Understanding safety perceptions, trust, and acceptance of autonomous shuttles: findings from a UK pre-trial survey

1 Introduction

Autonomous shuttles (AS) are increasingly framed within urban mobility policies as potential tools to enhance first-/last-mile connectivity, accessibility, and transport sustainability (Nordhoff, et al., 2021; Diba, et al., 2025; Hafiz, et al., 2025; Taamneh and Makahleh, 2025) particularly within closed or semi-closed settings (e.g., campuses, exhibition centres, pedestrianised zones, and low-speed corridors) (Nordhoff, et al., 2021; Makahleh, et al., 2025). However, for autonomous mobility services to be successful in real-world settings, three criteria should be met: 1) the technology must be able to perform well and safely, 2) policy and regulatory frameworks must permit deployment and operation, and 3) members of the public must be willing to use the service. While the former two criteria have attracted increasing attention in both academic and policy circles, the role of public acceptance, awareness, and perceptions is often more difficult to address, particularly at the early stage of AS deployment.

A significant body of behavioural transport literature has established that acceptance of automated mobility is strongly influenced by psychological factors, including perceived safety, trust in the automation, comfort, and perceived risk (Paddeu, et al., 2020; Pascale, et al., 2021; Nordhoff, et al., 2022; Molina, et al., 2021; Kenesei, et al., 2025; Pascale, et al., 2021). Negative expectations of system unreliability, unpredictability, and unfamiliarity, for example, are likely to significantly outweigh expected benefits such as improved accessibility, reliability, or environmental performance in contexts in which they have not been directly experienced (Nordhoff, et al., 2018; Celina, et al., 2021). Qualitative and mixed-method evidence further suggests that comfort and trust may be positively influenced by experience and exposure (e.g., after riding or observing rides) (Nordhoff, et al., 2021; Makahleh, et al., 2024a; Makahleh, et al., 2026), while safety concerns are often not mitigated even after exposure. As such, it is important to understand attitudes and perceptions both after individuals have had an opportunity to trial the technology, but also at the earlier stage when individuals first become aware of AS services in their locality and are asked to decide whether to engage with them at all. In this context, it is also important to consider that pre-use perceptions may be influenced by prior exposure to information, communication materials, or informal awareness of planned deployments, even in the absence of direct experience.

In this study, acceptance is conceptualised as a multidimensional construct reflecting individuals’ overall evaluation of autonomous shuttle systems, encompassing both psychological perceptions (e.g., trust, perceived safety, and risk) and behavioural intention. This conceptualisation is consistent with established frameworks such as the Technology Acceptance Model (TAM), the Theory of Planned Behaviour (TPB), the Unified Theory of Acceptance and Use of Technology (UTAUT) (Venkatesh, et al., 2003), and more recent models specific to automated mobility such as the Model for Acceptance of Automated Vehicles (MAVA) (Nordhoff, et al., 2019; Aasvik, et al., 2025), which collectively conceptualise acceptance as being shaped by perceptual factors and expressed through intention to use and actual behaviour. Building on this, the present study adopts a structured conceptualisation in which acceptance is understood to emerge from underlying perceptual evaluations and to be expressed through behavioural intention. In this context, willingness to ride (WTR) is operationalised as a context-specific measure of behavioural intention, capturing the extent to which individuals are prepared to use an autonomous shuttle under different operational conditions prior to direct experience. This conceptual distinction is particularly important in pre-trial settings, where actual usage behaviour cannot be observed and behavioural intention serves as the primary indicator of anticipated engagement with the service at the pre-adoption stage, rather than realised adoption behaviour. Recent studies on automated shuttle buses and autonomous public transport similarly conceptualise intention to use as a key outcome of acceptance-related processes (e.g. (Aasvik, et al., 2024; Aasvik, et al., 2025; Marsja, et al., 2025; Mourtzouchou, et al., 2025; Prasetio, et al., 2025)), reinforcing the relevance of WTR as an empirically grounded proxy for anticipated engagement with autonomous shuttle services in the early stages of the adoption process. Accordingly, the analytical framework of this study assumes that perceptions of safety, trust, and service expectations influence behavioural intention (WTR), which in turn reflects overall acceptance in a pre-use context. In this sense, willingness to ride (WTR) is interpreted as a measure of anticipated or stated behavioural intention, which provides an indication of potential engagement with the service but does not constitute adoption behaviour itself. Its relationship to realised behaviour is therefore indirect and mediated by subsequent experience, contextual conditions, and system performance during actual deployment.

As the majority of members of the public have yet to experience riding in an autonomous vehicle, a priori perceptions formed prior to direct exposure play an important role in shaping early-stage engagement decisions, such as whether individuals choose to seek further information, participate in trials, or consider using the service when first introduced. However, consistent with the Intention–Behaviour Gap, such pre-use perceptions and stated behavioural intentions do not necessarily translate directly into realised behaviour, particularly in contexts where experience, situational constraints, and system performance may subsequently influence decision-making. Rather than being treated as direct predictors of behaviour, pre-exposure perceptions are conceptualised in this study as antecedent conditions that shape initial willingness to engage with autonomous shuttle services and frame how subsequent experiences are interpreted and evaluated. This distinction is particularly important in early-stage pilot contexts, where participation itself represents a critical behavioural step, and where understanding the drivers of initial engagement provides actionable insight for trial design, communication strategies, and policy interventions.

