Is that what people want? An initial study on the intention to use self-driving taxis in the city of Zurich

1. Theoretical background

Automated vehicles have the potential to fundamentally transform road traffic. As a consequence of the considerable technological progress made in recent years, automated driving has undergone a substantial and rapid development (Hartwich et al., 2018). This development gave rise to the core technology of autonomous vehicles (Garidis et al., 2020), which are capable of fully assuming the tasks and skills of human drivers, independently transporting passengers from point A to point B (Both & Weber, 2013). However, it is not only technological innovation that is critical to the future development of self-driving vehicles. Equally essential are social acceptance, the intention to use, and the integration of this new technology, as these factors are key to ensuring the continued success of technological advancement (Venkatesh et al., 2012). Current public debates are shaped by controversies surrounding safety, utility, and the impact on road traffic (Bazilinskyy et al., 2015).

Nevertheless, it can be assumed that autonomous vehicles will become a common mode of transportation in the future, as automotive companies are actively pursuing their implementation in passenger transportation. Experts also predict that autonomous driving will lead to significant changes in society, the economy, and importantly, in individuals' lives (Garidis et al., 2020).

1.1 Autonomous driving

To differentiate between the various levels of automation, the six classification levels established by the Society of Automotive Engineers (SAE) have become widely recognized. These levels are used to distinguish the capabilities of vehicles, ranging from Level 0 (no driving automation) to Level 5 (full driving automation) (Society of Automotive Engineers International, 2021). At Level 4, vehicles are capable of fully operating without a driver within a defined operational domain. If the vehicle operates exclusively within this domain, it no longer requires a steering wheel or foot pedals. The strict distinction from Level 5 lies in the defined operational domain. At the final stage of automation, vehicles are independently capable of handling all traffic situations, and passengers are simply occupants of the vehicle.

Highly automated Level 4 vehicles, operating driverless in cities, have become a reality. Since 2023, autonomous electric taxis, such as those operated by Waymo, have been deployed without additional safety drivers in cities like San Francisco (Waymo, n.d.).

In Switzerland, among other developments, a conditionally automated Level 3 bus was introduced at the end of April 2023. A pilot route was launched in the canton of Schaffhausen to test the technology. The bus is capable of independently performing all driving tasks, while a safety driver can intervene in unforeseen situations. Its routes must be pre-programmed. The project aims to bridge the gap between individual and public transportation (C. Nägeli, personal communication, November 28, 2023).

In late November 2024, the Canton of Zurich and the Swiss Federal Railways (SBB), coordinated by the Swiss Transit Labs, initiated a pilot project for automated driving in public transport in the Furttal area. The project aims to enhance the transport offerings by deploying self-driving Nissan Ariya vehicles equipped with WeRide¹ technology, particularly in rural and suburban areas. These vehicles are designed not only to complement the existing public transport services but also to expand regional mobility in terms of time and space. The project partners view automation as an opportunity to sustainably shape the future of public transport and to develop cost-effective and flexible mobility solutions. Initially, up to four of these vehicles will be deployed, with plans to expand the fleet by 2026 to include minibuses that will operate either on fixed routes or on-demand, depending on the need (swisstransitlab, n.d.).

In the following, the term autonomous taxis refers to Level-4 autonomous electric taxis, unless otherwise specified.

1.1.1 Advantages of automated driving

The automation of vehicles has the potential to enhance road safety (Hartwich et al., 2018). According to the Federal Statistical Office (BFS), in 2023, a total of 18,254 road traffic accidents with personal injury occurred on Swiss roads, with almost 20% of these caused in the canton of Zurich (BFS, 2024). In the most recent traffic accident statistics provided by the Traffic Department of Zurich, a total of 1,282 accidents involving personal injury occurred in 2023. Approximately 82% of these accidents were attributed to the behavior of the drivers (Dienstabteilung Verkehr, 2024). Deublein (2020) even attributes about 95% of the causes of accidents to human errors, such as fatigue, inattention, or excessive speed. Half of these accidents could already be prevented today if all vehicles were equipped with modern driver assistance systems, reports the Swiss Council for Accident Prevention (BFU, 2020). It implies that automation can reduce human error, thereby making automated vehicles safer than those operated by humans (Jabbari et al., 2022). This is due to the fact that critical traffic situations can be identified with greater precision and reliability. If necessary, vehicles can make corrections autonomously and respond more adequately than humans, not least because they do not get distracted. Consequently, driving comfort would increase, and there would be a relief as well as a reduction in stress for the occupants (Walter et al., 2015).

In addition to the expected positive impact on road safety, other advantages to self-driving vehicles are anticipated. Self-driving vehicles could provide improved accessibility for individuals with mobility impairments. For instance, for those who are excluded from mobility due to age or health reasons (Lemmer, 2019), as well as for individuals who do not possess a driver's license (Hörl et al., 2019). Vehicle automation also presents significant potential for fuel savings. When these vehicles are powered electrically, they can have a more positive environmental impact than conventional modes of transport, partly due to more efficient driving behaviors (Favarò et al., 2017). Additional expected benefits include improved traffic flow and more effective use of infrastructure (Fleischer et al., 2022). Furthermore, riding in a self-driving taxi eliminates the need for parking, which is often a concern in large cities. Therefore, there is an opportunity to alleviate parking shortages (Hörl et al., 2019).

