User experience, safety perceptions, and adoption intentions for autonomous shuttles: findings from a UK trial-phase survey

Autonomous shuttles (AS) are increasingly being tested in real-world environments, yet robust empirical evidence from live UK trials remains limited. This study investigates public awareness, user experience, and behavioural intentions during a public AS trial conducted at the National Exhibition Centre (NEC) campus in Birmingham, UK. An in-situ survey was conducted during the live AS trial, with 321 respondents participating. Of these, 51% were aware of the service, and 71% of those aware had used the shuttle. Users evaluated ten service attributes, including waiting time, boarding, comfort, responsiveness, and perceived safety. Principal component analysis (PCA) was used to identify the main dimensions influencing passenger experience. PCA identified a single underlying dimension, termed “overall passenger experience,” as the primary influencing factor, with consistently high ratings across all service attributes. A majority of users (62%) indicated they would “definitely” use the shuttle regularly if available, suggesting strong initial acceptance and positive user experience. Among non-users, the most frequently reported barriers were lack of information (38%), no perceived need (32%), and safety concerns (30%). The findings indicate a clear exposure effect, whereby direct experience with the shuttle was associated with more positive evaluations and stronger behavioural intentions. The study provides novel UK-based evidence from a live autonomous shuttle trial and highlights the importance of communication, service design, and real-world exposure in shaping public acceptance. The findings suggest that increasing public familiarity and addressing informational barriers may play a critical role in supporting the wider adoption of autonomous shuttles in urban mobility systems.

Autonomous shuttles (AS) have been widely studied as a potentially key component of emerging mobility systems. In particular, their ability to support first- and last-mile/connectivity and circulation between and within large and complex sites has been a focal point of recent literature. Low-speed and highly automated, AS typically serve relatively constrained and operationally prescriptive domains, such as campuses, innovation districts, business parks, and multimodal hubs. In these contexts, AS are often framed as a pragmatic intermediary solution to AV deployment, where technological capacity and public exposure to automation may be explored in operationally meaningful, but operationally bounded, conditions (Hafiz, et al., 2025; Feys, et al., 2020; Golbabaei, et al., 2022).

Emerging research suggests that AS can support first- and last-leg travel, enhance circulation within large, complex sites, improve the overall attractiveness of public transport systems, and support decarbonisation and sustainable mobility goals (Feys, et al., 2020; Golbabaei, et al., 2022; Liu, 2024; Xiao and Goulias, 2022). Slower operating speeds, sensor–suitcase–based localisation and obstacle-detection systems, and passive safety features are intended to effectively eliminate or reduce collision risk in these contexts and enable operation in mixed-traffic or semi-segregated environments (Rezwana and Lownes, 2024). Across recent European, Asian, and North American trials, they have also been shown to provide service to users that may be underserved by traditional public transport, such as older adults, people with mobility impairments, children, and non-residents unfamiliar with complex transport networks (Feys, et al., 2020; Sharma, et al., 2025; Larsson, et al., 2023). Integrated with larger public transport networks, AS have also been reported to increase overall perceived accessibility, in cases where fixed-route services are infrequent or otherwise impractical (Golbabaei, et al., 2022; Larsson, et al., 2023; Makahleh, 2024).

In spite of these trends, AS deployment in operational settings is still accompanied by a high degree of uncertainty for both developers and prospective users. On the technological side, this manifests as questions regarding system reliability, responsiveness, and actual performance in varied operating conditions. On the user side, this relates to safety, trust, comfort, and the broader ability to understand and interact with an automated service. The AV literature as a whole strongly suggests that lived experience is a decisive factor in how these issues are framed and understood. The act of participating in a trial period in itself consistently reduces uncertainty and changes expectations around comfort, safety, and perceived reliability (Nordhoff, et al., 2018; Eyssartier, et al., 2023; Makahleh, et al., 2026a). This has been noted both in studies that compare users and non-users of AS and in trial-based pre- and post-experience evaluations, where direct exposure to an automated service has shifted user perceptions compared to initial expectations (Nordhoff, et al., 2018; Fonzone et al., 2024; Fonzone, et al., 2025; Makahleh, 2026).

The quality of the user experience during AS trials is shown to be highly sensitive to factors at the service level, such as the smoothness of the vehicle motion, braking and acceleration, responsiveness to obstacles and other road users, and quality of information provided to the passengers during the ride (Bellone, et al., 2021; Luo, et al., 2024; Bansal, et al., 2016). These experiential factors map onto key constructs in information systems and technology acceptance, where factors like perceived usefulness and ease of use have consistently been shown to predict ride intention in the AS literature (Golbabaei, et al., 2022; Luo, et al., 2024; Pascale, et al., 2021). Safety perception is consistently found to be the key determinant of these and other constructs. Across trial-based studies, a perceived ability to physically protect passengers from collisions or other forms of harm, high system reliability, and responsiveness to hazards are frequently identified as strong predictors of acceptance and willingness to use an autonomous public transport service (Sheykhfard, et al., 2026; Xiao and Goulias, 2022; Diba, et al., 2025; Korkmaz, et al., 2022). This is echoed in the broader AV literature, which consistently reports perceived collision risk and system unpredictability to be influential factors in adoption behaviour (Nazari, et al., 2026). These perceptions are also shown to be strongly conditioned by the characteristics of the operating environment, with high pedestrian densities, tight spatial layouts, or any other factors that necessitate frequent close interactions with people or other obstacles, leading to increased safety assessments (Bellone, et al., 2021; Makahleh and Makahleh, 2024; Makahleh, et al., 2024b).

