You are on page 1of 11

Hindawi

Journal of Advanced Transportation


Volume 2022, Article ID 8461212, 11 pages
https://doi.org/10.1155/2022/8461212

Research Article
Factors Influencing the Acceptance of Robo-Taxi Services in
China: An Extended Technology Acceptance Model Analysis

Mingyu Liu ,1 Jianping Wu ,1 Chunli Zhu ,2 and Kezhen Hu 3

1
Department of Civil Engineering, Tsinghua University, Beijing 100084, China
2
School of Information and Electronics & Advanced Research Institute of Multidisciplinary Science,
Beijing Institute of Technology, Beijing 100081, China
3
China Academy of Information and Communication Technology, Beijing 100191, China

Correspondence should be addressed to Jianping Wu; jianpingwu@tsinghua.edu.cn

Received 13 July 2021; Revised 7 March 2022; Accepted 24 March 2022; Published 9 April 2022

Academic Editor: Yuchuan Du

Copyright © 2022 Mingyu Liu et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

With the development of autonomous driving technologies, robo-taxis (shared autonomous vehicles) are being tested on real
roads. In China, in particular, people in some cities such as Beijing and Shanghai can book a robo-taxi online and experience the
service. To examine the influential factors on user acceptance of robo-taxi services, this study proposes and employs an extended
technology acceptance model (TAM) with four external factors: perceived trust, government support, social influence, and
perceived enjoyment. Data were collected through an online questionnaire in China, and responses from 403 respondents were
analyzed using structural equation modeling. Both typical TAM factors—including perceived ease of use, perceived usefulness,
and attitude—and external factors were found to play significant roles in predicting users’ intention to use robo-taxis. The four
external factors influenced the user acceptance indirectly via typical TAM factors. Improving users’ perceived trust is important
for increasing public adoption. A greater emphasis by manufacturers on safety concerns, wider dissemination of information on
data protection and safety systems, and government support through incentives for manufacturers and users can help improve
public adoption of robo-taxi services.

1. Introduction the cost of purchase, maintenance, repair, and insurance [4].


Moreover, an agent-based study showed that, in a system of
Recently, autonomous driving technologies have been shared AVs (SAVs), each SAV can replace 11 conventional
studied by many researchers. Autonomous vehicles (AVs) vehicles [5]. The robo-taxis can thus alleviate parking
have become a major challenge for manufacturers. Many problems in modern cities by reducing private car owner-
companies and research groups, such as Google, Daimler, ship levels.
and Deutsche Bahn, have focused on the design and testing China has made significant progress in the development
of AVs [1]. However, compared with manually driven ve- of robo-taxis. On June 21, 2019, Baidu’s Apollo robo-taxis
hicles, AVs involve high costs, with up to 50,000 USD per obtained 45 licenses for testing in Changsha City, Hunan
vehicle, which has become a considerable challenge [2]. Province [6]. In August 2019, testing of Apollo robo-taxis
Therefore, robo-taxis have been considered an effective began on actual roads in Changsha City [7]. On June 27,
approach to accelerate the industrialization of AVs. This 2020, Didi, a vehicle-for-hire company providing bike-
study considered SAE L5 AVs with sharing services [3], sharing and ride-hailing services in China, started to offer
robo-taxis that can meet travel demand with an autonomous robo-taxi services for testing in a specific area in Shanghai
driving system. Similar to ride-hailing services and taxis, [8]. Users can book robo-taxi services for free on their
robo-taxis are intended to provide traffic services to con- mobile phones. As a measure against the COVID-19 pan-
sumers. Sharing services can save money for users, including demic, these robo-taxis have devices that can recognize
2 Journal of Advanced Transportation