These pre-use perceptions shape whether individuals elect to participate in pilot trials, visit information stands or attend engagement events, and are also likely to shape how later experiences with the technology are interpreted, framed, and subsequently shared with other members of the public (Celina, et al., 2021; Launonen, et al., 2021; Fonzone, et al., 2025; Makahleh, et al., 2026a). As such, gaining a baseline understanding of these anticipatory attitudes is important for trial organisers and policymakers in terms of designing, communicating, and governing AS services in a manner that is proactive in addressing public concerns, rather than reactive to them.

This study forms part of a broader evaluation programme of autonomous shuttle deployment at the National Exhibition Centre (NEC) in Birmingham, United Kingdom, which included two distinct survey phases. The present paper reports findings from the pre-trial survey conducted between 1 August and 6 September 2024, prior to any direct public exposure to the autonomous shuttle service. A separate companion study reports findings from a trial-phase survey conducted during live shuttle operations, focusing on user experience, safety perceptions, and adoption intentions following exposure to the service (Makahleh et al., 2026b). The two studies are analytically complementary, capturing public perceptions at different stages of engagement with the technology, but they are based on independent cross-sectional samples and were conducted as separate data collection exercises, with no respondent overlap. Together, they provide a more comprehensive understanding of how perceptions of autonomous shuttles evolve from pre-use expectations to post-experience evaluation.

To date, while a growing number of AS trials have taken place internationally (Kenesei, et al., 2025), few have sought to empirically measure public perceptions prior to exposure in the United Kingdom. Studies have primarily focused on hypothetical acceptance measures or post-use evaluations following trial rides or demonstrations (Bansal, et al., 2016; Fonzone, et al., 2024; Paddeu, et al., 2020; Molina, et al., 2021; Nordhoff, et al., 2021; Makahleh, 2026; Makahleh, 2024). While important, these provide limited insight into the psychological processes that may be at play in influencing willingness to engage with automated services in the earliest stages of deployment (Makahleh and Makahleh, 2024; Makahleh, et al., 2024b; Makahleh et al., 2026a). In particular, less is known about how individual perceptions of safety, trust, and service-level expectations interact to shape a willingness to ride in an AS under different levels of operational scenarios before any experience has taken place.

To address this gap, the current study draws on a large-scale pre-trial survey conducted at the National Exhibition Centre (NEC) in Birmingham, United Kingdom, to examine pre-use attitudes and willingness to ride an autonomous shuttle under a range of operational scenarios with differing levels of human supervision. By focusing specifically on perceptions formed prior to direct exposure, this study provides baseline evidence on anticipatory perceptions and behavioural intention, which represent the first stage of the adoption process and complement post-experience evaluations of actual use.

This study’s key contributions to the literature and practice of AS trials in the United Kingdom are as follows:

• It provides large-scale empirical evidence on public attitudes and willingness to ride autonomous shuttles prior to direct exposure, based on a real-world trial setting at the NEC Hub in Birmingham.
• It advances understanding of how perceptions of safety, trust, and service reliability shape pre-use acceptance of autonomous shuttles under different levels of human supervision.
• It offers policy-relevant insights to support the design, communication, and governance of early-stage autonomous shuttle deployments, with implications for trial organisers and transport authorities.

The analysis is guided by the following key research questions (RQs):

• RQ1: What underlying patterns of association can be identified among attitudinal variables related to perceptions of autonomous shuttles in a United Kingdom pre-trial context?
• RQ2: How are these attitudinal patterns, particularly those related to safety perception, trust, and service expectations, associated with behavioural intention (operationalised as willingness to ride, WTR) in a pre-exposure context?
• RQ3: What implications do these pre-trial perceptual patterns have for trial operators and policymakers when designing autonomous shuttle demonstration and public engagement strategies?

2 Methods

2.1 Shuttle and route

The pre-trial survey was conducted in preparation for a trial of AS around the NEC campus in Birmingham, United Kingdom. The National Exhibition Centre (NEC) in Birmingham is a major metropolitan leisure and events campus, hosting over 750 events annually, attracting more than 7 million visitors each year, and supporting approximately 3,800 employees across five venues, making it one of the largest multi-purpose destinations in the United Kingdom (NEC Group, 2025). The NEC was selected as a trial site due to the mix and size of its heterogeneous user group and its strategic location as a transport interchange between national rail, motorway access and local employment sites. The proposed shuttle route would have linked Birmingham International Rail Station with the NEC complex, Resorts World, and Birmingham Business Park to create a circulatory service covering a number of high-demand nodes. A 700 m pilot route on the NEC campus formed the first phase of a planned multi-phase shuttle project. The vehicles were zero-emission electric shuttles with a capacity of approximately 10–12 passengers and equipped with automated driving sensors, including lidar, cameras, and radar (see Figure 1). Although the trial design allowed for progressive automation, a human steward was present onboard during operations to support safety management and customer assistance.

2.2 Respondent recruitment and procedure

Questionnaire data was collected in three time periods (1–4 August, 10–11 August and 12 August–6 September 2024). Visitors were recruited purposively from the wider population attending exhibitions, events or leisure facilities at the NEC, and staff from those working at the NEC, Resorts World, the surrounding hotels and Birmingham Business Park. Sampling was convenience-based, with recruitment conducted across different days, times, and locations within the NEC site to capture a broad cross-section of visitors and staff. Also, an attempt was made to capture a range of respondents from across the site by recruiting on different days, at different times of day, and at different locations on the NEC site.