A construct closely associated with safety is trust in automation. Users tend to develop trust through observing competent and predictable system behaviour, system/operator communication, and consistent performance over repeated trips (Nordhoff, et al., 2018; Nordhoff, et al., 2022; Luo, et al., 2024; Yang and Kim, 2025). Prior studies have also found evidence of the joint effects of safety perception and trust on behavioural intention, with perceived safety and trust reinforced through positive ride experiences that increase users’ willingness to reuse automated mobility services (Sheykhfard, et al., 2026; Golbabaei, et al., 2022; Paddeu, et al., 2020; Nordhoff, et al., 2021; Kenesei, et al., 2025). However, the evidence base remains heavily skewed towards hypothetical or scenario-based survey methods as opposed to observations based on actual use of operational services (Eyssartier, et al., 2023; Nazari, et al., 2026). Studies simultaneously capturing multiple components of user experience, safety perceptions, non-user barriers, and adoption intentions are also still somewhat limited, especially for the UK context. While evidence from international deployments and field experiments is growing rapidly, including in emerging-market contexts (Wiedyanoe, et al., 2024), UK-based, trial-phase empirical evidence has remained scarce to date. Recent reviews have therefore called for more empirically grounded evaluations of autonomous mobility, considering behavioural, infrastructural, and contextual factors (Makahleh, et al., 2024a).

In the UK context, several AS trials have been facilitated by government agencies such as the Centre for Connected and Autonomous Vehicles (CCAV) and the Department for Transport (DfT), including the Cambridge Autonomous Shuttle Trial, CAVForth in Edinburgh, and the Solihull and Coventry Automated Links Evolution (SCALE) trial operating on the National Exhibition Centre (NEC) campus in Birmingham (Connecting Cambridgeshire, 2021; Transport Scotland, 2020; NEC Birmingham, 2024; Solihull Council, 2025). These trials represent a unique opportunity to study public perceptions of and engagement with automated mobility in real-world operating conditions. At the same time, these trials offer the chance to collect regulatory, organisational, and operational learning which can inform longer-term deployment planning and development. Importantly, these evaluations of user responses also allow the service to be assessed not only in relation to technology and performance itself but also to the way the service is presented to and integrated into a local transport environment. It should be noted that the trials were conducted within the UK regulatory framework, including the DfT Code of Practice for automated vehicle trialling and relevant BSI PAS guidance (e.g., PAS 1881, PAS 1883, PAS 1884), which define safety assurance, operational design domains, and the role of safety operators (BSI Group, 2021).

Against this background, this paper reports findings from a trial-phase survey conducted during live public operation of an autonomous shuttle at the NEC campus in Birmingham, UK (April–May 2025). The study draws on in-situ responses from both users and non-users exposed to the service environment, providing empirical evidence on awareness, actual use, service evaluations, perceived safety, barriers to uptake, and behavioural intentions in a real-world trial context. This study forms part of a broader evaluation programme of the NEC autonomous shuttle trial. Within this programme, two complementary studies provide additional context. The first is a pre-trial survey conducted prior to public deployment, which examined expectations, safety perceptions, trust, and willingness to use the service before direct exposure. The second is a published focus group study, which explored deeper reflections on trust, communication, and policy readiness. The present study builds on these by examining how perceptions and behaviours are expressed during live operation, enabling direct comparison between users and non-users within the same operational setting. There is no respondent overlap between the datasets.

The first is a companion pre-trial survey study conducted before public operations began, which examined expectations, safety perceptions, trust, and willingness to use the shuttle prior to direct exposure. That study established baseline attitudes and highlighted safety, trust, and willingness-to-use as key issues requiring examination under real-world trial conditions. The present paper builds on that earlier work by examining how these issues are expressed during live public operation, and there was no respondent overlap between the two surveys. The first is a companion pre-trial survey study conducted before public operations began, which examined expectations, safety perceptions, trust, and willingness to use the shuttle prior to direct exposure (Makahleh et al., 2026b). That study established baseline attitudes and highlighted safety, trust, and willingness-to-use as key issues requiring examination under real-world trial conditions. The present paper builds on that earlier work by examining how these issues are expressed during live public operation. There was no respondent overlap between the two surveys.

The second related study is a published focus group paper based on the same NEC autonomous shuttle trial (Makahleh et al., 2026a). That study explored participants’ deeper reflections on autonomous shuttle adoption, particularly the roles of trust, visible human support, communication, and policy readiness. By contrast, the present paper provides broader survey-based evidence from a larger trial-phase sample and enables quantitative assessment of awareness, differences between users and non-users, service evaluations, safety perceptions, and behavioural intentions. In this way, the current study complements the focus group findings by testing, at a broader sample level, several themes that emerged qualitatively while also addressing questions that qualitative evidence alone could not answer. These limitations in the existing literature highlight several unresolved issues that are particularly relevant for real-world deployment contexts.

Taken together, the literature and the related NEC studies point to three specific gaps that motivate the present paper. First, robust UK-based trial-phase survey evidence from live autonomous shuttle operations remains limited. Second, few studies simultaneously compare users and non-users within the same live operational setting. Third, limited evidence is available on how service evaluations, perceived safety, and behavioural intentions are structured during actual trial exposure rather than under hypothetical or pre-trial conditions. To address these gaps, the present study uses data from the NEC live autonomous shuttle trial to examine how exposure, service experience, safety perception, and communication shape trial-phase acceptability in a real-world UK context.

While survey-based studies of user perception are well established in the literature, the present study advances the field in a more integrated and empirically grounded way. First, it is based on in-situ data collected during live autonomous shuttle operations in a real-world UK trial, rather than hypothetical or pre-exposure scenarios. Second, it enables direct comparison between users and non-users within the same operational environment, capturing both experiential and non-experiential perspectives on awareness, safety, concerns, and adoption pathways. Third, it examines user experience through a broad set of service-attribute measures, including waiting time, boarding and alighting, information provision, perceived speed and responsiveness, comfort, and service frequency, thereby offering a detailed account of how the shuttle journey is evaluated in practice. Fourth, the study integrates user experience, safety perception, and behavioural intention within a single empirical framework, providing a more comprehensive understanding of trial-phase acceptability. Finally, by applying principal component analysis, the study identifies the latent structure underlying service evaluations and examines how this experiential dimension relates to intention to use the shuttle regularly. Taken together, these contributions provide novel UK-based evidence on how exposure, service design, and communication shape the acceptability of autonomous shuttles in a live trial context. The analysis is guided by the following research questions (RQs): RQ1: How do users and non-users differ in demographic characteristics, awareness and exposure to the autonomous shuttle service?

