Skip to content
Frontiers
Search

Original Research

Frontiers in Future Transportation ·

Volume 7 - 2026 | https://doi.org/10.3389/ffutr.2026.1687363

Driven by trust? Social norms increase trust in shared autonomous public transport

  • Sander Vassanyi1*corresponding author
  • Ole Aasvik1,2
  • Pål Ulleberg1
  1. 1Department of Psychology, University of Oslo, Oslo, Norway
  2. 2Institute of Transport Economics, Oslo, Norway
Article metrics0 viewsView details
PDFCiteMetrics

Abstract

Public adoption of novel transport technology depends on the willingness of potential users to engage with unfamiliar systems. Trust is a key determinant of this willingness, yet little is known about how trust in shared autonomous vehicles (SAVs) can be shaped. The current pre-registered study investigates to what degree trust predicts the intention to use SAVs and whether static and dynamic social norms can influence the degree of trust in SAVs. Static norms convey the current consensus in a population, while dynamic norms highlight that attitudes or behaviors are changing over time. An online experimental survey was conducted in which participants (n = 1,032; n = 630 after filtering) were randomly assigned to one of three conditions: control, static norm, or dynamic norm. Each condition presented distinct normative statements embedded in informative vignettes about SAVs. The moderating effect of propensity to trust (interpersonal and automation-specific) on the relationship between normative framing and subjective trust was also examined. One-way ANOVA and regression analyses revealed that both static and dynamic normative framings led to significantly higher subjective trust in SAVs compared to the control group. No significant difference was found between the two normative conditions. Propensity to trust (both interpersonal and automation-specific) had direct positive effects on subjective trust but did not moderate the effect of normative framing. Safety evaluation emerged as the strongest predictor of subjective trust. Subjective trust, in turn, was strongly associated with intention to use SAVs. These findings indicate that norm-based interventions providing social cues about others’ trust can effectively shape public attitudes toward SAVs, even in the absence of firsthand experience. However, safety perceptions remain paramount, suggesting that normative strategies should complement, rather than substitute for, efforts to demonstrate and communicate the safety of autonomous public transport.

Keywords

Highlights

  • Trust in SAVs is crucial: Social normative influences significantly affect public trust in shared autonomous vehicles (SAVs), a key factor for adoption.
  • Static and dynamic norms: Both static and dynamic social norm interventions were found to increase subjective trust in SAVs although dynamic norms did not outperform static norms significantly.
  • Safety is paramount: Evaluation of safety was the most substantial predictor of trust in SAVs, impacting user intention to use these vehicles.
  • Propensity to trust matters: Individuals with higher propensity to trust, especially in automation, displayed greater subjective trust in SAVs.
  • Interventions influence adoption: Norm-based interventions that provide social cues about trust in SAVs can effectively shape public attitudes towards adopting autonomous public transport.

1 Introduction

The rapid urbanization and growth of urban populations in modern societies demand innovative solutions to tackle congestion and greenhouse gas emissions from vehicles. Autonomous vehicles offer a promising solution, potentially reducing accidents, enabling climate-neutral transport, and providing mobility to marginalized groups (Hult et al., 2016; Nordhoff, Kyriakidis, van Arem, et al., 2019; Xu et al., 2018). However, the proliferation of private autonomous vehicles might exacerbate congestion without a corresponding improvement in communication and coordination among these vehicles (Ruter, 2019).

In this context, shared autonomous public transport, particularly small, shuttle-like, on-demand buses, emerge as a crucial alternative. These systems promise to enhance mobility, especially for first and last-mile connections, while potentially reducing traffic accidents and offering more personalized mobility services. However, there is a still a need for knowledge about how to best implement these systems. Public adoption of shared autonomous vehicles (SAVs) is still understudied, particularly using solid psychological frameworks (Greifenstein, 2024).