The questionnaire was accessed through a QR code displayed on flyers, posters and banners across the site, and handed out directly by Transport for West Midlands (TfWM) and NEC staff. At the time of data collection, the autonomous shuttle service had not yet commenced live operations, and participants had no direct experience of riding the system. Public communication about the upcoming trial was limited to general awareness and engagement activities conducted on-site, including informational posters, banners, and staff-led interactions. These materials provided high-level information about the planned service, including its purpose, route, and the presence of an onboard steward for safety and customer support during early deployment phases. However, detailed operational experience, real-time performance, and direct exposure to the technology were not available to participants at this stage. As such, responses captured in this study reflect anticipatory perceptions shaped by limited prior information rather than experiential knowledge. Respondents accessed the online survey on their own devices, or if required, on a tablet provided by the research team. Participants were provided with information about the study, gave informed consent, and were assured anonymity before completing the questionnaire. An incentive in the form of a prize draw for two £50 Love2Shop vouchers was offered to participants. Overall, 922 valid responses were obtained (staff = 320, visitors = 602). This pre-trial survey was conducted as a distinct data collection phase prior to the live autonomous shuttle trial and is separate from the later trial-phase survey reported elsewhere; the two surveys did not involve overlapping respondents. Responses with missing data or with duplicate entries (confirmed by checking email/time stamps) were excluded from the dataset. This sample size exceeds recommended limits for multivariate analysis, giving sufficient observations for PCA and regression modelling (Hair and Sabol, 2025; Hair, et al., 1995).

2.3 Questionnaire content

The questionnaire was organised into six thematic groupings (Table 1). These captured participant background characteristics, current travel behaviour within and around the NEC site, psychological attitudes towards autonomous shuttles, willingness to ride under different operational configurations, expectations regarding service and operational performance, and perceived barriers and benefits of use. A short staff-only module captured commuting patterns and on-site working arrangements. The full questionnaire is provided in Supplementary Material S1. This structure was designed to provide contextual, behavioural, and psychological information relevant to pre-use acceptance, while remaining concise and suitable for completion in a busy real-world trial environment. For clarity, each thematic grouping is described below in terms of item content, number of items, and response format.

TABLE 1. Overview of questionnaire structure and content.

Psychological attitudes and service expectations (Q10–Q15; Q20–Q26). The design of the questionnaire was informed by established behavioural frameworks commonly used in transport and technology adoption research, particularly the Technology Acceptance Model (TAM) and the Theory of Planned Behaviour (TPB). These frameworks informed the conceptual selection and operationalisation of key constructs, and behavioural intention in shaping acceptance of emerging mobility technologies. In this study, these constructs were operationalised through five-point Likert-scale items capturing psychological attitudes (Q10–Q15; six items) and service expectations (Q20–Q26; seven items), which were subsequently analysed using Principal Component Analysis (PCA) to identify underlying latent dimensions of pre-use acceptance. In addition, the structure of the questionnaire aligns with broader constructs identified in UTAUT and MAVA, particularly those relating to perceived safety, trust, and behavioural intention in automated mobility contexts. These Likert-scale items were designed to capture latent perceptual constructs and were subsequently analysed using Principal Component Analysis (PCA).

The attitudinal items used to capture psychological constructs such as trust, perceived safety, interest, and anxiety were not drawn from a single validated instrument, but were developed through a theory-informed and literature-based approach. Specifically, items were adapted from prior empirical studies on automated vehicle acceptance (e.g., Nordhoff et al., 2018; Paddeu et al., 2020; Molina et al., 2021) and aligned with constructs commonly operationalised within the Technology Acceptance Model and Theory of Planned Behaviour frameworks. Where existing measures were not directly transferable to the autonomous shuttle context, items were developed and contextualised to reflect the characteristics of low-speed, shared, and supervised autonomous mobility services in a real-world trial environment. As such, the questionnaire represents a context-specific adaptation of established constructs rather than the direct application of a standardised psychometric scale. This approach is consistent with prior transport studies examining emerging technologies in applied settings, where context-sensitive measurement is often required. This approach is consistent with prior transport research where context-specific adaptations of established constructs are required for emerging technologies.

Willingness to ride (Q16–Q19). For the willingness-to-ride scenarios (Q16–Q19), participants were asked to indicate whether they would use the autonomous shuttle under four distinct operational configurations representing varying levels of human supervision: (1) a driver and onboard steward present, (2) a driver only, (3) a steward only (i.e., no driver, but with onboard staff supervision), and (4) a fully autonomous configuration with no onboard human staff. Responses to these four items were recorded as binary variables (Yes/No). These scenarios were designed to reflect a progression from conventional supervised operation to full automation, allowing assessment of how acceptance varies with decreasing levels of human oversight. In line with the study’s conceptual framework, responses to these items are treated as a behavioural proxy for intention to use, and therefore represent the observable component of acceptance in a pre-use context. The scenarios were presented together within the same section of the questionnaire to enable direct comparison between configurations. Participants were provided with general contextual information about the planned service, including the presence of an onboard steward to support safety and customer assistance during early deployment. However, the specific operational roles of “driver” and “steward” were not formally defined within each question item. As such, responses may reflect individual interpretations of these roles, as well as potential framing or comparative effects arising from the presentation of the scenarios. The remaining sections of the questionnaire included participant profile (Q1–Q9, Q27), travel behaviour (Q28–Q41), and barriers and benefits (Q42–Q46), which were measured using a combination of categorical, ordinal, and open-ended response formats, as summarised in Table 1.