RQ2: How do users evaluate key service attributes of the autonomous shuttle, and what underlying dimensions of user experience can be identified through principal component analysis?

RQ3: What factors shape users’ safety perceptions, including those that enhance feelings of safety and those that contribute to feeling unsafe, and do these concerns change after first use?

RQ4: What barriers, concerns, information needs and improvement preferences influence non-users’ consideration of future autonomous shuttle use, and how do these compare with users’ adoption intentions?

By addressing these RQs within a real-world trial context, the study advances understanding of public acceptance of AS and offers practical guidance for authorities, operators and technology developers. In particular, it provides evidence to support refinement of service design, communication strategies and safety-related interventions to foster public readiness for wider deployment of automated mobility.

Shuttle service and operational route

The study was embedded within the live-trial phase of the Solihull & Coventry Automated Links Evolution (SCALE) project at the NEC Hub, Birmingham, in the United Kingdom. The fleet comprised fully electric AS manufactured by Ohmio that accommodated 10–14 passengers and operated under onboard supervision by a safety operator, consistent with the UK regulatory framework for automated passenger services (Figure 1). The term safety operator is used to describe the on-board supervisory role during the trial. The AS was and deployed on a fixed route across the campus. During this phase, the vehicles ran the same defined loop used by other vehicles on the Hub, thereby placing users and non-users in an operationally relevant setting. Each journey was monitored in real time from the regional control centre, ensuring operational safety and data integrity (Solihull Council, 2025). The service was accessible to both existing users of the campus transport network and visitors who had not yet used the shuttle, enabling comparison of both groups.

Respondent recruitment and procedure

The trial-phase survey was conducted during the public operation of the AS trial at the NEC campus in April and May 2025. This survey formed a separate phase of data collection from an earlier pre-trial survey conducted before public operations began. No respondents were repeated across the two surveys. A total of 321 respondents completed the questionnaire through on-site intercept surveys using tablets or QR codes, including at shuttle stops, along the walking corridors of the NEC campus, and on board the AS. The recruitment strategy was consistent with previous AS acceptance research, involving naturalistic in-situ recruitment of both riders and non-riders (Bellone, et al., 2021; Eyssartier, et al., 2023). Respondents first indicated if they were aware of the AS trial. Of the 321 total respondents, 163 (50.8%) were classified as Aware and 158 (49.2%) were classified as Not Aware. Given that user adoption and non-adoption can only be determined amongst those who are aware of the service, analysis for the remainder of the study is only focused on the 163 respondents who were aware of the service. Of those who were aware of the AS trial, respondents were then asked if they had used the service during the trial. Of the 163 aware respondents, 116 were classified as Users (71.2%) and 47 were classified as Non-Users (28.8%). Users were individuals who had used the shuttle at least once during the trial, and Non-Users were individuals who were aware of the AS service but had not travelled on it. This two-step classification process of first differentiating respondents based on awareness, followed by user status, is methodologically consistent with previous autonomous mobility trials, and is recommended for ensuring strict separation of exposure-based and perception-based responses (Nordhoff, et al., 2018; Sheykhfard, et al., 2026). All respondents were at least 18 years of age and gave informed consent to participate. Respondents were not incentivised for their participation, and all responses were anonymous.

Questionnaire content

The questionnaire was structured to collect in-use perceptions from both users and non-users during a live trial of an autonomous shuttle service. A structured survey was preferred to enable a systematic comparison of users and those aware but choosing not to use the service. This method is appropriate for trial-phase evaluations of emerging transport technologies, where it is important to gather at scale and in an operational setting measure of behavioural intention, safety perception, and experience-based responses. The questionnaire was therefore organised into blocks based on exposure level, to ensure that only users answered experience-based questions and non-users could only report on barriers, concerns and information needs. The questions are provided in Supplementary Material S1.

Demographics and participant characteristics (Q1–Q7)

The questionnaire first collected demographic and contextual information about the respondents. Respondents were asked to provide their age (Q1) and gender (Q2), before being asked to select their ethnicity from a list of categories (Q3). Respondents were then asked to self-report having a disability (Q4) and Blue Badge permit that allows access to reduced mobility parking (Q5). Respondents were then asked to provide their location (Q6), by answering whether they lived in the West Midlands, in other areas of the UK, or outside the UK. Finally, they were asked to select their main purpose of visiting the NEC Hub (Q7), with options including commute, lunch within working hours, general leisure, activity-based leisure, or other. The above variables were used to profile the sample and used in comparing Users and Non-Users. The evidence that such demographic and situational variables impact attitudes to AVs was used in this consideration.

Awareness and AS usage (Q8–Q10)

The following section of the questionnaire determined exposure and awareness of the AS. First, participants were asked if they knew that an autonomous shuttle was currently operating on the NEC campus (Q8). If answered “Yes” to this question, the next question asked if the person had ever taken a ride on the shuttle (Q9). This allowed for segregation of Users (yes to both questions) and Non-Users (yes and no, respectively) of the shuttle. Users were then asked how many rides they had taken on the AS (Q10) from one of four categories that corresponded to one to seven or more trips. The selection of questions in this portion of the questionnaire was done so in stages to ensure that only those with exposure to the service would be required to answer experience-based questions, while Non-Users would also be able to report barriers and concerns without prior assumed experience.