This study aims to explore the formation of trust in SAVs and the role social norms play in shaping public acceptance. Trust is a pivotal factor; without it, the willingness to use new technological solutions, such as SAVs, is significantly hindered (Nordhoff et al., 2023). Through an examination of whether static and normative social norms and their influence on trust, this study seeks to uncover the psychological processes underlying trust formation in the context of SAVs. It investigates how trust affects the intention to use SAVs and whether interventions based on static and dynamic norms can enhance trust levels. Static norms are current and consistent expectations inferred through others behavior and attitudes. Dynamic norms on the other hand, are defined by the belief that some notion, attitude or behavior is currently changing, and hence promote conformity through expected future change (Sparkman and Walton, 2019). By focusing on these aspects, the study contributes to understanding the psychological underpinnings of adopting autonomous public transport, aiming to provide insights for successful integration and adoption strategies.

Furthermore, this study contributes to the literature by addressing the role of normative influence on trust formation in the context of shared autonomous vehicles (SAVs). Specifically, it examines how static and dynamic social norms shape subjective trust, a key determinant of adoption intention (Nordhoff et al., 2023). Furthermore, it explores the moderating role of individual propensity to trust, providing new insights into the psychological processes underpinning trust in emerging transportation technologies. By filling these gaps, this research may offer actionable recommendations for policymakers and industry stakeholders aiming to promote public acceptance of SAVs. First, we will review relevant literature with our associated hypotheses embedded.

1.1 Trust and propensity to trust

Trust has a large impact on the interactions individuals have with each other, and with the world around us. It is a complex construct and is plagued by the same terminological and psychometric issues as other composite and latent psychological terms (Cronbach and Meehl, 1955). Interpersonal trust and trust in automation and technology have been researched for decades. More often than not, the latter has its’ roots in the trust we exhibit in human to human interaction (Lee and See, 2004). Trust has been defined as an attitude, a belief, a disposition, a personality trait, and as a mental state (Hoff and Bashir, 2015). No single interdisciplinary definition has been established (Harrison McKnight and Chervany, 2001), and the construct is largely dependent on the context in which it is being examined.

Somewhat lesser investigated than trust itself, is the base rate of which the individual exhibit trusting or distrusting attitudes and behavior: their propensity to trust (Frazier et al., 2013; Schoorman et al., 2007). Propensity to trust is viewed as dispositional trust, or the tendency to trust, when information and other influencing factors are minimal. To understand context specific trust, it is essential to investigate the fundamental structures, such as propensity to trust, because it serves as an underlying baseline that shapes how people perceive and react to trust-related cues in various situations individually, ultimately influencing their trust decisions in specific contexts based on other determinants. When information about something is scarce, we seek other ways to inform an evaluation. Because autonomous vehicles is not yet a common technology, people may rely more on others’ testimony and beliefs in their own attributions (Hoff and Bashir, 2015). Mayer et al. (1995) propose that propensity to interpersonal trust, or the general willingness to trust others, affects or moderates trust attribution. Trust propensity functions as a baseline for trusting different trustees. Propensity to trust is believed to be stable (Mayer et al., 1995; Rotter, 1967) and function like a personal trait. This trait functions as a substitute for lack of information and moderates the degree of negative or positive influence from new information. For example, a trustor that meets someone for the first time will exhibit trust appropriate to their own propensity in addition to any available information. If the trustor has a low propensity to trust, the more lack of information contributes to distrust.

Building on this understanding, we hypothesize that individuals with higher propensity to interpersonal trust will exhibit a greater effect of normative framing on subjective trust in SAVs (H1). This is grounded in the idea that a higher baseline tendency to trust others facilitates the incorporation of social cues, such as normative statements, into evaluations of unfamiliar technologies. This hypothesis aligns with prior findings that individuals with a strong interpersonal trust disposition are more likely to rely on external influences, such as social norms, when forming trust judgments (Hoff and Bashir, 2015).

In addition to moderating the impact of social norms, we hypothesize that higher propensity to interpersonal trust is directly and positively related to subjective trust in SAVs (H2). This is based on the notion that dispositional trust often generalizes across contexts, enabling individuals with a greater tendency to trust others to extend this trust to automated systems as well. This hypothesis aligns with prior research suggesting that interpersonal trust can influence attitudes toward unfamiliar technologies, particularly when information is limited (Mayer et al., 1995; Hoff and Bashir, 2015).

Lastly, we hypothesize that propensity to interpersonal trust and propensity to trust automation are positively correlated (H3). While these constructs represent distinct dimensions of trust, prior studies suggest that individuals with high interpersonal trust are more likely to exhibit trust in automated systems, reflecting an underlying general trust disposition (Frazier et al., 2013; Jessup et al., 2019). Investigating this relationship contributes to a deeper understanding of the interplay between personal dispositions and context-specific trust.