whether users have worn a mask before they board the robo- However, limited studies have used a social-psycho-
taxi. During the trip, users can interact with the robo-taxi logical perspective to examine the adoption and acceptance
system. Moreover, they are accompanied by two people, one of robo-taxi services or SAVs, especially in China. Liu et al.
to ensure the users’ safety and the other to make sure the investigated the effects of SI, PU, and perceived ease of use
robo-taxi is driven normally. In recent years, robo-taxis have (PEU) on public adoption of robo-taxis [22]. These factors
undergone various developments, and they will likely be were found to be strong predictors of consumers’ behavioral
widely introduced soon. Therefore, before their arrival on responses to IU robo-taxis. Yuen et al. combined TPB with
the market, the factors that influence consumers’ acceptance UTAUT2 to analyze the factors influencing the adoption of
of robo-taxi services must be explored. SAVs in Vietnam [23]. UTAUT2 factors, including per-
Some studies have investigated public adoption of SAVs formance expectation, effort expectation, habit, price value,
using discrete choice experiments (e.g., [4, 9, 10]). However, and hedonic motivation, were mediated by attitudes toward
major roadblocks to the widespread adoption of AVs are not SAVs, a TPB factor. All TPB and UTAUT2 factors were
technical but psychological in nature [11]. Therefore, user effective predictors of the intention to use SAVs.
acceptance of robo-taxis must be investigated from a psy- Government support (GS) is an important factor that
chological perspective. To examine user acceptance of in- influences people’s acceptance of new technologies. How-
formation technologies from such a perspective, researchers ever, few studies have explored its effects using a social-
have proposed many models and frameworks, such as the psychological model. Khoo and Ong introduced a govern-
theory of planned behavior (TPB) [12], technology accep- ment policy factor into the subjective norm construct, a basic
tance model (TAM) [13], unified theory of acceptance and part of TPB [24]. They demonstrated this factor to be a
use of technology (UTAUT [14]), and UTAUT2 [15]. TPB strong predictor of users’ intention to use sustainable
was developed based on the theory of reasoned action transport. However, the factor’s items were designed to
(TRA), and it can be applied to explain many types of be- understand the acceptance behavior of sustainable transport,
haviors, such as pedestrian behavior. To understand users’ which was already on the market. Although robo-taxis have
acceptance on technologies specifically, TAM was proposed. been tested in several countries worldwide, they are not yet a
In fact, typical TAM is highly extendable; UTAUT and mature product. Therefore, new items for the GS factor,
UTAUT2 are extended theories based on typical TAM. specifically related to robo-taxis, should be designed.
Compared with TAM, although UTAUT and UTAUT2 This study, therefore, developed robo-taxi-related items
consider more factors, some factors are not suitable to learn for the GS factor, which can also be used for other products
robo-taxi services, such as the factor of habit in UTAUT2, in development. To the best of the authors’ knowledge, this
which requires experience on robo-taxis. Many researchers study is the first to introduce GS into the TAM framework to
have thus investigated users’ adoption of various technol- investigate the role of government policy on public adop-
ogies based on the TAM framework. For example, Vanduhe tion. The proposed extended TAM includes four external
et al. extended TAM with task technology fit, social influence variables: GS, PT, PE, and SI. Structural equation modeling
(SI), and social motivation to investigate the factors that (SEM) was conducted to test several hypotheses from the
influence employees’ intention to use gamification for extended TAM to explain the effect of these factors on public
training [16]. Salloum et al. proposed an extended TAM to adoption of robo-taxi services. The results can be good
investigate e-learning acceptance with eight external vari- references for manufacturers, and some policies were also
ables, which were summarized into two constructs: system provided for governments and manufacturers to improve
characteristics and individual factors [17]. Rahman et al. public adoption of robo-taxi services.
developed a model of driver acceptance of driver support Following this introduction, which presents a literature
systems by combining TAM with TPB; the model provides review of works relating to robo-taxis and using social-
an improved understanding of the formation of driver ac- psychological models, the remainder of this paper is orga-
ceptance [18]. nized as follows. Section 2 describes the research model and
Social-psychological models have also been used to various proposed hypotheses. Subsequently, the study
examine AVs. Madigan et al. proposed a UTAUT mod- methodology—including survey design, data collection,
el—which included the factors of performance expectancy, participants, and statistical analysis methods—is presented.
effort expectancy, and SI—to investigate people’s acceptance Thereafter, the results of the reliability and validity and SEM
of automated road transport systems (ARTS) [19]. Re- analyses are reported. Lastly, and before concluding the
spondents in Lausanne (Switzerland) and La Rochelle paper, the effects of the investigated factors and some related
(France) had used autonomous vehicles as part of the implications are discussed.
CityMobil2 trials. Wu et al. introduced environmental
concerns into TAM, replacing perceived usefulness (PU) 2. Research Model
with green PU, to study public adoption of autonomous
electric vehicles (AEVs) and the effects of environmental 2.1. Typical TAM. A typical TAM includes PU, PEU, attitude
factors on AEVs [20]. Xu et al. explored the influence of the (ATT), and intention to use (IU). PU can be defined as the
direct experience of an automated vehicle (Level 3) and extent to which users believe that a specific technology or
explained and predicted public acceptance of AVs through service can enhance overall work performance [13]. PEU can
an extended TAM, which included perceived trust (PT) and be defined as the degree to which users believe that
perceived safety [21]. employing a specific technology or service could be free from
Journal of Advanced Transportation 3

mental and physical effort [13]. ATT can be defined as users’ Perceived
preference when using specific technologies or systems [25]. Usefulness