2.4 Analysis of responses

Survey data were transferred into Microsoft Excel and IBM SPSS Statistics v29 for processing and analysis. Items were screened for completeness, missing values and duplication. Incomplete questionnaires were removed from the data set. Descriptive statistics such as frequencies, means, standard deviations, and 95% confidence intervals were calculated to summarise the distribution of all items. Principal Component Analysis (PCA) was applied to the attitudinal Likert-scale items (Q10–Q15 and Q20–Q26) to identify underlying patterns of association among attitudinal variables underlying pre-use perceptions of autonomous shuttles. PCA is widely used in behavioural transport research to reduce correlated attitudinal indicators into a smaller number of interpretable constructs related to safety, trust, and service expectations [e.g., (Nordhoff, et al., 2018; Celina, et al., 2021; Fonzone, et al., 2024)]. In this study, PCA is applied as an exploratory technique to identify underlying perceptual dimensions rather than to estimate a confirmatory latent variable model. This approach is appropriate given the absence of a pre-validated measurement model for autonomous shuttle acceptance in pre-trial contexts and the study’s focus on identifying emergent attitudinal structures. While latent variable modelling approaches such as Structural Equation Modelling (SEM) are commonly used to examine relationships between constructs, the present study adopts an exploratory approach. This is appropriate given the absence of a pre-validated measurement model for autonomous shuttle acceptance in pre-trial contexts. PCA is therefore used to identify emergent perceptual structures rather than to test a predefined causal model.

Demographic, behavioural, and open-ended items were excluded as they are nominal, ordinal, or qualitative and therefore unsuitable for component extraction. Sampling adequacy was assessed using the Kaiser–Meyer–Olkin (KMO) measure and Bartlett’s Test of Sphericity to determine whether the attitudinal items shared sufficient common variance and inter-item correlation to justify the application of Principal Component Analysis. All items were standardised prior to extraction. Components were retained based on eigenvalues greater than one, cumulative variance explained, inspection of the scree plot, and interpretability of the factor structure. A Promax rotation was applied to allow for correlation between components, consistent with expectations for related psychological constructs (Hendrickson and White, 1964). Items with loadings ≥0.40 were retained. Willingness-to-ride responses (Q16–Q19) were analysed using Cochran’s Q test to assess whether acceptance differed across the four operating scenarios, as responses were binary (yes/no) and provided repeatedly by the same respondents; where significant effects were identified, post hoc McNemar tests with Bonferroni adjustment were used to compare adjacent scenarios. Component scores were subsequently used in correlation analysis to examine associations with demographic and contextual variables. These component scores were then interpreted in relation to willingness to ride (WTR), which is operationalised as a proxy for behavioural intention, thereby enabling examination of how attitudinal structures relate to intended use in a pre-exposure context. Pearson correlations were applied where variables were continuous or ordinal, consistent with established practice in behavioural and acceptance studies of automated transport systems (Nordhoff, et al., 2021; Molina, et al., 2021).

3 Results

This section is structured to distinguish between contextual descriptive findings and the key analytical results addressing the research questions. Sections 3.1, 3.4, 3.6 provide background on participant characteristics and travel behaviour, while Sections 3.2, 3.3 present the core analyses addressing RQ2 (behavioural intention) and RQ1 (latent perceptual constructs), respectively.

3.1 Participant profile and background (Q1–Q9, Q27)

The sample consisted of 922 respondents: 602 visitors (65%) and 320 staff (35%). Age (Q1) differed between groups, with visitors spread across all age bands (largest: 45–54 and 65–74, each 20%) and staff concentrated in younger groups (25–34: 28%; 35–44: 23%). Gender distribution differed meaningfully between groups, with women comprising 60% of visitors compared to 40% of staff, representing a 20-percentage-point difference between the two samples. Most respondents identified as White (Q3; visitors 81%, staff 75%). Employment status (Q4) reflected expected differences: 80% of staff were in full-time work compared with 42% of visitors, whose responses also included retirement, part-time work, and self-employment. Education levels varied across both groups, with 26% of respondents reporting a Bachelor’s degree. Disability (Q6) rates were comparable (staff 8%, visitors 9%). Financial circumstances (Q7) varied more among visitors, while staff most often reported that bills were affordable or manageable. Environmental consideration (Q8) was similar across groups, with “some of the time” being the most common response (staff 41%, visitors 40%). Prior AV experience (Q9) was low: 80% had never used an AV, and under 2% had used one often. Technology adoption orientation (Q27), shown in Figure 2, also differed. Staff were more open to early adoption compared to visitors. These descriptive characteristics provide important contextual background for interpreting the subsequent analyses of perceptions and behavioural intention but are not directly used to address the research questions.