1.2 Trust in automation

Automation is an integral part of modern society, performing tasks more efficiently, safely, and reliably than humans in many cases (Fröhlich et al., 2019). Shared autonomous vehicles (SAVs) exemplify this trend, where users often lack the technical knowledge to evaluate the system’s workings. Despite all the promises to improve transport infrastructure, considerable opposition remains towards AVs (Jin and Roy, 2024; Zhanhang and Jingyang, 2024). Much of the public opposition seem to revolve around the capability and trustworthiness of the novel technology. Other important factors for determining trust in AVs are thought to be competence/reliability, understandability/predictability, and intention of developers (Körber, 2019). These components have recently been empirically verified as important for users perception of trust in AVs (Aasvik, 2025). These components are hypothesized to be malleable, although little research has been done to explore how trust can be impacted. In fact, although trust regularly achieves large coefficients in predicting AV acceptance (Mengel et al., 2025; Zhang et al., 2023), the most commonly used frameworks lack a clear integration of trust as a predictor. For example, one of the most used frameworks, the Unified Theory of Acceptance and Use of Technology (UTAUT), does not mention trust (Venkatesh et al., 2016). Trust is not forefront in either the multi-level model on automated vehicle acceptance (MAVA), nor the Technology Acceptance Model (TAM) (Nordhoff et al., 2019; Venkatesh and Bala, 2008). Improving our understanding of the conceptualization and improvement of trust in AVs thus seems pertinent.

Trust in AVs depends on external factors, such as social norms and conformity, which shape perceptions (Hoff and Bashir, 2015). Trust formation occurs through analytical (evidence-based), analogical (heuristic-based), and affective (emotion-based) processes (Lee and See, 2004). Analytical trust arises from evaluating available information, whereas analogical trust relies on social cues, past experiences, and heuristic associations. For instance, evaluated SAV pilots show that, while trust remains critical for user intentions, it can also dampen enthusiasm when the technical capabilities of the vehicles are sub-optimal (Aasvik et al., 2024). Trust has been found to be higher when participants travel facing forwards rather than backwards and lower levels of trust when travelling at higher speeds (Paddeu et al., 2020). Trust is fragile and can erode following negative events, fostering algorithm aversion (Jussup et al., 2019). Emotions further influence trust by providing subconscious feedback, guiding decisions about whether to trust automated systems. Trust in the context of this study refers specifically to subjective trust in shared autonomous vehicles (SAVs), encompassing both trust in the technology’s reliability and the service’s ability to meet user expectations. This includes perceptions of safety, functionality, and the ability to operate autonomously in complex environments. Drawing on prior research (Hoff and Bashir, 2015; Nordhoff et al., 2023), we conceptualize trust as a multidimensional construct influenced by individual dispositions (e.g., propensity to trust), contextual factors (e.g., safety evaluations), and social cues (e.g., normative influence). By focusing on subjective trust, the study aims to uncover how these factors interplay to shape user intention to adopt SAVs.

Extending this perspective, we hypothesize that individuals with higher propensity to trust automation will exhibit a greater effect of normative framing on subjective trust in SAVs (H4). As trust in AVs is shaped by both dispositional and contextual factors, those with a greater baseline tendency to trust automated systems are expected to be more receptive to normative influences, particularly in novel contexts. Furthermore, we predict that individuals with higher propensity to trust automation will also demonstrate increased subjective trust in SAVs overall, regardless of normative framing (H5). This follows from prior research showing that a strong baseline trust in automated systems often translates into greater confidence in new and unfamiliar technologies (Hoff and Bashir, 2015; Lee and See, 2004).

Moreover, we hypothesize that safety evaluations are positively correlated with subjective trust in SAVs (H6). Safety perceptions are a critical component of trust in automated systems, shaping both emotional and cognitive assessments of reliability. Prior research has consistently identified safety as a central determinant of trust in autonomous vehicles, influencing users’ willingness to adopt such technologies (Nordhoff et al., 2019; Hoff and Bashir, 2015).