IU can be defined as users’ desire to use particular services


and technologies [13]. Previous research has found positive External Intention to
relationships between PEU and PU, PU and ATT, PEU and Attitudes
Variables Use
ATT, and ATT and IU, as shown in Figure 1. Therefore, the
following hypotheses are proposed:
Perceived
H1: PU (of robo-taxis) is positively associated with user Ease of Use
ATT (toward robo-taxis)
Figure 1: Typical technology acceptance model.
H2: PU is positively associated with IU (robo-taxis)
H3: PEU (of robo-taxis) is positively associated with
user ATT should use a particular technology [14]. SI plays an im-
portant role in public acceptance of new technologies, such
H4: PEU is positively associated with PU as AVs [19]. Acceptance of AVs by friends or family
H5: ATT is positively associated with IU members increased users’ confidence in the technology as
well as their intention to purchase or actual purchase
[26, 31]. SI has also been tested in other psychological
2.2. Perceived Trust. PT indicates the degree to which models, such as UTAUT [14] and UTAUT2 [15]. It has been
consumers generally trust a particular technology system regarded as a basic factor when exploring users’ acceptance
[26]. Many studies on the public adoption of AVs have of new technologies. Therefore, SI was added to TAM as an
focused on the effect of PT or perceived risk on user ac- external factor to explore its effect on user acceptance of
ceptance. Safety factors have been demonstrated as strong robo-taxi services.
predictors of public adoption [26–28]. In fact, consumers
have always been concerned about the safety of robo-taxis
and AVs. This safety concern includes data safety concerns 2.5. Perceived Enjoyment. PE here is similar to hedonic
and body safety concerns. On the one hand, users may be motivation in UTAUT2; it indicates the perceived degree
concerned about personal data collected illegally by others; of fun or pleasure experienced by users when using a
on the other hand, whether robo-taxis can perform better particular technology [15]. With the development of
and safer than normal vehicles requires to be ensured by economics and technology, traffic tools can meet people’s
users themselves. Therefore, PT can be regarded as an es- travel demands more easily. Furthermore, comfort and
sential factor in investigating user acceptance of robo-taxis. fun during a trip are being increasingly considered by
users. However, PE or hedonic motivation has only been
studied under the UTAUT2 model. Hedonic motivation
2.3. Government Support. GS indicates the extent to which
has been found to be a strong predictor of user accep-
users’ willingness to use is promoted when the government
tance of ARTS [32] and autonomous delivery vehicles
implements supportive policies or provides incentives for a
[33]. Besides, hedonic motivation was validated to be a
particular technology. The Chinese government has sup-
good predictor in other products, such as ride-hailing
ported new developments in many technologies. For ex-
services [34], bike-sharing [35],and social networks for
ample, in 2010, the Central People’s Government of the
education [36]. Therefore, PE from robo-taxis can be
People’s Republic of China (CPGPRC) selected five cities,
hypothesized to be influential in the context of robo-taxi
including Shanghai, Changchun, Shenzhen, Hefei, and
acceptance.
Hangzhou, for a pilot project on electric and hybrid vehicles.
Based on the typical TAM, these external variables may
Each plug-in hybrid electric vehicle can cost up to 50,000
affect users’ intention to use robo-taxis indirectly via PU and
yuan, while each pure electric vehicle can cost up to 60,000
PEU. Therefore, PT, GS, SI, and PE were hypothesized to
yuan [29]. One study showed that government incentives
affect PU and PEU (Figure 2). The proposed hypotheses are
can significantly promote users’ intention to purchase; when
as follows:
people knew more about this policy, this effect was found to
be more significant [30]. Besides, some other policies can H6: PT (in robo-taxis) is negatively associated with PU
also be regarded as government support. For example, if
H7: PT is negatively associated with PEU
robo-taxis can use bus lanes or autonomous vehicle lanes,
they will save travel time for users. Therefore, users will be H8: GS is positively associated with PU
more likely to accept robo-taxi services. In the present study,
GS is hypothesized to be an external TAM factor that can H9: GS is positively associated with PEU
influence public adoption of robo-taxi services. H10: SI is positively associated with PU
H11: SI is positively associated with PEU
2.4. Social Influence. SI indicates the degree to which an H12: PE (from robo-taxis) is positively associated with PU
individual perceives that important people believe he or she H13: PE is positively associated with PEU
4 Journal of Advanced Transportation

Perceived trust

Perceived usefulness Attitudes

Government support

Social influence

Perceived ease of use Intention to Use

Perceived enjoyment

Figure 2: Extended technology acceptance model.

3. Methodology were taken. Respondents with the same IP address or


username could only answer the questionnaire once. The
3.1. Survey Design. The data were collected using an online response time for each item should be less than 5 minutes.
questionnaire that consisted of three sections. The first Some attention-check questions randomly appeared while
section was related to demographic information, including answering the questionnaire, for example, “This is a quality
age, gender, education, and income, with a total of seven control question. Please select all the options, including the
items. The second section included items for all predictive word ‘frog’.” Furthermore, all items in the second section
factors of the proposed model (Table 1): PT (three items), GS appeared in a random order. However, the output was
(three items), SI (three items), PE (two items), PU (three ordered normally. To encourage more participation in the
items), PEU (three items), and ATT (three items). Based on survey, bonuses were provided for valid answers.
different categories of Chinese policies for supporting new
technologies, three scenarios were mentioned in the GS
construct: providing incentives for users, providing incen- 3.3. Participants. A total of 914 questionnaires (540 com-
tives for manufacturers, and providing traffic policy support pleted) were received, with 403 valid questionnaires. The
for robo-taxis. A five-point Likert scale was used in this demographic information of the respondents is shown in
section, with 1 being “strongly disagree” and 5 being Table 2. Of the valid respondents, 53.3% were women, while
“strongly agree.” The third section included items for IU. 46.7% were men. Furthermore, more than 90% of the re-
This section too employed a five-point Likert scale, with 1 spondents were between 18 and 44 years of age, indicating
being “extremely unlikely” and 5 being “extremely likely.” that this study is more useful for understanding young
people’s adoption of robo-taxis in China. Most participants
were well-educated, with a bachelor’s degree or above
3.2. Data Collection. The survey was conducted using (85.6%). Moreover, more than half of the respondents
Wenjuanxing (https://www.wjx.cn/), one of the largest (64.5%) had a personal annual income of between 40,000
online survey platforms in China. A link was disseminated and 200,000 yuan. Notably, 18.4% had no income and were
through social networks (e.g., Weibo and Wechat) and the likely to be students. Most respondents had driving licenses
Wenjuanxing platform itself. Respondents from different (73.4%), and more than half had their own vehicles (65%).
age groups, education levels, and income groups were in- Lastly, 58.1% chose to drive by themselves during a trip.
cluded in the sample. Data were collected from February
2020 to March 2020. The initial questionnaire was designed
in English by using English references. Several translators 3.4. Statistical Analysis. Statistical analyses were performed
were asked to independently translate the questionnaire into using SPSS version 23 and AMOS version 22. First, Cron-
Chinese and then translate it back into English with random bach’s alpha was calculated using SPSS to determine the
answers. Respondents were provided with a brief intro- reliability of the factor structure. A reliability coefficient of
duction on robo-taxi services to ensure proper under- 0.7 or more is generally considered acceptable [38].
standing of such services. To ensure the effective Moreover, confirmatory factor analysis (CFA), based on
administration of the questionnaire, the following measures the maximum likelihood estimation method, was used to
Journal of Advanced Transportation 5

Table 1: Instrument measurement items and their sources.