User experience and safety perceptions (users only: Q11–Q15)

Respondents who had used the AS rated their experience in terms of nine attributes on a five-point scale ranging from “Very Poor” to “Excellent.” These included the length of the waiting time before boarding (Q11.a1), the ease of boarding (Q11.a2) and alighting (Q11.a3), the amount and quality of information received during the ride (Q11.a4), the sense of safety onboard (Q11.a5), the perception of vehicle speed (Q11.a6), the vehicle’s responsiveness to the road layout and obstacle (Q11.a7), the sense of frequency (Q11.a8), comfort (Q11.a9), and journey experience (Q11.a10). The second comfort question was removed due to its close similarity with the first one. Respondents were also asked to specify in an open-ended question (Q11.b) the reason behind a “Poor” or “Very Poor” rating for any of the items above. AS users also reported the factors that contributed most to their sense of safety (Q12) and the ones that made them feel “somewhat” or “very” unsafe (Q13), up to three options for each question. Questions Q14.a and Q14.b were designed to capture the effect of exposure on perception. Respondents indicated whether they had any concerns before using the shuttle for the first time (Q14.a) and whether they still had those concerns after their experience (Q14.b). Finally, they were asked to indicate their likelihood of using the shuttle as a regular mode of transport if it were available (Q15.a) and were given an open-ended field to specify their reasoning (Q15.b).

Non-user barriers, concerns and intention (non-users only: Q16–Q20)

Respondents who indicated that they were aware of the AS but had not used it completed a block of questions specific to non-use. Non-users first listed all reasons that applied to them for not using the shuttle (Q16). Reasons included lack of awareness, safety, no need, and lack of convenience or route alignment. Non-users then selected which areas, if any, concerned them with regard to the possibility of using the shuttle in the future (Q17). Response options included safety, service reliability, technology reliability, comfort/convenience, and “none of these.” Respondents also selected which types of information would make them more likely to use the service (Q18). Options included information about the safety of the shuttle, instructions for use during an emergency, reliability of the technology, fares and payment information, and information about the frequency of the AS or environmental benefits. Non-users then selected which features, if improved, would make them more likely to consider riding the shuttle (Q19). Options included improvements to the route design, an increased frequency, safety or increased reliability and on-time performance. Finally, non-users selected the purpose for which they would most likely use the shuttle in the future, if allowed to use it (Q20). Response options included commuting, lunchtime and recreation. These items allowed for a comprehensive analysis of the elements that contribute to non-user trepidation and potential routes to future adoption.

Open-ended feedback and comments (Q21)

All respondents were offered an opportunity to provide further comments about the shuttle service in an open-ended question (Q21). These narrative responses and feedback supplemented the quantitative findings and offered additional insight into expectations, concerns, and broader attitudes towards AS deployment at the NEC campus.

Analysis of responses

All analyses were conducted using IBM SPSS Statistics v29, with Microsoft Excel used for data cleaning and recoding. Descriptive statistics were first used to summarise demographics, trip purpose, awareness, and usage patterns, providing an overview of the sample and supporting comparisons between Users and Non-Users. A binary logistic regression model was then estimated to examine which demographic and contextual factors were associated with shuttle use, with user status specified as the dependent variable (User = 1, Non-User = 0). In addition to descriptive reporting, the analysis was designed to explore relationships between exposure, overall passenger experience, and stated behavioural intention. The combined use of logistic regression, PCA, correlation analysis, and group-based comparisons enabled examination of both sample characteristics and associations between key trial-phase variables.

For respondents who had used the shuttle, principal component analysis (PCA) was applied to the ten service-experience items (Q11.a1–Q11.a10) to identify underlying dimensions of user experience and to reduce correlated items into a single interpretable construct. The suitability of the data for PCA was confirmed using the KMO statistic and Bartlett’s test. A single component was retained based on the scree plot and eigenvalues, and internal consistency was assessed using Cronbach’s alpha. The resulting standardised composite score (PCA1: Overall Passenger Experience) was used in correlation analyses and in modelling regular-use intentions.

Changes in safety concerns before and after first use of the shuttle were assessed using a McNemar test, which is appropriate for paired categorical responses. For Non-Users, multi-response items relating to barriers, concerns, and information needs (Q16–Q19) were coded as binary variables and analysed using frequencies and chi-square tests where applicable, allowing identification of the most salient deterrents to use. Intended future trip purposes (Q20) were summarised using percentages. Finally, open-ended responses (Q11.b, Q21) were analysed inductively using thematic analysis to capture recurring themes that complemented the quantitative findings. All statistical tests were two-tailed and assessed at p < 0.05.

Demographic and background characteristics (Q1–Q7)

A total of 321 respondents completed the trial-phase survey during the period of public deployment of the autonomous shuttle at the NEC campus (April–May 2025). The sample shows a wide spread of demographic and situational characteristics; indicative of the relatively heterogeneous mix of visitors and workers present on the NEC Hub. While gender was considered in the analysis, no robust differences were identified. However, this finding should be interpreted cautiously due to the male-dominated sample.

Respondents were spread across a wide range of age groups, with the largest proportions in the 35–54 age band, but also good representation from both younger and older groups. The sample was male-dominated (77%), which reflects the makeup of visitors and employees present on the NEC site during the trial period. In terms of ethnicity, the largest group of respondents identified as White (77.9%), with some respondents also identifying as Asian (15.0%), Black (5.6%) and mixed/other ethnic groups. Although this is a less diverse sample than that of Birmingham as a whole (where ∼ 52% of residents identify as non-White), this ethnic make-up of the sample is broadly in line with the observed visitor and workforce mix at the NEC campus during large-scale business and exhibition events, rather than with that of the residential population in the surrounding area. A large majority of respondents reported having no disability and no Blue Badge status. There was a small but visibly present proportion of respondents reporting reduced mobility needs. Slightly more than half of the sample were based outside of the West Midlands but within the UK, while a smaller proportion were international visitors, reflecting the NEC’s profile as a national and international venue for events. Trip purposes were quite varied, including business travel, leisure/exhibition attendance, on-site work, as well as social/recreational visits. The demographic and background characteristics suggest that the sample broadly covers a mix of user groups that is relevant to the NEC context, and supports the subsequent comparisons between users and non-users of the AS. Full descriptive statistics are reported in Supplementary Table S1; Supplementary Material S2, while the main text here highlights the key patterns relevant for interpretation of the results.