3.2 Behavioural intention (WTR) across operational scenarios (Q16–Q19)

This section directly addresses RQ2 by examining behavioural intention, operationalised as willingness to ride (WTR), under different operational configurations. In line with the study’s conceptual framework, WTR is interpreted as an indicator of pre-use acceptance rather than a direct measure of realised adoption. Behavioural intention to use the autonomous shuttle, operationalised through willingness to ride (WTR), was assessed across four operating configurations with differing levels of human involvement: driver + steward (Q16), driver only (Q17), steward only (Q18) and no human staff (Q19). Distribution of responses for staff and visitors is summarised in Table 2. Visitor willingness was extremely high in the presence of a driver and steward (95%, Q16) and high for the driver-only scenario (87%, Q17). Willingness to ride for the steward-only scenario was substantially lower (65%, Q18) and was sharply reduced in the fully driverless configuration (31%, Q19). This trend is also depicted in Figure 3, which highlights a marked decrease in acceptance as human supervision is removed. Staff respondents exhibited the same directional pattern as visitors, with willingness to ride decreasing as levels of human supervision were reduced. However, acceptance of the fully driverless scenario (Q19) was higher among staff (40%) compared to visitors (31%), suggesting a greater tolerance for full automation among staff. Willingness to ride the AS was expressed by 95% of staff for the driver + steward (Q16) and driver only (Q17) scenarios, 63% for the steward only (Q18) and 40% for no human staff (Q19). This suggests that, while both staff and visitors are cautious about full automation, the former are slightly more amenable to higher levels of automation, perhaps as a result of greater familiarity with the NEC. The higher level of acceptance observed for the driver + steward configuration compared to the driver-only scenario may reflect a perceived additive effect of human presence, whereby additional onboard staff are interpreted as enhancing safety, supervision, or reassurance. Given that the scenarios were presented together, participants may also have evaluated them comparatively, rather than independently, which could have influenced responses.

TABLE 2. Willingness to ride the autonomous shuttle under four conditions (Q16–Q19).

A Cochran’s Q test was run on the full sample (responses to Q16–Q19 were coded as Yes/No). Results revealed a highly significant effect of scenario on willingness (Cochran’s Q = 1,155.2, df = 3, p < 0.001), indicating that willingness to ride the AS differs systematically across the four scenarios. Post-hoc McNemar tests (Bonferroni-adjusted) showed that each adjacent scenario differed significantly from the next (Q16 > Q17 > Q18 > Q19) and that both removing the driver and removing all staff significantly decreased acceptance. The shift from a driver-plus-steward configuration to a driver-only configuration produced a statistically reliable decline in willingness (p < 0.001). A much steeper decline occurred when moving from a driver to a steward-only scenario (p < 0.001), consistent with the substantial drop from 87% to 65%. The largest reduction was observed when all staff were removed, with willingness falling from 65% to 31% (p < 0.001). The sharp decline between Q18 and Q19 likely reflects persistent discomfort with full automation on the part of those with little or no prior experience of AV technology. This is broadly in line with wider evidence on AV acceptance, which has found the presence of supervision, trust and perceived control to be important during early deployment (Yang and Kim, 2025; Kenesei, et al., 2025). These findings provide direct empirical evidence of how acceptance varies with operational conditions and form the behavioural outcome against which perceptual constructs (RQ1) are interpreted. These results therefore reflect variation in behavioural intention across operational conditions, which represents the observable expression of acceptance in a pre-use context, rather than acceptance as a fully realised behavioural outcome.

3.3 PCA findings: attitudinal components (Q10–Q15, Q20–Q26)

This section addresses RQ1 by identifying underlying patterns of association among attitudinal variables. These components summarise clusters of related attitudinal items reflecting how perceptions are structured in the dataset of safety, trust, and service expectations are structured in a pre-trial context. PCA was performed on the attitudinal items (Q10–Q15) and service-importance items (Q20–Q26) to identify the main psychological dimensions that influence people’s perception of autonomous shuttles. All the five-point variables were recoded so that the higher the value, the more positive the attitude, including a reverse-scoring of Q15. The Kaiser–Meyer–Olkin measure of sampling adequacy was 0.86, and Bartlett’s Test of Sphericity was significant (p < 0.001), confirming the correlation structure was adequate for PCA.

For clarity, psychological acceptance is defined in this study as the affective and cognitive evaluation of autonomous shuttle technology, including trust, perceived safety, interest, and anxiety. In contrast, service and operational expectations refer to users’ anticipated evaluation of service performance, including reliability, frequency, pricing, usability, and safety-related service features.

In line with our analytical framework, a two-component solution was retained, explaining 62.4% of the total variance. Inspection of the scree plot (Figure 4) indicated a clear elbow after the second component, supporting the parsimony and interpretability of the two-factor solution.

Component 1 captures a cluster of items related to affective responses, trust, and perceived safety toward autonomous shuttles, including high loadings on the interest, excitement, trust, and safety items (reverse-scored for fear, 0.72–0.81). The second component is a factor of service and operational expectations, with high factor loadings for the pricing, service frequency, speed, booking procedure, on-board safety, technology reliability, and emergency information items (0.58–0.77).

In Table 3, the correlations between the two PCA components and key demographic variables are reported. The PCA components have small-to-moderate positive correlations with one another, which is not unusual for a two-component PCA, but it also means that emotional and expectational evaluations are partially related, which suggests that these attitudinal groupings can be interpreted as distinct but related patterns within the data (full correlation matrix in Supplementary Material S2).

TABLE 3. Pearson correlation matrix.

p < 0.05 for |r| ≥ 0.12; p < 0.01 for |r| ≥ 0.16.

More generally, the patterns are consistent with the broader behavioural evidence about the multidimensional nature of people’s attitudes to autonomous vehicles. PCA suggests that responses to AS cluster around two broad attitudinal groupings. The strongest and most coherent underlying dimension is emotional trust and safety, followed by a factor of service- and operational-related expectations.

The two PCA components provide an empirically grounded summary of attitudinal patterns for explaining the behavioural intentions and underpin the patterns observed in Section 3.2, where we saw a drop in acceptance as human oversight was removed. This result also helps to address the limitations of earlier analyses by clarifying the reasons for different reactions under different operational scenarios, and provides a firmer theoretical basis for the later modelling and discussion.