1.1.2 Disadvantages of automated driving

In the course of technological development, also various challenges are to be considered. These include not only legal aspects and economic consequences but also concerns regarding software and information security of autonomous vehicles (Hőgye-Nagy et al., 2023). For instance, respondents in an online survey across 109 countries on public opinion about automated driving conducted by Kyriakidis et al. (2015) were most concerned about software hacking or misuse of automated systems. In this context it is feared that user data could be stolen, or the vehicle could be manipulated.

Another possible obstacle is a lack of trust in this new technology among people. Serious accidents and the reporting on such incidents might considerably influence public acceptance of autonomous vehicles (Othman, 2021). Since this technology has not yet been implemented in everyday traffic, people and their opinions are still heavily swayed by such influences. These perceptions in turn shape the trust in these technologies (Manchon et al., 2023). According to Zmud et al. (2016) the lack of trust and the perception of safety are cited as the main reasons for a depreciating attitude towards self-driving vehicles.

Furthermore, a challenge could be the interaction and communication among all the traffic participants. It will take time before a majority of vehicles in daily road traffic are highly automated. During this period, so-called mixed traffic is expected. Mixed traffic refers to the simultaneous movement of conventional and automated vehicles along with other road users such as pedestrians, public transportation, and cyclists (Deublein, 2020).

1.2 Characteristics of the transportation system in Zurich

Switzerland has experienced continuous growth in resident population. This trend is similarly observed in the canton of Zurich, which is the most populous canton in Switzerland (BFS, 2022).

This increase has led to a steady rise in transportation demand in the city of Zurich (Tiefbau- und Entsorgungsdepartement, n.d.). Data from the Mobility and Transport Microcensus Switzerland reveal that in 2021, residents of the Greater Zurich area traveled an average of approximately 13 km per day by car. Public transport usage averaged around 8 km per day, which exceeded the overall average for Switzerland (Bundesamt für Raumentwicklung [ARE] & BFS, 2023). Zurich operates a highly efficient public transport system by international standards, thereby resulting in elevated user expectations with respect to service reliability (Thibault & Bayen, 2023). According to the annual report of the Zurich Public Transport Services (VBZ, 2023), approximately 297.68 million passengers had been transported in 2023.

According to the Mobility and Transport Microcensus 2021, residents of Zurich used ride services such as taxis or ride-hailing services like Uber for an individual average of about 18 km or 29 minutes annually (ARE & BFS, 2023). Uber itself has a strong presence in the city of Zurich, and its usage is above the national average for Switzerland (BFS, 2023c). Although traditional taxis still operate on the streets of Zurich, their number of employees is decreasing (BFS, 2023a). According to current statistics from 2023, ride-sharing systems are most frequently used by individuals aged 18 to 24 and in urban areas (BFS, 2023b).

Whether the introduction of self-driving taxis on urban streets would lead to an increase or decrease in individual traffic volume remains highly uncertain (Hörl et al., 2019). However, as public space in the city of Zurich is a very limited resource, the shift to smaller modes of transportation, such as self-driving taxis, could further strain this resource (Hackenfort, 2023).

1.3 Intention to use of new technologies

Based on the study by Benleulmi & Ramdani (2022), the intention to use as a predictor variable is employed in recognized theoretical models to anticipate actual use of several technologies. It’s also an important concept (Panagiotopoulos & Dimitrakopoulos, 2018), particularly given that self-driving taxis are not yet accessible to the public in the city of Zurich.

1.3.1 Definition of intention to use

According to Ajzen (1991), the term "intention to use" is used to describe an individual's willingness to engage in a specific behavior. This intention is a critical determinant of whether the behavior in question is ultimately carried out. It encapsulates the motivating factors that shape behavior, providing insight into the level of effort an individual is willing to invest and their inclination to engage in a particular action. The likelihood of a given behavior being carried out depends on the strength of the intention to perform it.

1.3.2 Intention to use explained based on theoretical models

Why individuals use or do not use a new technology has been examined in various theoretical models (Venkatesh et al., 2003). According to Ajzen’s Theory of Planned Behavior (TPB) (1991), behavioral intention is influenced by three key factors: attitudes toward the behavior, subjective norm, and perceived behavioral control. Similarly, Davis’s (1989) original Technology Acceptance Model (TAM) (1989) posits that perceived usefulness and perceived ease of use influence attitude toward using, which in turn influences the behavioral intention. Additionally, perceived usefulness has a direct effect on behavioral intention (Davis, 1989).

The TPB and the TAM are two of eight theoretical models that are subsumed under the Unified Theory of Acceptance and Use of Technology (UTAUT) (2003). It is one of the most comprehensive technology acceptance models, which seeks to explain human behavior in the context of the acceptance and utilization of novel technologies. UTAUT integrates these concepts, suggesting that intention to use is influenced by performance expectancy, effort expectancy, social influence, and facilitating conditions. Additionally, gender, age, experience, and voluntariness of use are identified as moderator variables (Venkatesh et al., 2003). As the UTAUT model concentrated on technology acceptance within a professional and organizational context, the subsequent iteration, UTAUT 2 (Venkatesh et al., 2012), facilitated a more comprehensive understanding of technology acceptance and utilization within the private sector. In addition to the four principal constructs, the UTAUT 2 incorporates three supplementary elements: hedonic motivation, price value, and habit. Furthermore, the moderating effects, with the exception of voluntariness of use, have been incorporated into the UTAUT 2 (Venkatesh et al., 2012).

