles on public roads: Direct experience changed
ambivalent people to be more positive. Technological Forecasting and Social Change, 151(November
2019), 119827. https://doi.org/10.1016/j.techfore.2019.119827
Liu, P., Xu, Z., & Zhao, X. (2019). Road tests of self-driving vehicles: Affective and cognitive pathways in
acceptance formation. Transportation Research Part A: Policy and Practice, 124(June 2018), 354–369.
https://doi.org/10.1016/j.tra.2019.04.004
Madigan, R., Louw, T., Wilbrink, M., Schieben, A., & Merat, N. (2017). What influences the decision to use
automated public transport? Using UTAUT to understand public acceptance of automated road transport
systems. Transportation Research Part F: Traffic Psychology and Behaviour, 50, 55–64.
https://doi.org/10.1016/j.trf.2017.07.007
Moták, L., Neuville, E., Chambres, P., Marmoiton, F., Monéger, F., Coutarel, F., & Izaute, M. (2017). Les
variables prédictives des intentions d’utilisation d’une navette autonome : aller au-delà du MAT et de la
TCP ? Revue Europeenne de Psychologie Appliquee, 67(5), 269–278.
https://doi.org/10.1016/j.erap.2017.06.001
NHTSA. (2016). Federal automated vehicles policy: Accelerating the next revolution in roadway safety. US
Department of Transportation.
Nordhoff, S., de Winter, J., Madigan, R., Merat, N., van Arem, B., & Happee, R. (2018). User acceptance of
automated shuttles in Berlin-Schöneberg: A questionnaire study. Transportation Research Part F: Traffic
Psychology and Behaviour, 58, 843–854. https://doi.org/10.1016/j.trf.2018.06.024
Nowakowski, C., Shladover, S. E., Chan, C. Y., & Tan, H. S. (2015). Development of California regulations to
govern testing and operation of automated driving systems. Transportation Research Record, 2489, 137–
144. https://doi.org/10.3141/2489-16
Penmetsa, P., Adanu, E. K., Wood, D., Wang, T., & Jones, S. L. (2019). Perceptions and expectations of
autonomous vehicles – A snapshot of vulnerable road user opinion. Technological Forecasting and Social
Change, 143(March), 9–13. https://doi.org/10.1016/j.techfore.2019.02.010
19
Pyrialakou, V. D., Gkartzonikas, C., Gatlin, J. D., & Gkritza, K. (2020). Perceptions of safety on a shared road:
Driving, cycling, or walking near an autonomous vehicle. Journal of Safety Research, 72, 249–258.
https://doi.org/10.1016/j.jsr.2019.12.017
Rahman, M. T., & Dey, K. (2022). Perceptions of Vulnerable Roadway Users on Autonomous Vehicle
Regulations. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.4051556
Reig, S., Norman, S., Morales, C. G., Das, S., Steinfeld, A., & Forlizzi, J. (2018). A field study of pedestrians and
autonomous vehicles. Proceedings - 10th International ACM Conference on Automotive User Interfaces and
Interactive Vehicular Applications, AutomotiveUI 2018, 198–209. https://doi.org/10.1145/3239060.3239064
Salonen, A. O. (2018). Passenger's subjective traffic safety, in-vehicle security and emergency management in the
driverless shuttle bus in Finland. Transport Policy, 61(April 2017), 106–110.
https://doi.org/10.1016/j.tranpol.2017.10.011
Sheela, P. V., & Mannering, F. (2020). The effect of information on changing opinions toward autonomous
vehicle adoption: An exploratory analysis. International Journal of Sustainable Transportation, 14(6), 475–
487. https://doi.org/10.1080/15568318.2019.1573389
Shi, X., Wang, Z., Li, X., & Pei, M. (2020). The Effect of Ride Experience on Changing Opinions Toward
Autonomous Vehicle Safety. July 2021, 1–10. https://doi.org/10.13140/RG.2.2.15902.56648/1
Templeton, B. (2020). EasyMile Self-Driving Shuttle Banned After Sudden Stop Hurts Passenger — Are Seatbelts
Needed? https://www.forbes.com/sites/bradtempleton/2020/03/03/easymile-self-driving-shuttle-banned-
after-sudden-stop-hurts-passenger---are-seatbelts-needed/?sh=2bcc6c677ad5
Washington, S., Karlaftis, M., Mannering, F., & Anastasopoulos, P. (2020). Statistical and econometric methods
for transportation data analysis. Chapman and Hall/CRC.
Xing, Y., Zhou, H., Han, X., Zhang, M., & Lu, J. (2022). What influences vulnerable road users' perceptions of
autonomous vehicles? A comparative analysis of the 2017 and 2019 Pittsburgh surveys. Technological
Forecasting and Social Change, 176(August 2020), 121454. https://doi.org/10.1016/j.techfore.2021.121454
Xu, Z., Zhang, K., Min, H., Wang, Z., Zhao, X., & Liu, P. (2018). What drives people to accept automated
vehicles? Findings from a field experiment. Transportation Research Part C: Emerging Technologies,
95(June), 320–334. https://doi.org/10.1016/j.trc.2018.07.024
20
APPENDIX: Questionnaire