Awareness and use of the AS (Q8–Q10)

Of the 321 participants, 163 (50.8%) indicated they were aware that the autonomous shuttle was operating on the NEC campus, while 158 (49.2%) did not. Of those aware of the service, 116 (71.2%) of the respondents indicated they had taken at least one ride on the shuttle. The distribution of trips taken among shuttle users was as follows: Most had taken it only a few times. 62.1% of respondents indicated they had taken the shuttle only once, 29.3% took it two or three times, and 8.6% took it four or more times. These results show that the shuttle was used for the most part only a few times, rather than repeatedly or habitually, during the trial. To provide a clearer overview of the progression from awareness to actual use, Figure 2 illustrates the distribution of respondents across the full sample, those aware of the service, and those who subsequently used the shuttle. The figure highlights a noticeable drop-off between awareness and usage, indicating that awareness alone does not necessarily translate into behavioural engagement with the service.

PCA findings: user experience with the AS (Q11.a1–Q11.a10)

Ten items assessed respondents’ experience with the AS (Q11.a1–Q11.a10), including waiting time, ease of boarding/alighting, ride information, perceived safety, vehicle speed, system responsiveness, service frequency, travel comfort, and the overall journey experience. The suitability of the data for PCA was evaluated. All ten experience items were positively inter-correlated at moderate-to-high levels (full correlation matrix in Supplementary Table S2; Supplementary Material S3). This pattern of inter-correlations evidences the conceptual relatedness of the aspects of experience with the AS that are measured by each item and, thus, justifies using PCA. Figure 3 presents the mean ratings across all service attributes, providing a visual representation of the consistently high evaluations reported by users. The narrow range of scores reinforces the PCA findings, suggesting that users perceive the service as a uniformly positive and cohesive experience rather than distinguishing sharply between individual attributes.

PCA was conducted on the z-standardised item scores. Component eigenvalues were inspected to determine the number of components to retain. The scree plot (Figure 4) shows a clear elbow after the first component: While Component 1 has an eigenvalue of 5.20 and accounts for 51.8% of the total variance, only the second component marginally exceeded an eigenvalue of 1.0 (Component 2 = 1.14) and all other components had eigenvalues well below 1.0 (e.g., Component 3 = 0.89). These results provide strong statistical evidence for retaining a single-component solution, in line with conventional criteria (Kaiser criterion) and with the findings of previous empirical studies that have examined users’ experience of AS (Nordhoff, et al., 2018; Golbabaei, et al., 2022; Eyssartier, et al., 2023).

A one-factor solution was therefore retained and interpreted as PCA1: Overall Passenger Experience. All ten items loaded positively and strongly onto this component, with standardised loadings ranging from 0.62 to 0.86. Thus, each item meaningfully contributed to the latent construct. The highest loadings were observed for overall journey experience (0.86), system responsiveness (0.84), perceived safety (0.82), and travel comfort (0.76), while the lowest loading was observed for service frequency (0.62), which was, however, well above accepted thresholds. These results suggest that the respondents experienced these service attributes cumulatively rather than as distinct or separable dimensions of experience.

The internal consistency of the ten-item scale was high (Cronbach’s α ≈ 0.89), suggesting that the items reliably measure a unified latent construct. Descriptive statistics are consistent with the PCA results. Mean scores across the ten items ranged from 4.23 to 4.55, while standard deviations ranged from 0.63 to 0.78, indicating that the ratings were consistently positive and exhibited moderate dispersion. The highest ratings were given to the ease of boarding (M = 4.55; SD = 0.64), alighting (M = 4.55; SD = 0.65), and ride information (M = 4.55; SD = 0.64), while the lowest (yet still favourable) ratings were given to service frequency (M = 4.23; SD = 0.78) and vehicle speed (M = 4.32; SD = 0.69). Based on these results, a composite score was computed that represents PCA1: Overall Passenger Experience. This composite was standardised (M = 0; SD = 1) and was used in the correlation analyses with demographic and user-status variables (Table 1) as well as in the subsequent models of adoption and regular-use intentions. This analytical step allowed the study to move beyond item-by-item description and to examine whether a broader experiential construct was associated with downstream behavioural intention.

User safety perceptions and the effect of exposure (Q12–Q14)

Perceived safety was measured with three survey items that inquired (i) which factors made respondents feel safe riding the shuttle (Q12), (ii) which factors made them feel unsafe (Q13), and (iii) whether they held any concerns before and after exposure (Q14) (Table 2). Among the 114 users who responded to Q12, the most commonly reported factors contributing to feeling safe were the presence of a safety operator onboard (81.6%), the shuttle’s operation and speed (74.6%), and its safety features such as collision sensors and emergency braking (54.4%). Signage and instructions were less frequently selected (34.2%), and few respondents selected other factors (6.1%).

Figure 5 provides a comparative visualisation of the factors influencing perceived safety and sources of unease among users. While most respondents reported feeling safe, the figure highlights that concerns remain concentrated around specific operational aspects, particularly obstacle handling and vehicle behaviour.

Fewer users (N = 19) selected at least one response in Q13 (Table 2), in which users were asked which factors contributed to feeling unsafe while riding the shuttle. As in Q12, the most commonly cited factor was the shuttle’s handling of obstacles or pedestrians (10 users; 52.6%), followed by operation and speed (8; 42.1%), and the shuttle’s advanced safety features (8; 42.1%). Signage and instructions were again much less commonly mentioned (2; 10.5%), and a Similar result where the vast majority of users feel safe, but a minority remains apprehensive about unlikely but very consequential operational edge cases have also been found in previous AV trials (Tengilimoglu, et al., 2023; Naiseh, et al., 2024). Before riding the shuttle, most users did not feel that there were any safety concerns (99 of 116 users; 85.3%; Table 3)). Of users who indicated that they held safety concerns and answered the follow-up Q14.b question (N = 14), half (7; 50.0%) no longer felt they had safety concerns after riding the shuttle. Five (35.7%) of users reported that some safety concerns persisted, while 2 (14.3%) stated that all their safety concerns remained unchanged.