These theories examine the various constructs that influence behavioral intention and actual use (Korkmaz et al., 2021). The TAM (Davis, 1989), along with its various adaptations, as well as the UTAUT (Venkatesh et al., 2003) and UTAUT 2 model (Venkatesh et al., 2012), has been widely used in recent years to better understand the factors influencing the intention to use self-driving vehicles (Benleulmi & Ramdani, 2022).

1.4 Current status of research / Relevance

Benleulmi and Ramdani (2022) report that there is no consensus on the specific motivations for the factor 'intention to use', noting a broad spectrum of contradictory results. Their study identifies several factors that influence the intention to use autonomous vehicles, including affective factors like trust, instrumental factors like performance expectancy and hedonic motivation, and symbolic factors such as personal innovativeness and social influence (Benleulmi & Ramdani, 2022).

The confidence and safety with which individuals approach new technologies in road traffic and automated vehicles are important factors in shaping their opinions and intentions regarding their use (Bazilinskyy et al., 2015). Jabbari et al. (2022) demonstrated that the perception of safety has a substantial influence on the acceptance of autonomous vehicles. Additionally, a survey conducted by Garidis et al. (2020) on user acceptance of autonomous driving in Germany found that safety was identified as a pivotal factor, either as a facilitator or an obstacle.

Personal innovativeness influences perceptions of usefulness and user-friendliness (Nastjuk et al., 2020). Agarwal and Prasad (1998) noted that individuals with higher innovative behavior are more likely to have favorable attitudes towards technologies and adopt them earlier. Similarly, Benleulmi and Ramdani (2022) found a correlation between an individual's innovativeness and their intention to use automated vehicles.

Social influence refers to the degree to which a person believes that important people around them expect them to use a new system. Observing the use of new technology in social contexts increases the likelihood of adoption (Venkatesh & Davis, 2000). Madigan et al. (2017) found empirical evidence of this factor's relevance in adopting new technologies through a survey of participants in Greece who tested an autonomous minibus. Similarly, a study by Nordhoff et al. (2020) with 9,118 participants across eight European countries confirmed that social influences are crucial for behavioral intention. The actions of others, particularly when technology is new, greatly impact our decisions (Graf-Vlachy et al., 2018).

The impact of sociodemographic variables like gender, age, and educational attainment on the intention to use self-driving vehicles has been extensively studied. Madigan et al. (2017) and Zmud et al. (2016) argue that gender and age are not significant factors, whereas Hulse et al. (2018) and Charness et al. (2018) found that men and younger adults are more favorable towards these technologies. A survey by Schoettle and Sivak (2014) across the USA, the UK, and Australia showed mixed opinions, with safety concerns being prominent, especially among women. Additionally, individuals with higher education levels generally expect autonomous vehicles to reduce accidents and are thus more inclined to use them. These findings align with Othman (2021), who noted that higher education correlates with a more favorable view of autonomous vehicles. However, the scientific consensus on these impacts remains unclear.

It is incontestable that technological development is progressing at an accelerated rate and is consequently gaining increasing attention and relevance (Hörl et al., 2019). Therefore, it is imperative that research in this area be conducted. As the Swiss Federal Council asserts intelligent mobility represents a megatrend (Bundesamt für Strassen, 2018). It is essential that Switzerland adopts a strategic approach to prepare for this transition (Wicki & Bernauer, 2018).

Zurich represents a rare high performance mobility environment in which autonomous services would compete with one of the world’s most efficient public transport systems (Thibault & Bayen, 2023).This allows us to examine acceptance mechanisms under conditions where traditional drivers of adoption, such as poor alternatives or limited accessibility, are largely absent. Moreover, Zurich’s socio-economic profile, characterized by high purchasing power, reduces financial barriers commonly present in other study settings (Statista, 2025). This enables us to isolate acceptance factors that extend beyond classical cost benefit considerations and remain underrepresented in the existing literature. Finally, Zurich’s status as a highly regulated and innovation-oriented test environment provides an internationally relevant benchmark (Ogorodnikova, 2023). Insights derived from such a mature urban mobility system therefore offer clear theoretical and empirical contributions.

1.5 Objective of the study and research question

The objective of this study is to examine the potential use of self-driving taxis in the city of Zurich, including the identification of user groups and the routes they would take. Furthermore, we are interested in whether there are any differences in the intention to use them during the day or at night. The findings will be used to explore and record an initial picture of the disposition in the city of Zurich, which will then be used to draw up initial implications. No hypotheses were formulated for this kind of exploratory research (Häder, 2019).

The following research question is investigated in this study: Under what conditions would self-driving taxis be used in the city of Zurich, and what factors influence the intention to use them?

2. Method

In this quantitative study, we employed a non-experimental cross-sectional design, using an online survey with a multifactorial approach. This method is particularly suitable for surveying a large number of individuals and gaining an initial understanding of the topic (Häder, 2019).

2.1 Operationalization of variables

The items from the self-developed questionnaire were derived from the existing literature on the UTAUT 2 model (Venkatesh et al., 2012). The individual questions were deduced from the factors of performance expectancy, effort expectancy, facilitating conditions, social influence and hedonic motivation. Price value and habit were not included in the survey due to the unavailability of data on these factors, which could not be surveyed because respondents had no prior experience with this new technology.

In her meta-study, Keszey (2020) proposes the inclusion of additional new variables in future research on automated vehicles that extend beyond the conventional models, such as the UTAUT 2. To this end, we have devised a novel questionnaire, augmented with questions drawn from other traffic studies, which should facilitate the capture of any intention to utilize self-driving taxis, particularly in Zurich.