Factor Item Literature source
PT1: I generally have concerns about using robo-taxis
Modified from
Perceived trust (PT) PT2: robo-taxis are somewhat frightening to me
[26]
PT3: I have concerns about system security and data privacy of robo-taxis
GS1: I am more likely to use robo-taxis if the government provides incentives for users
GS2: I am more likely to use robo-taxis if the government provides policy support, such as
Government support
permission to use them on bus lanes Self-developed
(GS)
GS3: I am more likely to use robo-taxis if the government provides incentives for
manufacturers
SI1: I would be proud if people saw me using a robo-taxi
Modified from
Social influence (SI) SI2: people whose opinions I value would like to use robo-taxis
[19, 26]
SI3: I would be more likely to use robo-taxis if my friends and family members used them
Perceived enjoyment PE1: using robo-taxis would be very fun
Modified from [15]
(PE) PE2: using robo-taxis would be very entertaining and relaxing
PU1: I think robo-taxis would be useful in meeting my transportation needs
Perceived usefulness Modified from
PU2: I think using robo-taxis would make me feel safer (than using normal vehicles)
(PU) [26]
PU3: I think the use of robo-taxis would decrease accidents
PEU1: I think it would be easy to learn to operate robo-taxis
Perceived ease of use Modified from
PEU2: my interactions with robo-taxis would be clear and easy to understand
(PEU) [20]
PEU3: I think it would be easy for me to use robo-taxis to go anywhere I wanted
ATT1: I think using robo-taxis would be a good idea
Modified from
Attitude (ATT) ATT2: I think using robo-taxis would be a wise idea
[37]
ATT3: I like the idea of robo-taxis
IU1: assuming robo-taxis come into use, I intend to use them Modified from
Intention to use (IU)
IU2: assuming robo-taxis come into use, I intend to recommend them to others [20]

Table 2: Demographic information of the respondents.


Variables Category Frequency (%)
Male 188 (46.7%)
Gender
Female 215 (53.3%)
18–24 125 (31.0%)
25–34 185 (45.9%)
Age
35–44 67 (16.6%)
Over 45 26 (6.5%)
Master’s degree or above 49 (12.2%)
Education Bachelor’s degree 296 (73.4%)
High school or below 58 (14.4%)
None 74 (18.4%)
Less than 40,000 54 (13.4%)
Annual income (yuan) 40,000–100,000 142 (35.2%)
100,000–200,000 118 (29.3%)
More than 200,000 15 (3.7%)
Yes 296 (73.4%)
Driving license ownership
No 107 (26.6%)
Yes 262 (65.0%)
Private vehicle(s) ownership
No 141 (35.0%)
Yes 234 (58.1%)
Drive by yourself
No 169 (41.9%)

assess the suitability of the whole scale. Items with factor the square root of AVE for all constructs should be larger
loadings lower than 0.4 were removed (e.g., [39]). Subse- than the correlation coefficients of all constructs [43].
quently, convergent and discriminant validity were calcu- Furthermore, to obtain a good model fit, a single index
lated to evaluate the construct validity of the measurement cannot provide a reliable measure across situations; multiple
instrument. Convergent validity (using average variance indices should be reported [42, 44]. Therefore, the following
extracted, AVE) denotes whether items in the same con- indices are reported here: normed chi-square, normed fit
struct have internal consistency [40], while discriminant index (NFI), comparative fit index (CFI), Tucker–Lewis
validity refers to the degree to which a given construct differs index (TLI), incremental fit index (IFI), root mean square
from another construct [41]. Generally, an AVE of more error of approximation (RMSEA), and standardized root
than 0.50 is acceptable [42]. Regarding discriminant validity, mean square residual (SRMR). Acceptable values for a good
6 Journal of Advanced Transportation

Table 3: Results of the measurement model fit, reliability, and convergent validity.
Constructs Items Means Standard deviation Factor loadings Composite reliability AVE
PT1 3.04 0.98 0.73 0.78 0.54
Perceived trust (α � 0.78) PT2 2.55 0.95 0.71
PT3 3.36 1.05 0.77
GS1 4.05 0.74 0.78 0.78 0.54
Government support (α � 0.79) GS2 3.97 0.76 0.69
GS3 3.96 0.70 0.74
SI1 3.44 0.79 0.65 0.72 0.46
Social influence (α � 0.72) SI2 3.70 0.80 0.74
SI3 3.96 0.74 0.64
PE1 3.87 0.80 0.70 0.69 0.53
Perceived enjoyment (α � 0.68)
PE2 3.63 0.86 0.75
PU1 3.63 0.74 0.74 0.71 0.46
Perceived usefulness (α � 0.74) PU2 3.28 0.86 0.66
PU3 3.30 0.86 0.62
PEU1 3.90 0.83 0.73 0.73 0.48
Perceived ease of use (α � 0.74) PEU2 3.73 0.73 0.67
PEU3 3.68 0.81 0.67
ATT1 3.84 0.81 0.80 0.80 0.57
Attitude (α � 0.80) ATT2 3.68 0.79 0.76
ATT3 3.96 0.78 0.71
IU1 3.86 0.77 0.81 0.74 0.59
Intention to use (α � 0.73)
IU2 3.68 0.91 0.72

Table 4: Results of correlation matrix.