1.3 Social normative impact on trusting intention and behavior

Social norms are a cornerstone component in several behavioral models in psychology (Ajzen, 1985). Seeing significant others use and endorse autonomous vehicles have been found to increase personal decision making regarding adoption of the technology, and people tend to view the normative consensus as a guide to the perceived benefits of autonomous vehicles (Acheampong and Cugurullo, 2019). While some fail to replicate the effect of social influence on intention to use SAVs (Goldbach et al., 2022), most studies find this to be the case (Greifenstein, 2024; Yuen et al., 2020). Even the perceived ease of use of autonomous vehicles is believed to be affected by how normative the use is in the population. As normative information is theorized to be of importance in analogic trust formation (Hoff and Bashir, 2015), social norms could be a major influential factor regarding trust too. This is especially true when little information is available, and the testimony and experience of others are a primary source of information.

Social norms and their ability to affect behavior is well documented (Bicchieri et al., 2018; Sherif, 1936). Static norms represent the current consensus in a population, conveying stability and reliability. We hypothesize that static norms will positively influence subjective trust in SAVs compared to no normative statement, as they provide clear and consistent information about others’ trust (H7). Dynamic norms, on the other hand, highlight behavioral change over time, creating a sense of momentum and anticipated conformity (Sparkman and Walton, 2019). This mechanism is expected to amplify trust further by portraying public trust in SAVs as not only established but also growing, which could make individuals more likely to align with evolving group behaviors. Accordingly, we hypothesize that dynamic norms will have a greater positive effect on subjective trust in SAVs than static norms (H8). Prior research on climate-positive behaviors demonstrates that dynamic norms not only influence immediate attitudes but also foster greater long-term engagement compared to static norms (Sparkman and Walton, 2019). In the context of SAVs, dynamic norms may similarly evoke stronger trust by portraying increasing public acceptance as an ongoing trend, potentially reassuring individuals of the technology’s reliability and safety. Additionally, overcoming personal psychological barriers to change can be assisted through dynamic norm intervention (Sparkman and Walton, 2019). Given that it can impact sustainable behavior, could the same interventional strategy be applied to increase trust in shared autonomous vehicles? The current study seeks to explore this, however, as discussed, trust attribution is heavily reliant on the context in which it takes place.

Finally, we therefore hypothesize that subjective trust in SAVs will positively predict intention to use SAVs (H9). Trust is a well-established antecedent of behavioral intention in technology adoption models, including the Theory of Planned Behavior (Ajzen, 1985). By examining this relationship, the study aims to explore whether subjective trust translates into practical adoption intentions, a critical step for the successful implementation of SAVs. In total, the current study seeks to investigate (1) to what degree trust impacts intention to use shared autonomous vehicles and (2) whether static and dynamic norms can be applied to impact degree of trust in shared autonomous vehicles. To examine this, an experimental online survey was utilized, with one control-group and two intervention groups.

1.4 Summary of hypotheses

The reviewed literature identifies trust as a key determinant of SAV adoption, yet little empirical work has examined how trust can be actively shaped. Two overarching research questions guide this study: (1) to what degree does trust predict intention to use SAVs, and (2) can static and dynamic social norms influence the degree of trust in SAVs? The following hypotheses address these questions by targeting three domains: the direct and moderating role of dispositional trust (H1–H5), the relationship between safety perceptions and trust (H6), and the effect of normative framing on trust and its downstream link to usage intention (H7–H9). Based on the reviewed literature, we propose the following hypotheses:

H1 individuals with higher propensity to interpersonal trust will exhibit a greater effect of normative framing on subjective trust in SAVs.

H2 Higher propensity to interpersonal trust is directly and positively related to subjective trust in SAVs.

H3 Propensity to interpersonal trust and propensity to trust automation are positively correlated.

H4 Individuals with higher propensity to trust automation will exhibit a greater effect of normative framing on subjective trust in SAVs.

H5 Individuals with higher propensity to trust automation will also demonstrate increased subjective trust in SAVs overall, regardless of normative framing.

H6 Safety evaluations are positively correlated with subjective trust in SAVs.

H7 Static norms will positively influence subjective trust in SAVs compared to no normative statement, as they provide clear and consistent information about others’ trust.