Correlations between demographic variables and the PCA components were generally weak. While a small positive association was observed between gender and psychological acceptance (PCA1), the magnitude of this relationship was low and not considered substantively meaningful for interpretation. As such, the analysis focuses on the broader psychological constructs rather than demographic differences. The correlation matrix presented in Table 3 was carefully verified to ensure internal consistency and symmetry. All reported coefficients are derived directly from the statistical output, and diagonal elements are standardised to 1.00. As expected in behavioural transport studies, correlations between demographic variables and attitudinal constructs are generally weak, and therefore not central to the explanatory framework of the study (Nordhoff et al., 2021; Molina et al., 2021). Together, these results establish the perceptual foundation required to interpret behavioural intention (RQ2), linking latent attitudes to willingness-to-ride outcomes. While these components are not estimated within a confirmatory latent variable framework, they provide an empirically grounded basis for understanding the structure of perceptions and their relationship to behavioural intention. This exploratory approach is consistent with early-stage research on emerging mobility technologies, where measurement models are still developing. It is important to note that PCA is an exploratory data reduction technique, and the components identified here do not represent confirmed latent constructs. Rather, they provide a parsimonious summary of correlated attitudinal variables, which can inform—but do not in themselves establish, underlying theoretical structures.

3.4 Travel behaviour and NEC mobility patterns (Q28–Q41)

This section provides contextual behavioural background and does not directly address the research questions but supports interpretation of acceptance patterns. Staff and visitors’ travel behaviour characteristics at a monthly scale show significant differences in multimodal use (Figure 5). Both groups were highly dependent on the car, with 85.3% of staff and 83.9% of visitors using it at least once a week (Q28). Walking was the second most prevalent mode for both groups, particularly for visitors (95.3% use weekly), which aligns with on-site, visitor-led trips to the NEC (Q31). Visitor use of the bus was considerably higher than staff use (38.7% and 28.1% use at least weekly, respectively; Q29), whereas train use in both groups was moderate and similar (around 23%). All other active travel and shared micromobility modes were low, with less than 8% of respondents cycling (Q32–Q33) and around 1% using shared e-scooters or hire bikes at least once a week (Q34). This is similar to wider transport patterns in the United Kingdom, where private car use remains high and micromobility modes are not widely adopted. Figure 5 shows weekly travel frequencies by mode and respondent group.

Staff and visitors also reported different levels of spatial travel activity across the NEC estate (Table 4). A large share of staff members reported regularly travelling within the NEC campus (60.0% do at least once every 2 weeks; Q35) and between Birmingham Business Park and the NEC (63.8%; Q36). In comparison, visitors had low levels of mobility in these areas (17.8% and 19.4%, respectively). These differences in trip patterns suggest that a shuttle could have dual and complementary uses to provide regular, short-distance trips for staff with work-related travel requirements and support visitor trips during events Mode choice (Q37) reflected typical NEC behaviour, with the car dominating (72% of staff and 54% of visitors). Walking was the second most popular mode (31% of visitors and 22% of staff). Respondents were almost universally likely to travel alone (Q38) when moving around the site (67% visitors, 74% staff). The most popular reasons for choosing a mode (Q39) were travel time and convenience. 78% of visitors and 81% of staff selected travel time, and 69% and 74% selected convenience. Cost was selected by 42% of visitors but only 18% of staff. Sustainability was selected by ∼20% across both groups, indicating it was not a top priority. Travel times (Q40) were bimodal with morning and afternoon peaks. Staff travel between sites peaked in the morning (63% of staff travelled in 08:00–10:00, but 54% of visitors). Visitor arrivals peaked between 10:00 and 13:00. Afternoon travel peaked in similar proportions (aligned with visitors attending late events, and staff shift patterns). Trip purposes (Q41) were as expected. 72% of staff were travelling for work duties, meetings, or operational activities, while 68% of visitors travelled for events, leisure, or wayfinding between NEC venues. 18%–22% of respondents from each group reported moving to meet someone else on-site.

TABLE 4. Frequency of NEC & BBP–NEC trips.

These observed travel patterns provide important context for interpreting willingness-to-ride outcomes. The high reliance on private car use suggests the presence of established and habitual travel behaviours, which may act as a structural barrier to modal shift, even where stated willingness to use autonomous shuttles is high. At the same time, the frequent and predictable intra-site travel reported by staff indicates a clear functional need for short-distance, reliable mobility, which aligns closely with the intended operational design of autonomous shuttle services. This may help explain the slightly higher acceptance of higher levels of automation observed among staff compared to visitors. In contrast, visitor travel patterns are more intermittent and event-driven, with lower levels of intra-site mobility. This suggests that their engagement with the shuttle is more situational and less necessity-driven, which may contribute to more cautious acceptance, particularly in fully autonomous scenarios. Together, these findings indicate that travel behaviour is not only descriptive of mobility patterns but also plays a role in shaping the perceived usefulness and acceptance of autonomous shuttle services.