The questionnaire was developed based on the findings of several studies, including the Questionnaire on the Acceptance of Automated Driving (QAAD; Weigl et al., 2021), the research conducted by Nordhoff et al. (2018), and the study by Günthner & Proff (2021). Additionally, questions from the survey conducted by Schoettle and Sivak (2014) were incorporated. Appendix Table A-1 provides a detailed overview of the source of each question, along with a rationale for their selection.

All original studies and questionnaires were translated into German using DeepL (2023) and subsequently discussed with the assistance of a native German speaker with a C2 level of English. To ensure a broad and diverse sample, the questionnaire was designed to minimize technical jargon.

In order to facilitate the comparability of the variables and to enhance the precision of the responses, two distinct types of scale were selected for the survey. The majority of items were rated on a seven-point Likert scale. The scale ranged from 1 (strongly disagree) to 7 (strongly agree) and was adapted from the UTAUT model (Venkatesh et al., 2003). The items pertaining to the topic of benefits and challenges, conducted by Schoettle and Sivak (2014), were measured on a four-point Likert scale, ranging from "very unlikely" (1) to "very likely" (4). The sociodemographic data were recorded using nominal and ordinal scales.

All questions were required to be answered, and no questions could be omitted. However, respondents were permitted to select "don't know" as an answer option for each item. This was done to prevent participants from randomly selecting an option if they did not have an answer to the question. To guarantee that respondents carefully read, understood and completed the online survey, a control variable was incorporated into the questionnaire as an attention control. To ensure consistency in responses, one question was posed twice. The response differential between the control variables was examined for the purposes of data cleaning. Answers with a high discrepancy were excluded from the study. This approach is recommended to identify careless responders (Meade & Craig, 2012).

In the introductory text to the survey, respondents were informed about the data protection and anonymity measures that would be in place. Furthermore, it was indicated that the questionnaire would require approximately 15 to 20 minutes to complete. To proceed, respondents were required to indicate their consent to participate in the study. The questionnaire was introduced by two brief informational videos concerning Level 4 autonomous vehicles. One video was created by the authors and depicts the potential ordering process and the arrival of the self-driving taxi. To provide respondents with the most realistic picture possible, additional scenes from a YouTube video of a real trip through Los Angeles in a Waymo taxi were incorporated into the survey (Electrek.co, 2023). The second video, produced by Cruise, another company that developed self-driving vehicles, showcases sample rides with different individuals (Cruise, 2022).

To ensure the questionnaire's quality and suitability, a two-week pretesting phase was conducted with ten people. In consideration of the feedback, three questions (No. 8, 13, and 48) were excluded.

The detailed reasons for exclusion can be found in the questionnaire in Table A-1 of the appendix. The data collection process was completed in approximately two months and utilized the online survey software Unipark (2019). The mean processing time was 13 minutes.

The questionnaire was divided into thematic areas to facilitate a more coherent and accessible format for the respondents.

2.1.1 Questionnaire component

Table 1 presents the thematic areas, along with sample items and their respective numbers.

To illustrate the potential operational area of self-driving taxis, a map of Zurich was shown in the intention to use questions. The participants were informed that the self-driving taxis are subject to operational limitations within a specified area of Zurich. In addition, in question 19, a series of illustrative routes within Zurich were presented, accompanied by the actual distances and associated fares. These were calculated in accordance with existing fare structures in the USA.

2.2 Sample collection and statistical methods

The objective was to gather a sample that was as diverse as possible, in order to approximate the characteristics of the general population. The recruitment process was conducted in multiple stages and across different locations in Zurich, both online and offline. Various age groups were approached on the street, in different local neighborhoods, and at universities. Additionally, 43 associations and organizations were contacted to reach further participants.

The data were analyzed using IBM SPSS Statistics version 29.0.2.0. An exploratory factor analysis was performed for each questionnaire component. The identified factors are described in the results. Three independent, hierarchical regression analyses were conducted to predict the respective dependent variables (DV): The first DV was the intention to use self-driving taxis if they would be permanently available. This is intended to assess general usage. (Chapter 3.4.1). The differences of the usages in the day or night time were examined with the second and third DV. The second DV was the intention to use them during the day (Chapter 3.4.2), and the third DV was the intention to use them at night (Chapter 3.4.3). The contribution of several predictors to explaining the DV was examined in stages. To ensure that only participants who had already formed an opinion about self-driving cars were included in the analysis, the “Don’t know” response option was treated as a missing value. This resulted in case-wise exclusion in the regression analyses. Consequently, the sample size varied depending on the model.

3. Results

3.1 Statistic data cleaning and processing

The analysis procedure entailed the following steps: (i) the creation of multiple response sets (question No. 19, 20, 23), (ii) the recoding of negatively formulated variables (question No. 3, 16), and (iii) the determination of the discrepancy between the control variables (question No. 5 and 28).

A total of 302 participants completed the questionnaire in its entirety. Due to the examination of the careless responders, 7 participants were excluded from the sample. Individuals exhibiting a discrepancy of greater than three points were immediately excluded from the study. For participants with a differential of two or three points, additional evaluation criteria were applied. The response time of the participants was then analyzed, as well as their overall response patterns. Through this screening process, an additional 23 participants were excluded. As a result, the final sample comprised 272 participants (N = 272). The objective of this cleaning process was to enhance the quality of the data (Ward & Meade, 2023).