H8 Dynamic norms will have a greater positive effect on subjective trust in SAVs than static norms.

H9 Subjective trust in SAVs will positively predict intention to use SAVs.

We theorize that normative messages about SAVs increase subjective trust relative to a neutral control (static and dynamic norms). Two dispositional factors: Propensity to interpersonal trust and propensity to trust automation, raise trust directly and were expected to strengthen message effects. Safety evaluations should covary with, and positively relate to, trust. Trust should, in turn, increase intention to use SAVs.

2 Materials and methods

2.1 Pre-registration and power analysis

The study was pre-registered using asPredicted.org on 8 November 2022 to promote transparency and minimize potential biases in the research process. The pre-registration contains the hypotheses of the study, key dependent variables and their measures, procedure of the experiment, analyses, and outlier-handling. The full pre-registration can be found in Supplementary Appendix B. There were some minor divergencies that should be noted, these can also be found in Supplementary Appendix B.

To determine the necessary sample size for the moderation analysis, an a-priori power analysis was conducted using GPower (Faul et al., 2009). The analysis was based on a small effect size (f* *2* = 0.02), a significance level of 0.05, and a desired power of 0.80. The power analysis indicated that a minimum sample size of 395 was required to adequately detect the expected effect. To address the potential impact of dropouts and nonresponse to questions, the aim was to recruit a sample of 600 respondents (see Supplementary Appendix C for plot).

Lastly, the study in its entirety was evaluated and approved by NSD - Norwegian Centre for Research Data, as well as the Department of Psychology’s internal research ethics committee (see Supplementary Appendix D).

2.2 Recruitment

Recruitment was conducted through three main channels: social media, posters, and a mailing list with potential respondents from a previous study. Social media yielded 159 respondents, email 835 and posters 38, for a total of 1,032 responses. Most respondents from social media were recruited through student-groups and peers including word-of-mouth. The respondents were not directly compensated for participation, however, all participants could choose to enter a raffle for three gift cards, should they wish to. This was communicated through all recruitment channels as an incentive to take part in the study. Email recruitment was conducted in two waves, yielding a response rate of 60.5% and 53.6% respectively. A total of 1,460 emails were sent. Social media recruitment was conducted via posts in Facebook-groups as well as LinkedIn. Posters were displayed at different locations, including on a university campus.

2.3 Procedure

First, participants recorded three sociodemographic variables: age, gender, and education. Then, three independent measures were presented: Propensity to trust others, technical competence, and propensity to trust automation. After this, the participant was randomly assigned one of three experimental conditions: Control group, static norm group, or dynamic norm group. The participants in the intervention groups also received a rumination task right after, as well as a reminder of the presented vignette. Then, two dependent variables followed: Subjective trust in shared autonomous vehicles and intention to use shared autonomous vehicles, further followed by a safety evaluation of shared autonomous vehicles. Additionally, a measure of automated vehicle acceptance (not used in the present study) were presented, then lastly the participants were asked whether they had been exposed to pilot projects testing SAV’s in real traffic.

2.4 Measures

2.4.1 Gender, age, and education

The question about gender included four options: “Female”, “Male”, “Other”, and “Do not wish to answer”. Age was collected in 10-year intervals to ensure anonymity. There were seven categories, with the first being “18–29” and the last being “80+”. Education was also collected as a categorical variable, asking the participant to note their highest completed educational level. The four options were “elementary”, “high-school”, “1–3 years university or higher education”, and “over 3 years of university or higher education”.

2.4.2 Propensity to trust

The scale used in the current study consists of five items adapted from Frazier et al. (2013), measured on a seven-point Likert scale where one and seven are denoted by strongly disagree and strongly agree respectively. The items with the highest loadings in their study are retained and used in the current study as a short form measure of propensity to trust (e.g., “I usually trust people until they give me a reason not to trust them” and “Trusting another person is not difficult for me.”; α = 0.84). Additionally, a fifth reversed item was added (“I have little trust in other people’s promises.”) to decrease the likelihood of pattern responses and further strengthen the measure. Additionally, the items used were sourced from validated previous work by Lee and Turban (2001) and McKnight et al. (2002).