3.5 Barriers, benefits, and perceptions (Q42–Q46)

Participants expressed a more consistent set of concerns about the autonomous shuttles (Q42). As illustrated in Table 5, safety (25%) and factors closely related to safety and trust in the technology (23% for trust, 23% for reliability) were the most commonly cited barriers. Barriers related to impacts on job security (13%) or customer service (8%) were less common, and all other factors, including cost, vandalism, and loss of human interaction, were selected by <1% of participants. These concerns broadly match established patterns of AV acceptance; safety and trust in the technology are commonly cited and established barriers. The perceived benefits (Q43) show the inverse of this pattern. Participants overwhelmingly saw AS as providing environmental and/or reliability benefits (33% and 17% respectively), and a less-hassle travel experience (16%) (Table 6). Improved comfort (14%) and faster boarding (11%) were also commonly identified positive attributes. Cost-related benefits and convenience were selected at extremely low rates (<1%), suggesting that the shuttle is not expected to be a cheaper or more convenient mode but instead an improvement in service quality. Open-text responses indicated strong demand for easily accessible information regarding safety and emergency procedures, real-time route/location information, and clarity around the steward’s role (Q44). The most commonly cited reasons for not using the existing NEC bus service were all largely practical (e.g., low awareness, not needed, preferred to walk/drive, timetable didn’t fit needs; Q45). Feedback on the current bus service was mixed, but broadly acceptable with only a few negative comments (e.g., unclear timetable, poor ride quality; Q46). These results indicate that overall acceptance is related to three broad dimensions: Safety and trust, Expected service quality, and Comfort and ride experience. These closely map to the three components derived from the PCA reported above, and show consistency between the factor structure of attitudinal patterns and stated barriers/benefits. These findings reinforce the PCA-derived constructs (Section 3.3) and the observed behavioural patterns (Section 3.2), providing convergent evidence on the role of safety, trust, and reliability in shaping acceptance.

TABLE 5. Main concerns about using autonomous shuttles (Q42).

TABLE 6. Main perceived benefits of autonomous shuttles (Q43).

3.6 Staff-only questions (Q47–Q48)

Staff mobility behaviours were analysed in isolation to identify workplace-specific travel patterns around Birmingham Business Park and the NEC. Commuting (Q47) was dominated by car use: 63% travelled as a car/van driver and 6% as a passenger, which demonstrates a reliance on private vehicles. Proportions for other modes were much lower: bus (6%), shuttle bus (6%), train (9%), walking (7%), taxi (2%) and insignificant shares for car sharing and other public transport (<1%). 1% reported not travelling to any of the sites. The current dominance of car use for staff commuting suggests a level of car dependency that could be disrupted by an AS offering an alternative for short-distance intra-campus and last-mile trips. Mobility around the workplace was spread across the week (Q48). 41% of respondents attended on 3–4 days per week, with similar shares as 5 or more days (35%) and 1–2 days (21%). 3% worked fully remotely. The overall distribution of responses points to a stable and regular demand for travel on weekdays, demonstrating the operational benefit of frequent and reliable AS services during peak commuting times.

4 Discussion and implications

4.1 Key findings

This study yields five key findings regarding pre-use acceptance of autonomous shuttles at the NEC:

Behavioural intention to use autonomous shuttles (operationalised as willingness to ride) is strongly conditional on the presence of human supervision. In line with the study’s conceptual framework, this finding reflects pre-use acceptance expressed through stated intention, rather than observed adoption behaviour. This indicates a strong reliance on human presence during early stages of deployment.

These findings align with established acceptance frameworks such as TAM, TPB, UTAUT, and MAVA, which emphasise that behavioural intention (operationalised here as willingness to ride) emerges from both affective evaluations (e.g., trust and perceived safety) and expectations of system performance, consistent with recent empirical studies on autonomous shuttle adoption.

Safety, trust, and reliability are the major concerns and driving forces for acceptance. Safety, trust and reliability were the most prominent ones among the variables in the WTR questions, the attitudinal components identified through PCA, and the barriers to using an autonomous shuttle identified. The above three factors are found to be the most prominent attitudinal patterns associated with behavioural intention of an autonomous shuttle in its early-stage deployment.

Travel behaviour patterns shape the functional acceptance of autonomous shuttles. Travel behaviour was found to influence how respondents perceive the usefulness and role of autonomous shuttles. Staff exhibited frequent, predictable, and function-driven mobility patterns within the NEC, which align well with the operational characteristics of autonomous shuttle services (short-distance, repetitive routes). This helps explain their slightly higher acceptance of higher levels of automation. In contrast, visitors demonstrated more occasional, event-driven travel patterns, with lower levels of intra-site mobility, suggesting that their engagement with the shuttle is less necessity-based and more discretionary. As a result, their acceptance appears more sensitive to perceived safety and trust concerns, particularly in fully autonomous scenarios. These findings indicate that acceptance is shaped not only by psychological perceptions but also by the degree to which the service aligns with existing travel needs and behaviours.

Staff members had slightly higher acceptance for higher levels of automation than the visitors. This could be due to staff being more familiar with the location than the visitors.

4.2 Relevance of findings for autonomous shuttle adoption in the United Kingdom context

In many ways, our results here are in agreement with what has been documented for other United Kingdom and European trials of autonomous shuttles; that the issues of trust, safety perception and human intervention are particularly important at early stages of deployment (Paddeu, et al., 2020; Nordhoff, et al., 2021; Celina, et al., 2021; Fonzone, et al., 2025) and that the presence of reduced human supervision is associated with a gradient of acceptance, as has also been noted in other United Kingdom testbeds such as Milton Keynes and Bristol, with the perception of on-board staff being a key reason for trial uptake due to the reassurance they offer.

In the United Kingdom policy environment, where the regulated nature of our roads and autonomous vehicle deployment being at a stage of pilots and testbeds means that large-scale immediate rollout is not an immediate option, such results must support a gradual, step-wise approach to automation deployment that emphasises high reliability of operations with visible human support and safety communications, instead of early roll-out based on perceptions of autonomous technology alone. The findings also suggest that messaging around autonomous shuttles as a mobility service for first-/last-mile and campus circulation, rather than as a public transport replacement service, will be important for acceptance.