PT GS SI PE PU PEU ATT IU
PT 0.735
GS −0.344 0.735
SI −0.416 0.583 0.678
PE −0.432 0.497 0.574 0.728
PU −0.582 0.504 0.590 0.567 0.678
PEU −0.391 0.594 0.549 0.582 0.548 0.693
ATT −0.520 0.594 0.614 0.664 0.663 0.637 0.755
IU −0.511 0.574 0.639 0.604 0.612 0.582 0.674 0.768
PT, perceived trust; GS, government support; SI, social influence; PE, perceived enjoyment; PU, perceived usefulness; PEU, perceived ease of use; ATT,
attitude; IU, intention to use.

model fit are as follows: χ 2/df < 3, NFI >0.90, CFI >0.90, TLI studies have shown that an AVE higher than 0.45 can also be
>0.90, IFI >0.90, RMSEA <0.08, and SRMR <0.10 acceptable [46, 47]. Therefore, all constructs had acceptable
[37, 44, 45]. convergent validity.
For discriminant validity, the square root of AVE was
4. Results used to correlate the latent constructs. Table 4 shows that the
square root of the AVE for all constructs was higher than the
4.1. Reliability and Validity Analysis. The reliability and pairwise correlations. Therefore, all constructs also had
validity of the research instrument were calculated before acceptable discriminant validity.
performing SEM. Table 3 presents the fitness of the mea- In addition, a CFA was conducted for PT, GS, SI, and PE;
surement model and the reliability and validity of each the standardized solution is shown in Figure 3. The model fit
construct. Most constructs had acceptable reliability, with indices were acceptable, with χ 2/df � 2.365, χ2 � 89.883,
Cronbach’s alphas of more than 0.70, except for PE df � 38, CFI � 0.966, IFI � 0.966, TLI � 0.951, NFI � 0.943,
(α � 0.68). However, Cronbach’s alpha of PE can also be SRMR � 0.047, and RMSEA � 0.058.
considered acceptable because it was only slightly lower than
0.70. Therefore, all the constructs used in this study can be
regarded as reliable. 4.2. Structural Equation Modeling. SEM was used to test the
CFA was used to assess the convergent and discriminant proposed hypotheses. The results indicated a good model fit:
validity of the measurement model. As shown in Table 3, all χ 2/df � 2.164, χ 2 � 411.201, df � 190, CFI � 0.946, IFI � 0.947,
constructs have composite reliability higher than 0.6 [42]. TLI � 0.935, NFI � 0.905, SRMR � 0.044, and RMSEA �
The AVE of PT, GS, PE, ATT, and IU was higher than 0.50, 0.054. The hypothesis testing coefficients are shown in Table 5
while that of SI, PEU, and PU were lower than 0.50. Some and Figure 4.
Journal of Advanced Transportation 7

.52

e3 PT1
53 .72
.72
e2 PT2 Perceived trust
58 .76

e1 PT3

-.44
.63

e6 GS1
49 .80
.70
e5 GS2 Government support
51 .71 -.55

e4 GS3

.78 -.60
.43

e9 SI1
55 .65
.74
e8 SI2 Social influence
41 .64 .66

e7 SI3

.81

.47

e11 PE1 .68


58 Perceived enjoyment
.76
e10 PE2

Figure 3: Standardized solution of confirmatory factor analysis.

Table 5: Results of hypotheses testing.


Hypotheses Hypothesized effect Significance and significance level Direct effect
∗∗∗
H01 PU ⟶ ATT 0.592
∗∗
H02 PU ⟶ IU 0.494
∗∗∗
H03 PEU ⟶ ATT 0.405
H04 PEU ⟶ PU (Not supported) —
∗∗
H05 ATT ⟶ IU 0.452
∗∗∗
H06 PT ⟶ PU −0.377
H07 PT ⟶ PEU (Not supported) —
H08 GS ⟶ PU (Not supported) —
∗∗∗
H09 GS ⟶ PEU 0.426

H10 SI ⟶ PU 0.317
H11 SI ⟶ PEU (Not supported) —

H12 PE ⟶ PU 0.351
∗∗∗
H13 PE ⟶ PEU 0.586

p ≤ 0.05, ∗∗ p ≤ 0.01, ∗∗∗ p ≤ 0.001; PU, perceived usefulness; PEU, perceived ease of use; ATT, attitude; IU, intention to use; PT, perceived trust; GS,
government support; SI, social influence; PE, perceived enjoyment.
8 Journal of Advanced Transportation

e12 e13 e14 e22 e23 e24

e3 PT1 .73
.71
e2 PT2 .77 Perceived trust PU1 PU2 PU3 ATT1 ATT2 ATT3

e1 PT3 .76
-.38 .66 .80 .71
.74 .62
e27

- .43 .59
Perceived usefulness Attitudes e25

.17
e6 GS1 .78
.69
e5 GS2 .74 Government support
.01 .49
e4 GS3
.32
-.55 - .09 .45

-.59 .78
.43
.41
.67 .35
e9 SI1 .65
.74
e8 SI2 .64 Social influence
-.03
e7 SI3
Perceived ease of use Intention to use e26

.73 .67
.67 .81 .72
.81 e28
.59

PEU1 PEU2 PEU3 IU1 IU2

e11 PE1 .70


.75 Perceived enjoyment
e10 PE2
e17 e16 e15 e21 e20

Figure 4: Standardized solution for the structural model.