2.4.3 Propensity to trust automation

The items follow the same structure as the previous measure, but are adjusted to target trusting behavior tendencies towards automated systems and are as such context specific in this regard. The scale had a seven-point Likert scale where one and seven are denoted by strongly disagree and strongly agree respectively. The respondents were instructed to answer with general automation-systems in mind, like computers or elevators (e.g., “I usually trust automatic systems until they fail” and “It is not hard for me to trust automatic systems”; α = 0.90). This was to avoid priming of very advanced technology, since the goal was to measure propensity in relation to more “everyday” automation. However, this might also introduce some priming as to what kinds of automation they evaluate. Avoiding the problems of contextual differences is difficult, but aligning the items with the propensity to interpersonal trust may yield a more proper general tendency measure. Still, this scale needs further testing. The scale had an Cronbach’s alpha value of 0.90, indicating a solid scale, with few redundant items (Tavakol and Dennick, 2011).

2.4.4 Technical competence

Technical competence, also called tech-savviness, is a self-developed measure which aims to capture the respondent’s subjective ability to handle, and their interest in, technology. The measure is included as a control variable when assessing trust. In this study, three items were developed to assess the construct (“I am among the first to hear about new technology”, “Friends and family come to me for help with technology”, and “I am above average interested in technology”; α = 0.89). These items function as an aggregate to how technology-savvy the individual sees themselves and how comfortable they are with interacting with technology. The scale had a seven-point Likert scale where one and seven are denoted by strongly disagree and strongly agree respectively.

2.4.5 Safety evaluation of SAVs

Safety Evaluation included five items which aimed to measure the subjective degree of how safe a shared autonomous shuttle would be in use, both from a personal perspective (e.g., “A bus like this would be safe to use”) and a societal perspective (e.g., “A bus like this would increase traffic safety”; α = 0.95). The scale is developed for this survey, with inspiration from Nordhoff et al. (2019) using a seven-point Likert scale where one and seven are denoted by totally disagree and completely agree respectively. The variable is designed to be a control variable, as safety is assumed to be of importance in trust evaluation of SAVs. The scale had a Cronbach’s alpha score of 0.95 which may entail that some items are redundant (Tavakol and Dennick, 2011). The measure may be concrete enough to not demand a five-item scale.

2.4.6 Subjective trust in shared autonomous vehicles

The first dependent variable was comprised of three items, specifically aimed at assessing the extent to which respondents would have trust in shared autonomous vehicles. The items were measured on a seven-point Likert scale where one and seven are denoted by totally disagree and completely agree respectively. The scale was developed for this survey, with inspiration from Choi and Ji (2015): “Self-driving buses are reliable”, “Self-driving buses will mostly do what is expected” and “All in all, I can trust self-driving buses”; α = 0.95. It represents the willingness to trust and be vulnerable, and not the actual trust-behavior, hence a proxy for probable behavior. The argument can be made that this is a form of institutional-based trust, more specifically a form of structural insurance (Harrison McKnight and Chervany, 2001), since it represents an expectation that the government and service operator have conducted the necessary testing. Still, the variable should capture the essence of trusting the technology in this context. The scale had a somewhat high Cronbach’s alpha score, possibly indicating some redundancy.

2.4.7 Intention to use shared autonomous vehicles

The second dependent variable was also comprised of three items. This measure was developed for this study, with the belief that intention is a viable predictor of behavior (Ajzen, 1985). It represents the trust-behavior as best as can be done without observing real-world behavior. The reason for including a three-item measure is to fully capture intention to use SAVs given that this offer is not available yet. For example, the intention may be masked by the respondent’s belief that this offer would not be applicable in their context or local area, and hence not measure their intention free of contextual boundaries. The items were measured on a seven-point Likert scale where one and seven is denoted by totally disagree and completely agree respectively: “I would be comfortable riding a bus like this”, “If I had the opportunity, I would be fine with riding on a bus like this”, and “I would avoid driving with a bus like this, even if it was the fastest and cheapest alternative (Reversed)”; α = 0.93. Additional notes about the measures can be found in Supplementary Appendix F.

2.5 Manipulation

After the independent measures, each respondent was randomly assigned to one of three conditions: control, static norm, or dynamic norm. All groups were shown an informative vignette (Translated from Norwegian, see full survey in Supplementary Appendix A):

A bus-service is now being developed in Norway, which will consist of small, self-driving busses. The busses drive on normal roads and in normal traffic but will not need a driver. Pilots (tests) with a service like this are now being tested.