In large and complex sites such as the NEC, autonomous shuttles seem most likely to be acceptable when part of a wider transport ecosystem which includes rail, walking and traditional public transport modes. Targeting initial service use to predictable staff travel with capacity and scope to flex to meet demand for visitors driven by events, may also offer a route to scaling up their use, while managing public perception and risk.

4.3 Policy recommendations

Building on the behavioural and attitudinal patterns reported in the results section, a set of targeted, evidence-based policy recommendations was developed to guide future AS deployments at the NEC and comparable United Kingdom campuses. These recommendations translate the statistical results, PCA components, and identified barriers and benefits into actionable measures for transport authorities and operators (see Table 7).

TABLE 7. Policy recommendations for autonomous shuttle deployment.

4.4 Limitations

The results of this study may not be generalised to other settings as they are representative of the opinions in only one controlled-site setting. Responses to the intention-based survey items capture stated behavioural intention rather than observed behaviour, and therefore may not directly translate into real-world usage, as behavioural intention is subject to change when individuals encounter real-world constraints, direct experience, and observed system performance during deployment. This reflects the well-established Intention–Behaviour Gap, whereby intentions are subject to change when individuals encounter real-world constraints, system performance, and direct experience with the technology. While the research questions are theoretically informed by frameworks such as TAM, TPB, and UTAUT, the present study does not employ a confirmatory latent variable approach. Instead, it adopts an exploratory strategy to identify attitudinal structures, which can inform future research using SEM to formally test relationships between constructs. The attitudinal structure identified through PCA aligns conceptually with established acceptance frameworks such as TAM and TPB, suggesting that affective evaluations (e.g., trust and perceived safety) and performance-related expectations may operate as distinct but related influences on behavioural intention. However, these relationships were not formally tested within a latent variable framework and should be examined in future research using SEM. Differences in sample sizes and group compositions between staff and visitors may have affected the results. Data for travel behaviour was self-reported and thus could have been biased by recall error. The length of responses to the open-ended questions was not sufficient for a more in-depth interpretive analysis. Other attitudinal variables could be considered to optimise the PCA structure in future studies. Although the study captures perceptions prior to direct exposure to the autonomous shuttle service, participants may have been influenced by prior awareness, on-site communication materials, or informal information about the planned deployment. While such exposure was limited to general engagement activities and did not involve direct interaction with the service, it may have shaped anticipatory attitudes and expectations. Future research could further isolate purely uninformed perceptions or compare responses across varying levels of prior awareness. In addition, the willingness-to-ride scenarios were presented together, which may have introduced comparative or ordering effects in participant responses. Participants may have interpreted the configurations as representing increasing or decreasing levels of safety, rather than evaluating each scenario independently. Furthermore, while general information about the presence of onboard staff was provided, the specific roles of “driver” and “steward” were not formally defined within the questionnaire items. This may have led to variability in interpretation and contributed to the observed preference for configurations with greater visible human presence. Future research could build on these findings by applying Structural Equation Modelling (SEM) or other latent variable frameworks to formally test causal relationships between constructs such as safety perception, trust, and behavioural intention, using the perceptual structure identified in this study as a foundation.

5 Conclusion

The research presented in this paper formed part of a wider study of a trial of driverless electric shuttles at the NEC in Birmingham, United Kingdom. Pre-trial surveys were undertaken with visitors to, and NEC staff to understand attitudes towards, and travel behaviour of, these groups and their willingness to ride AS under different service operating conditions. These provide novel evidence on pre-use public acceptance of autonomous shuttles in a large-scale United Kingdom trial setting. Pre-use acceptance was investigated through behavioural intention (willingness to ride), as direct observation of adoption behaviour was not possible at this stage.

This research makes three key contributions. First, it is found that acceptance of AS services is conditional on the presence of direct or remote human supervision. People are willing to ride on driver-operated AS services, but willingness to ride was substantially lower under steward-only or fully driverless operations, reflecting concerns related to safety, trust and reliability. Second, the results show two broad attitudinal groupings, identified through PCA, are associated with behavioural intention and perceptions of autonomous shuttles: psychological acceptance, related to trust and affective perceptions of safety; and service and operational expectations, including reliability, frequency, pricing and on-board safety. This suggests that both technology-related perceptions and service performance expectations are important for shaping acceptance. Third, analysis of travel behaviour between NEC staff and visitors revealed distinct mobility needs and travel patterns, which suggest that early-stage deployments should prioritise frequent and predictable short-range internal journeys, while retaining flexibility to meet visitor demand during events.

Taken together, these results suggest that AS could play a meaningful role in supporting internal mobility within large and complex destinations such as the NEC, if deployment strategies prioritise human oversight, visible safety and security, and reliable service provision. This paper advances the existing evidence base by providing large-scale empirical insight into pre-use public acceptance of AS services in a real-world United Kingdom trial setting, to inform the design, communication and governance of early-stage deployments. While future research may build on this work to further examine how perceptions of safety and trust, and the behavioural drivers of acceptance, change as users gain experience with autonomous shuttle services, the results presented here provide a robust evidence base on which to inform early-stage trial design and deployment decisions.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/ffutr.2026.1833505/full#supplementary-material

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Footnotes

Edited by: Jianhua He, University of Essex, United Kingdom

Reviewed by: Nichole L. Morris, University of Minnesota Twin Cities, United States

Erik Marsja, Linköping University, Sweden