3.2 Descriptive statistics

3.2.1 Descriptive sample description

Of the 272 participants, 46.70% (n = 127) were female, and 53.30% (n = 145) were male. The participants ranged in age from 16 to 89 years, with an average age of approximately 44 years (M = 43.90, SD = 15.37). The percentage of participants whose highest level of education corresponded to a university degree was 60.30% (n = 164).

3.2.2 Descriptive presentation of the results

A majority of participants regularly inform themselves about new technological innovations and they look for ways to experiment with them (see Table 2). They also have no difficulty relinquishing control to a machine. Moreover, participants showed a slightly above-average interest in self-driving taxis. When asked whether participants would trust such a vehicle, the mean response was also slightly above the midpoint. However, the question of whether they would feel safe in such a vehicle received a more critical assessment. There was agreement among respondents regarding the question of whether they could envision self-driving taxis as a useful means of transportation. The willingness to try a self-driving taxi during a test drive was very high. The highest level of agreement was observed for the question of whether participants would accept the use of such vehicles in their environment, even if they would not use them personally. And there was also an agreement to use this kind of vehicle, when the social environment uses it.

Regarding the participants’ views on the trip length, 41.18% stated that they would use it for medium distances, defined as between 3 km and 5 km (see Table 3). Similarly, 40.44% could imagine using such a vehicle for longer distances of 7 km or more, whereas only 20.96% would consider it for short distances (under 3 km). Also 35.66%of the participants responded with "I would not use it". Regarding trip duration, the preference for medium travel times, defined as between 15 and 30 minutes, was most common with 49.72% selecting this option over the others. Meanwhile, 29.04% of respondents indicated that they would not use a self-driving taxi. More than half of the respondents indicated that they would replace a traditional taxi (52.21%), followed by Uber (44.12%) and public transport (31.99%). A quarter (25.37%) would replace their private car with a self-driving taxi.

3.3 Factor analysis

To summarize and reduce the number of variables, thematically similar items were averaged, and an exploratory factor analysis was performed for each content construct. The suitability of the variables for a factor analysis was assessed using the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy and the Bartlett test. The KMO values varied between medium (KMO = .66) to good (KMO = .87). The Bartlett test indicated that the variables were correlated (p < .001). Based on the results of the statistical tests mentioned, the variables were considered suitable for the factor analyses. The maximum likelihood extraction and varimax rotation were applied (Costello & Osborne, 2005). Based on the data structure, six factors were identified. Subsequently it was reevaluated whether the items were conceptually coherent and shared a common meaning. The explained variances are presented in Table B-1 to B-6 and the results of the factor analyses in Table B-7 to B-12 in the appendix. The factors are described below:

1. Attitudes toward new technologies. This factor describes an open-mindedness and willingness to experiment with new technologies. The explained variance of this factor was 45.86% and included four items, e.g.: "If I heard about a new technology, I would look for ways to experiment with it."

2. Safety and trust. This factor reports on trust in self-driving taxis and perceived safety. The explained variance of this factor was 53.70%. It comprised eight items, e.g.: "I would trust a self-driving taxi."

3. Usefulness. This factor reflects the perceived ease of use and willingness to use. It explained 62.99% of the variance and contained four items, e.g.: "I believe that self-driving taxis will be very user-friendly and easy to use."

4. Attitudes toward mobility in Zurich. This factor contains the opinions about the transportation system of Zurich. The explained variance was 36.69% and the factor consisted of six items, e.g.: "I believe that self-driving taxis are compatible with all aspects of today’s Zurich transportation system."

5. Environmental and traffic efficiency. This factor includes assessments about self-driving taxis in an environmental and transportation context. The factor had four items and explained 45.62% of the variance, e.g.: "How likely do you consider the benefit of lower vehicle emissions?"

6. Social influences. This factor measures the influence of recommendations and acceptance in the social environment. It explained 53.73% of the variance and consisted of four items, e.g.: "People who are important to me think that I should use a self-driving tax."

3.4 Hierarchical regression

To examine the association between the factors and the three dependent variables (DV), a hierarchical regression analysis was conducted for each variable: starting with the intention to use if the self-driving taxis were permanently available, followed by the intention to use during the day and concluding at night.

An assumption for a regression analysis is normality. This was verified with histograms. To assess potential multicollinearities, Pearson correlations were examined. Values exceeding the threshold of r = .70, as suggested in the literature, indicate strong collinearity (Backhaus et al., 2023). The factor usefulness correlated strongly with safety and trust (r = .72, p < .001) and attitudes toward mobility in Zurich (r = .75, p < .001). Therefore, safety and trust and attitudes toward mobility in Zurich were removed. The remaining factor usefulness provides a reliable result due to the high correlation across all three. Consequently, it was renamed to safety and utility evaluation. Likewise, the variable measuring if respondents find self-driving taxis useful in Zurich correlated strongly with the variable measuring whether they would recommend them to their social environment (r = .70, p < .001). For this reason, the recommendation item was removed.

The hierarchical regression was performed with the four factors environmental and traffic efficiency, safety and utility evaluation, attitude toward new technologies, social influences and the sociodemographic variables gender, age and educational level. The data were all measured on an interval scale, except for gender, which was coded as a binary variable (1 = female, 2 = male) and educational level, which was included with three dummy variables. Overall, the analyses included five hierarchical models. These, along with their arrangement, are shown in Table 4.