First, the typical TAM hypotheses were tested. PU was important for public adoption. In fact, safety has always
found to significantly affect ATT (β � 0.592, p < 0.001) and been an important aspect of AV design [26]. However,
IU (β � 0.494, p < 0.01), supporting H1 and H2. PEU had a people may not know much about data protection and
significant effect on ATT (β � 0.405, p < 0.001), supporting safety measures and may thus have neutral attitudes [27].
H3. However, no significant relationship was found between Therefore, on the one hand, manufacturers should con-
PEU and PU; thus, H4 was rejected. Furthermore, ATT had tinue to make efforts to improve the safety of such
a positive effect on IU (β � 0.452, p < 0.01), supporting H5. technologies. On the other hand, whether robo-taxis are
Second, the hypotheses related to the four external as safe or safer than manually driven vehicles should be
variables were tested. PT had a significant negative effect on confirmed. One possible approach to disseminate safety
PU (β � −0.377, p < 0.001). However, no significant rela- information is that, in the early stage, people could be
tionship was found between PT and PEU. Therefore, H6 was allowed to use robo-taxis for free or even with some
supported but H7 was rejected. Furthermore, GS had a incentives to help them recognize that robo-taxis are safe.
positive effect on PEU (β � 0.426, p < 0.001), but it did not Moreover, to reduce losses when accidents do occur, some
have a significant effect on PU. Therefore, H8 was rejected related insurance can also be provided to users.
but H9 was supported. Regarding SI, it had a significant In addition, GS indirectly influenced IU through PEU. In
positive effect on PU (β � 0.317, p < 0.05). However, it did China, GS can increase people’s confidence in new tech-
not have a significant effect on PEU. Therefore, H10 was nologies and their intention to purchase [30]. Therefore, GS
supported but H11 was rejected. Lastly, PE had a significant could likely play an important role in increasing public
positive effect on both PU (β � 0.351, p < 0.05) and PEU acceptance of robo-taxi services. Governments can provide
(β � 0.586, p < 0.001), supporting H12 and H13. When we incentives for users, manufacturers, and service providers to
deleted the unsupported hypothesis, the standardized so- reduce related costs, thereby encouraging more people to
lution for the revised model is shown in Figure 5. All the accept robo-taxi services. Traffic rules prioritizing robo-taxis
paths were significant enough (p < 0.05). can also be implemented. For example, robo-taxis can be
permitted to use bus lanes to reduce travel time and improve
5. Discussion travel quality.
Regarding SI, it positively and indirectly influenced user
In line with previous research on typical TAM, this study acceptance through PU. As SI represents the influence of
found significant positive relationships between PU and friends and family members on users [19, 26], this result
ATT, PU and IU, PEU and ATT, and ATT and IU. However, indicates that the adoption of robo-taxis by friends and
PEU did not significantly affect PU, which contradicts family members can promote user acceptance of such ser-
typical TAM results. Nevertheless, some studies have shown vices. A motivation mechanism to encourage existing users
that the effect of PEU on PU is not strong [48]. to invite new users can thus be set to promote greater user
Furthermore, the four external variables PT, GS, SI, acceptance. For example, existing users who invite new users
and PE affected PEU or PU, indirectly influencing ATT can receive incentives or coupons. Due to the influence of
and IU. PT indirectly influenced user acceptance via PU, other people, new users will be more willing to accept robo-
indicating that improving users’ PT in robo-taxis is taxis.
Journal of Advanced Transportation 9

e12 e13 e14 e22 e23 e24

e3 PT1 .73
.71
e2 PT2 .77 Perceived trust PU1 PU2 PU3 ATT1 ATT2 ATT3
e1 PT3 .76
-.38 .74 .66 .62 .80 .71
e27

-.44 .57
Perceived usefulness Attitudes e25

e6 GS1 .78
.69
e5 GS2 .73 Government support
.46
e4 GS3
.47
-.55 .48
-.58 .80
.42
.43
.68 .26
e9 SI1 .64
.74
e8 SI2 .64 Social influence

e7 SI3
Perceived ease of use Intention to use e26

.73 .66
.67 .81 .72
.81 e28
.55
PEU1 PEU2 PEU3 IU1 IU2

e11 PE1 .70


.75 Perceived enjoyment
e10 PE2
e17 e16 e15 e21 e20

Figure 5: Standardized solution for the revised structural model.

Lastly, PE had an indirect effect on user acceptance of This study, however, has some limitations, which should
robo-taxis via both PEU and PU, indicating that when be noted here. Data were collected through an online survey,
people enjoy the convenience and comfort of traffic tools, which can lead to nonresponse bias, among other things. Of
their demand for such services increased. Therefore, a the respondents, 93.5% were in the 18–44 year age group and
comfortable and fun design should be adopted to develop 85.6% had a bachelor’s degree or above, thus reducing the
robo-taxi services. For example, providing movies and food generalizability of the results. However, although some
as an additional service can improve the user experience. companies have tested robo-taxis on actual roads, they will
take more time to successfully introduce robo-taxi services
into the market. Young and well-educated people are po-
6. Conclusion tential groups to first accept such services. Therefore, the
findings of this study are still useful for designers of robo-
As robo-taxis are gradually entering the market in China, taxis. Future research can survey people who have used
this study proposed an extended TAM model to explore the robo-taxi services to conduct a more in-depth investigation
factors that influence user acceptance of robo-taxi services. of how to improve user adoption of robo-taxis.
The analysis validated typical TAM, with PU and ATT
having a direct effect on IU, and PEU and PU having an
indirect effect on IU via ATT. Moreover, PT, GS, SI, and PE Data Availability
have indirect effects on IU via PEU and PU. All these factors
The data generated during the current study are owned by
were found to be effective predictors of IU.
the Future Transport Research Center, Tsinghua University,
Based on these findings, the following suggestions can be
and are not publicly available. The data can be obtained from
made to improve public adoption of robo-taxis. First, as
the corresponding author upon request.
safety is a major concern, manufacturers and service pro-
viders should improve the safety of robo-taxi services.
Moreover, information on the safety performance of such Conflicts of Interest
services should be disseminated to the public through ad-
vertisements and other forms of communication to increase The authors declare that there are no conflicts of interest
their PT. Second, GS can effectively improve robo-taxi regarding the publication of this article.
adoption. Providing incentives for users and manufacturers
can help users to save money and increase their confidence Acknowledgments
in robo-taxi services. Prioritizing robo-taxis on roads, such
as through permission to use them on bus lanes, can help The authors gratefully acknowledge the support provided by
passengers save time. Third, incentives and coupons can be the NSFC Zhejiang Joint Fund for the Integration of In-
provided to existing users who invite new users to further dustrialization and Informatization (U1709212), and “Re-
promote user adoption through SI. Lastly, robo-taxis can search on frontiers of intelligent transport system” funded
provide entertainment services, such as movies, food, and by the China Association for Science and Technology
games, to make passengers’ experience more enjoyable. (2018DX2QY04).
10 Journal of Advanced Transportation