Given the small subsample size, these results are reported descriptively. They indicate that some users reported fewer safety concerns after riding the shuttle, while others continued to express concerns following exposure.

Intention to use the AS regularly (Q15)

Among users (n = 116), most reported positive intentions to adopt the autonomous shuttle, with 62.1% indicating they would “definitely” use it and 24.1% indicating they would “probably” use it. In the free-text box, the most frequently given reason for regular use was convenience, which was expressed in terms of time saved, reduction in walking, and an easy way to get around the NEC campus. Qualitative analysis of the open-ended responses revealed a number of common underlying themes to intentions to use the shuttle regularly. Convenience-related reasons were most commonly cited, with just over two-thirds of respondents referencing reduced walking or general ease of movement, or the convenience of the service. Safety-related considerations were cited by approximately a third of respondents, either as a positive reason to use the shuttle (e.g., feeling safe, reassured) or, less commonly, as a reason not to use it. Enjoyment, novelty, or interest in new technology was raised by approximately a quarter of respondents, often by describing the shuttle as “fun,” “interesting,” or “exciting.” Practical considerations such as reliability, availability and usefulness for getting to and from specific destinations (commuting, getting between venues, etc.) were also commonly mentioned. Accessibility-related reasons, including reduced physical effort required or suitability for people with mobility impairments, were mentioned by a smaller but substantial minority of respondents. References to environmental benefits or support for innovation more generally were comparatively rare. The qualitative evidence, therefore, suggests that intentions to use the shuttle regularly were due to a mixture of practical utility and experiential appeal, rather than a single factor.

Non-user barriers, concerns and future use potential (Q16–Q20)

Factors given for not using the service (n = 47) include lack of information (38.3%) no need for the service (31.9%) and safety concerns (29.8%). Fewer respondents mentioned convenience-related issues or routing limitations. In a follow-up question (Q17) asking respondents to elaborate on their concerns about using the AS, safety again was the most commonly stated (40.4%), as well as concerns about the reliability of the technology (36.2%) and the punctuality of the service (34.0%) (Table 4). These are also consistent with common concerns that have been seen in other public trials, where people questioned the vehicle’s ability to detect obstacles and unexpected events or its consistency in service provision (Fonzone, et al., 2025). The largest proportion of non-users said they would be more willing to try the service if given clear safety information (42.6%), followed by on the reliability of the technology (34.0%) and on what to do in an emergency (27.7%). Only 8.5% said there was nothing that could change their mind (Table 4). As illustrated in Figure 6, informational barriers emerge as the most prominent constraint among non-users, followed by perceived lack of need and safety concerns. This visualisation reinforces the interpretation that non-use is driven more by informational and perceptual gaps than by direct negative experience.

When non-users were asked what improvement to the shuttle would make them more likely to use it (Q19), better safety features and reliability and on-time performance were mentioned most often, followed by a greater number of stops and routes. A smaller share of non-users said that no improvement would change their mind. For the question of potential future use (Q20), leisure and social trips were most commonly mentioned, followed by commuting and short trips within the site (Table 5).

Qualitative feedback from open-ended responses (Q11.b, Q21)

A small number of users provided open-ended comments in Q11.b (only for users) and Q21 (for users and non-users). As noted, most of participants did not provide any additional feedback, however, the few comments they provided shed some light on specific experiences and aspects of the autonomous shuttle.

Key themes in the qualitative comments (see Table 6) point to sudden braking or deceleration, particularly when the shuttle was near pedestrians and/or other vehicles as the most frequently mentioned issue or problem. A smaller number of comments also mention accessibility related issues, particularly with the ramp or lift, the seating, or handrail location. The comments also call for improved signage or explanation of safety measures and rules. Furthermore, a few of the comments call for better visibility or promotion of the service.

In addition to these issues, a larger number of comments in this qualitative feedback highlight positive experiences and provide positive descriptors of the shuttle, for example, as fun, innovative, convenient, or as having on-board staff that provided good support. The limited amount of qualitative data available is provided here as additional context that supports, but does not go beyond, the quantitative findings.

Overall, the limited qualitative comments are presented as supplementary context and do not introduce findings beyond those identified in the quantitative analysis.

Discussion

The discussion is structured around the four research questions outlined in Section 1, linking the empirical findings to each question in turn to provide a clearer interpretation of the results. This study provides empirically grounded, trial-phase evidence from a real-world UK deployment, offering insights that extend beyond conventional perception studies based on hypothetical scenarios. Rather than demonstrating whether AS are viewed positively in general, which has been shown elsewhere, the key contribution of this study lies in identifying which aspects of the experience dominate users’ evaluations during early exposure, and where remaining points of friction persist even in a controlled environment. Among users, ratings of the shuttle across ten service attributes were consistently high, and the PCA results indicate that these attributes were experienced as a single, cumulative dimension of overall passenger experience, rather than as separable elements. This suggests that, at the trial stage, users tend to form a holistic judgement of the service, in which comfort, safety, information, and operational performance are not independently evaluated but blended into a general impression. While similar unidimensional experience constructs have been reported in other early trials (Nordhoff, et al., 2018; Eyssartier, et al., 2023), the present study demonstrates this pattern within a UK trial context and using in-situ, post-ride data, rather than hypothetical or pre-exposure assessments. Taken together, these findings suggest that trial-phase acceptability is not shaped by isolated service attributes alone, but by the interaction between exposure, perceived ride quality, safety reassurance, and perceived usefulness in context. These findings should be interpreted in light of the controlled trial environment, where both the service design and the characteristics of the user population may have contributed to relatively favourable evaluations. In relation to RQ2, the findings demonstrate that users evaluate the shuttle service through a set of closely interrelated attributes that form a single underlying dimension of overall passenger experience.