3.4.1 Regression model with intention to use if self-driving taxis were permanently available

The independent variables accounted for 56.90% of the variance in intention to use (R²_adj = .57).

According to Cohen (1988), this constitutes a high level of explained variance. The highest change in R² occurred in Model 2 after the addition of the safety and utility evaluation (R²_adj = .37). By adding this factor, an additional 25.70% of the variance was explained. This shows the high influence of this factor. Another relevant change in adjusted R² was observed in Model 4 with the inclusion of the factor social influences (R²_adj= .57). This resulted in an increase of 11.80% of the variance in intention to use. The model summary for Model 1 to 5 is presented in Table 5.

The most important predictors in Model 5 were social influences and attitudes toward new technologies (see Table 6). The strongest factor was social influences (β = .44, t(174) = 6.68, p < .001). The attitudes toward new technologies followed (β = .25, t(174) = 3.97, p < .001). Likewise, safety and utility evaluation is an essential and significant predictor (β = .20, t(174) = 2.86, p = .005). No relation was found between the permanent intention to use and the sociodemographic variables. Model 5 is statistically significant (F(8, 174) = 30.98, p < .001). The ANOVA and the coefficients in Table of Model 1 to 4 can be found in appendix Table C-1 and Table C-2.

3.4.2 Regression model with intention to use during the day

All independent variables together explained 31.10% of the variance in the intention to use during the day (R²_adj = .31). According to Cohen (1988), this represents a strong explained variance. The highest change in R² was observed in Model 2 with the addition of the factor safety and utility evaluation (R²_adj = .22, R² Change = .14). This was followed by the second most important factor environmental and traffic efficiency in Model 1, which also contributed to the explained variance (R²_adj = .08). With a marginal difference, the factor social influences accounted for an additional 7.10% of the variance (R²_adj = .32, R² Change = .07). This result showed that, similar to the permanent intention to use if self-driving taxis were available in Zurich, the evaluation of safety and utility evaluation contributed the most to the explanation. The model summary is provided in Table 7.

The coefficients Model 5 are presented in Table 8. The strongest predictor was social influences (β = .35, t(172) = 4.19, p < .001). The predictive value of attitudes toward new technologies was also significant (β = .19, t(172) = 2.35, p = .020). The remaining factors did not significantly contribute to predicting the intention to use during the day. Overall, Model 5 is statistically significant (F(8, 172) = 11.14, p < .001). The Table of the coefficients Model 1 to 4, as well as the ANOVA, are shown in appendix Table C-3 and Table C-4.

3.4.3 Regression model with intention to use during the night

Together the independent variables explained 41.90% of the variance (R²_adj = .42), which constitutes a strong explanation of variance, according to Cohen (1988). Thus, the intention to use at night was explained nearly 11% more effectively by the factors compared to the intention to use during the day. The most influential factor was safety and utility evaluation (R²_adj = .34, R² Change = .28). This finding indicates that the evaluation of safety and usefulness has the greatest impact on the intention to use a self-driving taxi at night. The rating of environmental and traffic efficiency accounted for 7.10% of the variance and was the second most relevant factor (R²_adj = .07). This was followed by social influences (R²_adj = .42, R² Change = .06). The values of the model summary are presented in Table 9.

In Model 5 (see Table 10) the factor safety and utility evaluation was the strongest predictor (β = .34, t(174) = 4.17, p < .001). The factor social influences similarly showed a significant and high influence (β = .30, t(174) = 3.95, p < .001). Age emerged as a significant negative predictor of the intention to use during the night (β = -.12, t(174) = -1.98, p = .049). Model 5 is statistically significant (F(8, 174) = 17.42, p < .001). The ANOVA can be found in Table C-5 in the appendix.

Likewise, Table C-6 with the results of the coefficients of Model 1 to 4.

4. Discussion

4.1 Interpretation of the results

Although the respondents have not yet had any personal experience with self-driving taxis and seem to be generally very satisfied with the current means of transportation available in Zurich, a high level of interest could be revealed. In general, they keep up with new technologies and seek ways to experiment with them. Moreover, they are comfortable relinquishing control to a machine. The very high level of agreement on testing a self-driving taxi further highlights the participants’ openness to innovative technologies.

Furthermore, the participants generally perceive taxis as very useful. Potential benefits of self-driving vehicles such as optimized traffic flow, improved accessibility to transportation, reduced fuel consumption and increased road safety, seem to be established already (Panagiotopoulos & Dimitrakopoulos, 2018). The findings indicate that the intention to use self-driving taxis increases from medium travel distances onward. Moreover, nearly half of the respondents would choose such a taxi for medium travel times between 15 to 30 minutes. The perception of safety was answered more critically than trust in self-driving taxis. By way of contrast, this finding differs from the survey conducted by Schoettle and Sivak (2014), in which respondents expressed a high level of concern. This may be due to a better understanding among individuals that automation has made road traffic safer. Nowadays cars equipped with automatic parking aids, automatic braking systems or lanekeeping assistants are no longer a rarity. This indicates that individuals have become accustomed to this elevated level of automation in daily traffic and have developed a greater trust in this technology through their experiences.