References [19] R. Madigan, T. Louw, M. Dziennus et al., “Acceptance of


automated road transport systems (ARTS): an adaptation of
[1] K. Merfeld, M.-P. Wilhelms, S. Henkel, and K. Kreutzer, the UTAUT model,” Transportation Research Procedia,
“Carsharing with shared autonomous vehicles: uncovering vol. 14, pp. 2217–2226, 2016.
drivers, barriers and future developments - a four-stage [20] J. Wu, H. Liao, J.-W. Wang, and T. Chen, “The role of en-
Delphi study,” Technological Forecasting and Social Change, vironmental concern in the public acceptance of autonomous
vol. 144, pp. 66–81, 2019. electric vehicles: a survey from China,” Transportation Re-
[2] A. Stocker and S. Shaheen, “Shared automated vehicles: re- search Part F: Traffic Psychology and Behaviour, vol. 60,
view of business models,” International Transport Forum pp. 37–46, 2019.
Discussion Paper, 2017. [21] Z. Xu, K. Zhang, H. Min, Z. Wang, X. Zhao, and P. Liu, “What
[3] P. Planing and M. Dursun, “Acceptance OF shared autono- drives people to accept automated vehicles? Findings from a
mous vehicles-a correspondence analysis OF new car buyer field experiment,” Transportation Research Part C: Emerging
attitudes,” International Journal of Sales, Retailing & Mar- Technologies, vol. 95, pp. 320–334, 2018.
keting, vol. 7, no. 2, 2018. [22] M. Liu, J. Wu, C. Zhu, and K. Hu, “A study on public adoption
[4] R. Krueger, T. H. Rashidi, and J. M. Rose, “Preferences for of robo-taxis in China,” Journal of Advanced Transportation,
shared autonomous vehicles,” Transportation Research Part C: vol. 2020, Article ID 8877499, 8 pages, 2020.
Emerging Technologies, vol. 69, pp. 343–355, 2016. [23] K. F. Yuen, D. T. K. Huyen, X. Wang, and G. Qi, “Factors
[5] D. J. Fagnant and K. M. Kockelman, “The travel and envi- influencing the adoption of shared autonomous vehicles,”
ronmental implications of shared autonomous vehicles, using International Journal of Environmental Research and Public
agent-based model scenarios,” Transportation Research Part Health, vol. 17, no. 13, p. 4868, 2020.
C: Emerging Technologies, vol. 40, pp. 1–13, 2014. [24] H. L. Khoo and G. P. Ong, “Understanding sustainable
[6] http://www.sohu.com/a/322970509_782058. transport acceptance behavior: a case study of Klang valley,
[7] http://auto.sina.com.cn/zz/sh/2019-09-26/detail- Malaysia,” International Journal of Sustainable Trans-
iicezueu8620718.shtml. portation, vol. 9, no. 3, pp. 227–239, 2015.
[8] https://www.sohu.com/a/404927011_120058293. [25] F. D. Davis, “User acceptance of information technology:
[9] J. Webb, C. Wilson, and T. Kularatne, “Will people accept system characteristics, user perceptions and behavioral im-
shared autonomous electric vehicles? A survey before and pacts,” International Journal of Man-Machine Studies, vol. 38,
after receipt of the costs and benefits,” Economic Analysis and no. 3, pp. 475–487, 1993.
Policy, vol. 61, pp. 118–135, 2019. [26] I. Panagiotopoulos and G. Dimitrakopoulos, “An empirical
[10] T. Stoiber, I. Schubert, R. Hoerler, and P. Burger, “Will investigation on consumers’ intentions towards autonomous
consumers prefer shared and pooled-use autonomous vehi- driving,” Transportation Research Part C: Emerging Tech-
cles? A stated choice experiment with Swiss households,” nologies, vol. 95, pp. 773–784, 2018.
Transportation Research Part D: Transport and Environment, [27] P. Liu, R. Yang, and Z. Xu, “Public acceptance of fully au-
vol. 71, pp. 265–282, 2019. tomated driving: effects of social trust and risk/benefit per-
[11] A. Shariff, J.-F. Bonnefon, and I. Rahwan, “Psychological ceptions,” Risk Analysis, vol. 39, no. 2, pp. 326–341, 2019.
roadblocks to the adoption of self-driving vehicles,” Nature [28] H.-K. Chen and D.-W. Yan, “Interrelationships between
Human Behaviour, vol. 1, no. 10, pp. 694–696, 2017. influential factors and behavioral intention with regard to
[12] I. Ajzen, “The theory of planned behavior,” Organizational autonomous vehicles,” International journal of sustainable
Behavior and Human Decision Processes, vol. 50, no. 2, transportation, vol. 13, no. 7, pp. 511–527, 2019.
pp. 179–211, 1991. [29] http://www.gov.cn/gzdt/2010-06/04/content_1620735.htm.
[13] F. D. Davis, “Perceived usefulness, perceived ease of use, and [30] X. Sun and S. Xu, “The impact of government subsides on
user acceptance of information technology,” MIS Quarterly, consumer preferences for alternative fuel vehicles,” vol. 39,
vol. 13, no. 3, pp. 319–340, 1989. Journal of Dalian University, (In Chinese), pp. 9–16, 2018.
[14] V. Venkatesh, M. G. Morris, G. B. Davis, and F. D. Davis, [31] P. Bansal, K. M. Kockelman, and A. Singh, “Assessing public
“User acceptance of information technology: toward a unified opinions of and interest in new vehicle technologies: an
view,” MIS Quarterly, vol. 27, no. 3, pp. 425–478, 2003. Austin perspective,” Transportation Research Part C:
[15] V. Venkatesh, J. Y. Thong, and X. Xu, “Consumer acceptance Emerging Technologies, vol. 67, pp. 1–14, 2016.
and use of information technology: extending the unified [32] R. Madigan, T. Louw, M. Wilbrink, A. Schieben, and
theory of acceptance and use of technology,” MIS Quarterly, N. Merat, “What influences the decision to use automated
vol. 36, no. 1, pp. 157–178, 2012. public transport? Using UTAUT to understand public ac-
[16] V. Z. Vanduhe, M. Nat, and H. F. Hasan, “Continuance ceptance of automated road transport systems,” Trans-
intentions to use gamification for training in higher educa- portation Research Part F: Traffic Psychology and Behaviour,
tion: integrating the technology acceptance model (TAM), vol. 50, pp. 55–64, 2017.
social motivation, and task technology fit (TTF),” IEEE Access, [33] S. Kapser and M. Abdelrahman, “Acceptance of autonomous
vol. 8, pp. 21473–21484, 2020. delivery vehicles for last-mile delivery in Germany - extending
[17] S. A. Salloum, A. Qasim Mohammad Alhamad, M. Al-Emran, UTAUT2 with risk perceptions,” Transportation Research
A. Abdel Monem, and K. Shaalan, “Exploring students’ ac- Part C: Emerging Technologies, vol. 111, pp. 210–225, 2020.
ceptance of E-learning through the development of a com- [34] Y. Chen and W. Salmanian, User Acceptance in the Sharing
prehensive technology acceptance model,” IEEE Access, vol. 7, Economy: An Explanatory Study of Transportation Network
pp. 128445–128462, 2019. Companies in China Based on UTAUT2, 2017.
[18] M. M. Rahman, L. Strawderman, M. F. Lesch, W. J. Horrey, [35] P. K. Chopdar, M. D. Lytras, and A. Visvizi, “Exploring factors
K. Babski-Reeves, and T. Garrison, “Modelling driver ac- influencing bicycle-sharing adoption in India: a UTAUT 2
ceptance of driver support systems,” Accident Analysis & based mixed-method approach,” International Journal of
Prevention, vol. 121, pp. 134–147, 2018. Emerging Markets, 2022.
Journal of Advanced Transportation 11