When compared with findings from previous autonomous shuttle trials, the results show a high degree of consistency, particularly in relation to overall user experience and the role of direct exposure. Studies such as Nordhoff et al. (2018) on the acceptance of driverless vehicles study and Eyssartier et al. (2023) on autonomous shuttle acceptance, similarly, report that users tend to evaluate autonomous shuttle services positively following direct interaction, with experience reducing uncertainty and increasing trust. The identification of a single, unified experience dimension in this study aligns with findings by Golbabaei et al. (2022) who explored autonomous shuttle adoption intention, where perceived usefulness, comfort, and safety were found to operate as interrelated constructs rather than independent determinants.

These findings are also consistent with the NEC focus group study reported by Makahleh et al. (2026a), which likewise identified trust, human support, and clear communication as central to autonomous shuttle acceptance, although the present survey extends that work by quantifying how these issues differ across users and non-users during live operation.

However, the present study extends these findings by demonstrating this pattern within a live UK trial context using in-situ data, whereas much of the existing literature relies on pre-trial or hypothetical scenarios. In contrast to some earlier studies where safety concerns were more dominant prior to exposure, the current findings suggest that, once experienced, concerns shift toward specific operational characteristics such as braking behaviour and obstacle handling. This refinement highlights a transition from abstract risk perception to concrete service evaluation, which has been less explicitly captured in prior research.

At the same time, the findings highlight a specific and practically relevant tension. Although perceived safety scores were high overall, open-ended feedback repeatedly identified abrupt braking and sudden deceleration as sources of discomfort or concern for a subset of users. Abrupt braking may reflect not only conservative algorithmic behaviour but also system-level factors such as sensor detection thresholds, controller tuning, and braking system performance. These factors can influence perceived safety even in low-speed operations. This indicates that positive global safety ratings can coexist with sensitivity to particular vehicle behaviours, even in low-speed, supervised operations. Rather than contradicting the quantitative results, this pattern refines them by showing that movement quality, rather than abstract safety perceptions alone, remains a critical determinant of comfort during early deployment. This aligns with prior work highlighting the importance of public safety assurance and transparent communication in automated mobility trials, where user confidence depends not only on technical performance but also on how safety is communicated and demonstrated to the public (Myklebust, et al., 2020). This observation is consistent with prior work showing that first-time users are especially responsive to vehicle dynamics and system transparency (Madigan, et al., 2017), but the present study contributes concrete evidence of how these issues emerge during an operational UK trial. Addressing RQ3, the results show that safety perceptions are shaped both by system characteristics, such as operational behaviour and safety features, and by experiential factors, with direct exposure reducing uncertainty for some users while specific operational behaviours, such as abrupt braking, remain areas of concern.

Intentions to reuse the shuttle were high among users, with most indicating they would “definitely” or “probably” use the service again if available. Qualitative responses suggest that these intentions were primarily linked to convenience-related factors, such as reduced walking effort, ease of use, and site accessibility, alongside interest in the experience itself. Importantly, these motivations were grounded in the specific spatial and functional context of the NEC campus, where long walking distances and dispersed destinations make short internal trips salient. This contextual dependence suggests that early adoption intentions for AS may be shaped as much by site characteristics as by attitudes toward automation in itself. These findings further contribute to RQ3, indicating that perceived safety and comfort, combined with contextual usefulness, play a key role in shaping behavioural intention following first exposure.

In contrast, non-users’ responses point to a different set of constraints. Lack of information, uncertainty about safety, and limited perceived relevance of the route were the most commonly reported reasons for non-use. These findings do not demonstrate resistance to automation as such, but rather indicate that the absence of exposure and limited-service visibility act as practical barriers to initial engagement. The contrast between users and non-users, therefore, illustrates a clear exposure effect: those who experienced the service evaluated it primarily in terms of operational qualities, while those without experience focused on informational and perceptual uncertainties. In relation to RQ4, the results highlight that non-use is primarily driven by informational barriers, perceived lack of need, and residual safety concerns, rather than outright rejection of the technology.

The barriers identified among non-users are also consistent with findings from other trial-based studies. For instance, the work of Fonzone et al. (2025) on autonomous bus trial perception and Paddeu et al. (2020) on passenger comfort and trust in relation to autonomous shuttlea highlight that lack of information, perceived irrelevance, and safety uncertainty are common deterrents in early-stage deployments. The present study reinforces these findings within a UK context, but further demonstrates that these barriers are not necessarily rooted in negative attitudes toward automation itself. Instead, they reflect limited exposure and insufficient communication, suggesting that non-use is more strongly associated with informational deficits than with active resistance. This distinction strengthens the interpretation that increasing awareness and trial accessibility may play a critical role in bridging the gap between intention and actual use.

The results do not suggest that autonomous shuttles automatically generate trust or acceptance. Instead, they show that early exposure shifts the basis of evaluation, from abstract concerns about safety and control toward concrete judgements about ride quality, vehicle behaviour, and usefulness within a given environment. This distinction represents the main contribution of the study and has direct implications for how pilot programmes are designed, communicated, and evaluated in their early stages. Finally, addressing RQ1, the comparison between users and non-users demonstrates clear differences in awareness, exposure, and evaluation of the service, confirming that direct experience plays a central role in shaping perceptions and adoption pathways. The findings provide a coherent response to all four research questions, demonstrating how exposure, experience, and context interact to shape early-stage acceptance of autonomous shuttle services.

Policy recommendations

The behavioural and attitudinal data collected as part of the trial is the basis for a number of prescriptive recommendations for future AS use at the NEC and other UK sites. These practical steps are an attempt to distil the descriptive survey results, PCA outputs and qualitative findings into more digestible form for policymakers, operators and site managers (see Table 7).