The regression analysis indicates, in summary, that the intention to use self-driving taxis, if they were permanently available in Zurich can be attributed to the factors safety and utility evaluation, social influences and attitude toward new technologies. For the intention to use them during the day, only social influences and attitude toward new technologies have an impact, whereas during the night, age also has a significant effect. These results align with the statements of Benleulmi and Ramdani (2022), who suggested that affective motives (trust) and symbolic motives (innovativeness and social influence) contribute to the intention to use.

Whether participants would trust a self-driving taxi on Zurich’s roads and their sense of safety are important influences on the intention to use. The perceived safety in road traffic is also a crucial element. This indicates whether the vehicles are capable of responding adequately in unforeseen circumstances and how they interact with other road users. The compatibility of the vehicles within Zurich’s transport network is also an important aspect. A user-friendly operation should be a priority and a key focus.

As also illustrated by Nordhoff et al. (2020) social influence is a fundamental determinant of intention to use. It can be assumed that individuals often align their behavior and attitudes with those of their social environment, particularly when a technology is new and there is a lack of experience with it. As Venkatesh and Davis (2000) have also observed, individuals are more likely to accept and use a technology when they perceive that it is already being used by others in their social circle. The opinions of family members, friends, and colleagues toward self-driving taxis therefore play a decisive role in the evaluation and use of this technology.

Furthermore, attitudes toward new technologies is also an important factor in predicting the intention to use. The higher a person’s interest in new technologies, the more open they are. Therefore, they are more likely to try out and adopt this innovation in Zurich. These findings align with those of Benleulmi and Ramdani (2022), which confirm a relation between an individual’s level of innovativeness and their behavioral intention to use an automated vehicle.

The results demonstrate that neither gender, age, nor educational level constitute an additional explanation for the intention to use self-driving taxis, if the taxis were permanently available in Zurich. These findings are consistent with Madigan et al. (2017) and Zmud et al. (2016). They also assert that personal and psychological factors, such as the perception of safety or attitudes toward the use of technology, are more accurate predictors than sociodemographic characteristics. By way of contrast, the studies by Hulse et al. (2018) and Charness et al. (2018), identified the influence of sociodemographic variables on attitudes toward self-driving vehicles. Due to differing methodological approaches, it should be noted that the results may not be entirely comparable. Nevertheless, the primary conclusion that can be drawn from the regression analysis is that self-driving taxis are of interest to the overall sample, regardless of gender, age or educational level.

There are differences in the results between the intention to use during the day and at night. During the day the factor attitudes toward new technologies influences the intention to use, while at night, it has no effect. Here, instead, safety and utility evaluation emerge as the most important factor. This result leads to the conclusion that at night a sense of safety is of greater importance. During these hours, people tend to be more safety-conscious and exhibit an increased awareness of risks.

Moreover, during the night age becomes an important influencing factor. With increasing age, the likelihood of using a self-driving taxi at night decreases. It can be assumed that younger people participate more actively in Zurich’s nightlife and are out more often during the night. This may also highlight the increased concept of safety during nighttime.

4.2 Implication

In this study an initial questionnaire on the intention to use self-driving taxis in the city of Zurich was conducted. Based on the results, the factors safety and utility evaluation, social influences and attitude toward new technologies emerge as decisive predictors. In contrast, gender, age and level of education do not play a crucial role.

It is important for authorities and manufacturers to provide transparent reports on safety aspects. Safety in self-driving taxis encompasses not only the subjective feeling but also measurable actions that enhance road safety. The participants also value the usefulness and user-friendliness of such vehicles. This emphasizes that, for example, the effectiveness of self-driving taxis should be shown. Furthermore, it means that these vehicles need to be as practical and suitable for everyday use as possible. This includes a simple ordering process and a comfortable ride until the vehicle is left. In fact, using a self-driving taxi should not require any special technical skills.

The survey revealed that younger people have a need for additional means of transportation at night. This finding is less surprising, as public transportation generally operates less often during the nighttime. According to VBZ (2023), the demand during these hours is increasing. Self-driving taxis could provide a solution here and serve as an additional mode of transportation.

As the introduction of self-driving taxis in the USA has shown, they are tried out and used once they are available in the cities. This aligns with our survey results, which revealed that more than half of the participants would like to try out a test ride. Although many respondents are very satisfied with public transport, they consider self-driving taxis an exciting alternative in Zurich’s road traffic. Similarly, Hörl et al. (2019) demonstrated in their study that a fleet of self-driving taxis in the city of Zurich could be a beneficial addition to the current transport network.

4.3 Limitations and future research

4.3.1 Limitations

A large part of the empirical research on automated vehicles, which forms the basis of the UTAUT (Venkatesh et al., 2003) and the UTAUT 2 (Venkatesh et al., 2012) models, focuses mainly on the actual utilization of these technologies (Keszey, 2020). As self-driving taxis are not yet commercially available in Zurich, it is not possible to ascertain their actual use. Accordingly, this study primarily concentrates on the intention to use as an outcome variable. For this reason, a new questionnaire was designed that includes questions from UTAUT and UTAUT 2, but was also supplemented with questions from other transportation studies.

Given that these vehicles are not yet accessible to the public in Zurich, our findings reflect the participants' perceptions. According to Nordhoff et al. (2018) this can lead to biases in the responses. To counteract this effect, two informational videos about real existing self-driving taxi technology were provided at the beginning of the questionnaire. However, the participants’ current knowledge about this technology was not assessed, which can be considered a limitation. As no pricing models for self-driving taxis currently exist in Zurich, the fare examples used in the questionnaire were based on U.S. price structures. This may limit the relevance of the findings to the Swiss context. There were also constraints regarding the response scales. Using the seven-point Likert scale, respondents had the option to give a neutral answer. According to Simms et al. (2019), this is often used when individuals are uncertain about their response. Moreover, the response scale was unified. However, since different questions were adopted from various studies, the validity and reliability of the measurement results require closer examination. Adjustments to scales can disrupt their psychometric properties (Simms et al., 2019).