[36] A. a. Abu Gharrah and A. Aljaafreh, “Why students use social


networks for education: extension of UTAUT2,” Journal of
Technology and Science Education, vol. 11, no. 1, pp. 53–66,
2021.
[37] P. Cheng, Z. OuYang, and Y. Liu, “Understanding bike
sharing use over time by employing extended technology
continuance theory,” Transportation Research Part A: Policy
and Practice, vol. 124, pp. 433–443, 2019.
[38] M. Tavakol and R. Dennick, “Making sense of Cronbach’s
alpha,” International Journal of Medical Education, vol. 2,
pp. 53–55, 2011.
[39] R. C. McIlroy, V. H. Nam, B. W. Bunyasi et al., “Exploring the
relationships between pedestrian behaviours and traffic safety
attitudes in six countries,” Transportation Research Part F:
Traffic Psychology and Behaviour, vol. 68, pp. 257–271, 2020.
[40] R. P. Bagozzi, Y. Yi, and L. W. Phillips, “Assessing construct
validity in organizational research,” Administrative Science
Quarterly, vol. 36, no. 3, pp. 421–458, 1991.
[41] J. Hulland, “Use of partial least squares (PLS) in strategic
management research: a review of four recent studies,”
Strategic Management Journal, vol. 20, no. 2, pp. 195–204,
1999.
[42] J. F. Hair, W. C. Black, B. J. Babin, and R. E. Anderson,
Multivariate Data Analysis, Pearson Prentice Hall, Upper
Saddle River, New Jersey, 7th Edition, 2009.
[43] C. Fornell and D. F. Larcker, “Evaluating structural equation
models with unobservable variables and measurement error,”
Journal of Marketing Research, vol. 18, no. 1, pp. 39–50, 1981.
[44] R. B. Kline, Principles and Practice of Structural Equation
Modeling, Guilford publications, New York, US, 2015.
[45] R. A. Acheampong and F. Cugurullo, “Capturing the
behavioural determinants behind the adoption of autono-
mous vehicles: conceptual frameworks and measurement
models to predict public transport, sharing and ownership
trends of self-driving cars,” Transportation Research Part F:
Traffic Psychology and Behaviour, vol. 62, pp. 349–375, 2019.
[46] K. Matzler, B. Renzl, J. Müller, S. Herting, and
T. A. Mooradian, “Personality traits and knowledge sharing,”
Journal of Economic Psychology, vol. 29, no. 3, pp. 301–313,
2008.
[47] J. McNeish, M. Foster, A. Francescucci, and B. West, “The
surprising foil to online education: why students won’t give up
paper textbooks,” Journal for Advancement of Marketing
Education, vol. 20, no. 3, 2012.
[48] S.-Y. Chen, “Green helpfulness or fun? Influences of green
perceived value on the green loyalty of users and non-users of
public bikes,” Transport Policy, vol. 47, pp. 149–159, 2016.

You might also like