Limitations

The results of the current study should be viewed in the context of these limitations, and they also help to delimit the scope of the conclusions which can be drawn. First, the AS operated in a defined campus setting, at low speeds, on a predefined route, in light traffic, and with a safety operator on board. The environment was thus appropriate for the objective of assessing first-time exposure, user experience and perceived safety within a specific operational design domain. At the same time, however, the conditions differ from on-street deployments in mixed-traffic urban contexts and interactions with other road users, including conventional vehicles, cyclists and pedestrians, which are more complex and less predictable. The results of the study can therefore be expected to generalise to on-street trials under similar conditions, but not without further validation. The study does, however, provide evidence for acceptance and experience in campus-style or site-specific operations, such as in business parks, hospitals, airports or exhibition centres. Although the study incorporated several inferential techniques, it was not designed as a fully explanatory multivariate modelling exercise, and future research with larger samples could examine more complex relationships among experience, trust, safety perception, and adoption intention. In addition, the operational characteristics of the trial environment are likely to have influenced user perceptions and responses. The shuttle operated at low speeds, on a fixed and predictable route, within a semi-controlled campus setting, and under continuous supervision by a safety operator. These conditions may have contributed to higher perceived safety, comfort, and acceptance than might be observed in more complex, mixed-traffic urban environments. As such, the findings should be interpreted as reflecting early-stage exposure within a constrained operational design domain, rather than fully representative of broader real-world deployment conditions.

Second, the sample was derived through voluntary, in-situ intercept recruitment at the NEC campus, which introduces the potential for selection bias. Respondents were primarily visitors and on-site workers who were already present within the trial environment, and therefore more likely to be exposed to, or curious about, the shuttle service. This may have led to an overrepresentation of individuals with a baseline openness to innovation or a contextual need for mobility within the site. In addition, the sample was male-dominated and not fully representative of the wider population of Birmingham or the UK. While this reflects the observed composition of NEC users during the trial period, it limits the generalisability of the findings to broader and more diverse population groups. Third, the principal component analysis was only performed with users of the autonomous shuttle (n = 116), as the questions concerning service experience were not answered by the non-users. This is again appropriate from a methodological perspective, as the latent variable analysis was used to identify how service attributes were cognitively structured by those with direct experience of the service. The experience dimension identified by the PCA thus represents a post-exposure evaluation and does not speak to perceptions in the absence of direct experience with the shuttle. Fourth, the reported intention to use the shuttle on a regular basis is a self-reported behavioural intention and not an indicator of actual use. As such, it represents a first reaction following limited exposure during a trial and does not necessarily predict continued or habitual use, should it be made available after the end of the trial. This does not invalidate the measure, but it does require caution when generalising to long-term use, in particular if service routes are expanded, or operational conditions change. Longitudinal data or measures of revealed preference would be needed to see how this changes as services mature and are used in practice.

Finally, Open-ended responses were limited in number and unevenly distributed. They are therefore used to provide contextual support to the quantitative findings rather than to support standalone generalisations. The responses were provided in relatively low numbers and do not allow for general conclusions, but were included in the current report to provide contextual depth and understanding of the quantitative results. The identified limitations delimit the conclusions that can be drawn from this study, but do not invalidate the research. The results should be viewed as being well-suited to shed light on the use of AS in campus-based and site-specific deployment, and they call for further trials in more complex, real-world contexts to see if these perceptions hold more generally. All in all, these contextual and sample-related factors indicate that the findings are most applicable to early-stage, site-specific deployments and should be extrapolated to wider urban contexts with appropriate caution.

In terms of extrapolation, the findings of this study are most directly applicable to early-stage autonomous shuttle deployments operating within constrained environments, such as campuses, business parks, airports, and similar semi-controlled settings. The consistency of the results with international trial evidence suggests that key behavioural patterns, particularly the positive effect of direct exposure and the importance of perceived safety and service quality, are likely to generalise across comparable operational contexts. However, extrapolation to fully mixed-traffic urban environments should be approached with caution. In more complex settings, higher speeds, increased interaction with other road users, and reduced presence of safety operators may significantly alter user perceptions, particularly in relation to safety and trust.

Nevertheless, the core behavioural mechanism identified in this study, namely, the transition from abstract concerns to experience-based evaluation following direct exposure, is likely to remain relevant across different deployment contexts. This suggests that trial-based engagement and staged introduction strategies may be broadly transferable, even if specific service attributes and operational performance factors vary between environments.

Future directions

Future studies could assess AS performance and acceptance in more difficult operating conditions, such as mixed-traffic routes and under adverse weather, using longitudinal designs to track changes in perceptions as exposure is increased and human drivers are successively removed. There is also a need for further research to assess accessibility requirements for passengers with mobility impairments, as well as modelling of the behavioural impacts of improved service frequency, routing, and integration with regional public transport. Analysing stated intention data alongside revealed behaviour in longer-term trials would also provide stronger evidence of the potential for real-world take-up.

This trial-phase evaluation demonstrates that AS can provide a reliably positive experience in a constrained campus environment in the UK. User feedback on our service was high on comfort, safety and overall ride quality, and PCA results validated that these variables combined into a one-dimensional construct of passenger experience. Direct experience was the key determinant in shaping acceptance. Riders who have used the AS have a high intention of regular use, which was matched with clear use cases such as convenience, reduced walking effort, enjoyment, and perceived reliability. On the other hand, non-riders were primarily influenced by informational and safety barriers, with many claiming to be unaware of the service or not feeling they had enough knowledge to understand how the shuttle operated. This feedback points to a pattern which has been the case for multiple prior autonomous mobility pilots, where experience reduces uncertainty while non-exposure heightens perceived risk. Overall, the results indicate that well-planned pilots with visible communication and accessible services have the potential to meaningfully enhance public preparedness for autonomous mobility. The NEC trial provides an important case study in the UK context for how real-world experience with this new form of mobility can shape trust, comfort, and intention to use, and how attention to these and the operational, informational, and accessibility challenges they have exposed in future deployments can help achieve autonomous services that are dependable, inclusive, and an integrated part of daily urban mobility. These findings are broadly consistent with international autonomous shuttle trials, reinforcing the generalisability of exposure-driven acceptance while highlighting the importance of contextual factors in shaping user experience and adoption.