As the recruitment was conducted partly on social media, there is a possibility that the sample is not representative of Zurich. For example, the sample was homogeneous in terms of education. People with a lower level of education were under-represented, as 60% of the participants had a university degree. In addition, the sample included only a few elderly individuals or those with disabilities. This may be due to the survey being conducted online and this target group tends to be less familiar with technology. On the other hand, prior informational events would have been necessary in retirement homes and residences to introduce the topic effectively. Consequently, the findings should be interpreted as indicative of attitudes among an urban, relatively well-educated population in Zurich.

4.3.2 Future research

We consider older people an interesting target group for future research. Self-driving taxis could provide them with better accessibility in urban traffic. Therefore, this group should be given greater attention in future studies and explored qualitative as well. We also recommend the inclusion of individuals with disabilities. This is because autonomous vehicles could ensure access to the transportation system, which has been highlighted as a positive aspect in many studies. According to the literature review by Othman (2021), only a few studies have addressed whether this group would accept it at all.

Although the variance explained by our hierarchical analyses is substantial, it remains evident that additional factors influence the intention to use self-driving taxis. For example, the perceived security and usability of the entire process may play an important role. Furthermore, the cost considerations and the availability of the taxis may also be a key. Future research should focus on other models and factors that may affect the intention to use autonomous taxis.

While the general acceptance of automated vehicles has been extensively researched, there are only a few studies that specifically address the intention to use self-driving taxis.

4.4 Conclusion

The objective of this study was to capture an initial impression of attitudes toward self-driving taxis in Zurich. Overall, it can be concluded that openness and curiosity about this technology exists. Most respondents are willing to take a test ride. If taxis were available, they would primarily be used for medium and longer distances and for medium travel times. Most individuals would currently replace a normal taxi or Uber with a self-driving taxi. During the night the willingness to use increased, especially among younger people and the factor safety and utility evaluation is more relevant.

An essential element in the success or failure of autonomous transportation is societal acceptance. The adoption of a concept like those practiced in the USA or in other countries will depend on how safe people perceive self-driving taxis, their opinions about their social environment, and their attitude towards new technologies. An important finding from this study, however, is that gender, age and educational level had no influence on the intention to use self-driving taxis.

The mobility sector is undergoing extensive developments, and the city of Zurich is no exception. The impact on urban life and the environment is expected to be remarkable, whether in terms of shifts in societal mobility behavior or individual transportation (Hackenfort, 2023). The implementation of autonomous vehicles requires seamless integration across all platforms, including public transport such as trams and buses, rail services like Uber or regular taxis, and micromobility solutions such as e-scooters or bicycles. Self-driving taxis could be perceived as either an optimal

complement to or a competitor of the existing offerings, depending on their way of integration in the mobility system of the respective cities.

Acknowledgment

This article is based on our bachelor’s thesis entitled “Test Drive into an Autonomous Future? A First Investigation of the Intention to Use Self-Driving Taxis in the City of Zurich”, completed within the Applied Psychology program at the Zurich University of Applied Sciences (ZHAW). We would like to express our sincere thanks to Prof. Dr. Markus Hackenfort for his valuable guidance during the thesis process and for his collaboration in the preparation of this publication.

We also thank the Avenir Group for acknowledging the original work.

Special thanks go to the City of Zurich, and in particular to Dr. Wernher Brucks, for providing constructive feedback on our survey. Lastly, we are grateful to all individuals who participated in the study and contributed their time and honest responses.

CRediT contribution

Lea Häberli: Conceptualization, Formal Analysis, Investigation, Methodology, Writing—Original Draft. Sabrina Hofer: Conceptualization, Formal Analysis, Investigation, Methodology, Writing—Original Draft. Markus Hackenfort: Conceptualization, Formal Analysis, Investigation, Methodology, Supervision, Writing—review & editing.

Data availability

The data are available on request to the authors.

Declaration of competing interests

The authors report no competing interests.

Declaration of generative AI use in writing

During the preparation of this work the authors used:

DeepL (2023). DeepL Translator (Version DeepL Pro): https://www.deepl.com/translator

• Translation of studies
• Translation of text passages

Neuroflash (2023). Neuroflash (Free Version): https://neuroflash.com/de/

• Image creation

OpenAI. (2024). ChatGPT (Version 4.0): https://chatgpt.com

• Source of inspiration
• For summarizing literature
• To assist in understanding content
• For improving phrasing and language
• Translation of text passages

The output was reviewed and revised by the authors who take full responsibility for the content of the publication.

Ethics statement

At the time of its preparation, the study did not fall within the legal framework of the Human Research Act.

Funding

Open access funding provided by ZHAW Zurich University of Applied Sciences.

Editorial information

Handling editor: Sonja Forward, The Swedish National Road and Transport Research Institute (VTI), Sweden.

Reviewer: Ioni Lewis, Queensland University of Technology, Australia.

Submitted: 11 August 2025; Accepted: 9 February 2026; Published: 20 February 2026.

¹WeRide is a global company specializing in autonomous driving technology, based in China.