In addition, the control group was shown a neutral text in the same vignette: “The tests have been done in relevant areas with moderately dense populations. Most of the buses are red.” The static norm group was presented with the information, and additionally: “Recent studies by the Institute of Transport Economics show that more than 7 out of 10 Norwegians trust that self-driving buses work as they should.” Finally, the dynamic group was presented with the above information, as well as:

More people are positive towards a service like this, and recent studies by the Institute of Transport Economics show that more than 7 out of 10 Norwegians trust that self-driving buses work as they should. People are changing their opinion about a self-driving public transport offer.

After one of three vignettes, the respondent was given a rumination prompt, asking what they thought about the amount of people trusting a bus service like this. This was to foster reflection of the normative statement and to make the respondent interact actively with the information. The control group did not receive any rumination prompt. Before continuing to the dependent measures, the participants were shown a reminder vignette, identical to the one they were presented, which they could navigate back to if they wanted.

The intervention is hypothesized to affect subjective trust in SAVs, whilst the two types of propensity moderate the effect. The other variables act as controls. Subjective trust in SAVs is hypothesized to predict intention to use SAVs.

All groups were displayed one of two comprehension checks which had the goal of establishing whether the respondent had properly read and contemplated the vignette. The intervention groups were asked “How many Norwegians trust that self-driving busses work as they should?”, with the alternatives being over 3 out of 10, over 5 out of 10 or over 7 out of 10. The control group were presented with the following question: “What color does most of the busses in the testing of self-driving busses in Norway have?”, with the alternatives red, green, or yellow.

Last, the respondents were asked to estimate their exposure to self-driving busses in Norway: “Did you know of trials with self-driving buses in Norway before you participated in this survey?”. The scale was on a seven-point Likert, with “1” representing “None”, “2” representing “A little”, and then a gradually increase until “7”, which represented “A lot”.

2.6 Response rate and filtering

A total number of 1,032 responses were collected, and 958 completed the survey. Removing all responses which had a response time of less than 3 min, in accordance with the pre-registration, left 920 total responses. When controlling for failed answers to the comprehension checks, 630 valid responses remained and were retained for further analysis, well above the required 395 assessed with the a priori power analysis. After filtering, the control group had 230 participants, the static norm group 198, and the dynamic norm group 202. Table 1 presents the sociodemographic characteristics of participants across the three experimental conditions.

TABLE 1. Sociodemographic characteristics of participants for each intervention group from retained responses (N = 630).

VariableControl group (N = 230)Static norm (N = 198)Dynamic norm (N = 202)Full sample (N = 630)
n%n%n%n%
Gender
Female11047.88140.98240.627343.3
Male11650.411759.112059.435356.0
NA31.300.000.020.3
Other10.400.000.010.2
Age
18–292912.62613.12411.97912.5
30–39208.72412.1188.9629.8
40–493213.93115.72813.99114.4
50–595423.53819.24622.813821.9
60–696126.53819.25225.715124.0
70–792812.23417.23316.39515.1
80+62.673.510.5142.2
Educational level
Elementary62.610.542.0111.7
High school4318.72713.62311.49314.8
1–3 years higher education6026.15427.35426.716826.7
Over 3 years higher education12152.611558.112059.435657.0
NA00.010.510.520.3

3 Results

3.1 Sociodemographic descriptive statistics

Differences in demographic distributions across the groups were investigated using Chi-Square tests, including age given that age was collected as intervals. The results showed that there were no significant differences among the three groups, suggesting that the randomization process was successful.

3.2 Scale analysis: distribution and normality insights

Descriptive statistics for all measures, including Cronbach’s alpha, means, standard deviations, skewness, kurtosis, and the Shapiro-Wilk test to explore normality can be found in Table 2.

TABLE 2. Descriptive statistics for independent, dependent, and control variables (N = 630).

StatisticPropensity to trustPropensity to trust automationTechnical competenceSafety evaluation of SAVsSubjective trust in SAVsIntention to use SAVs
N630630630623630630
Missing a000700
Mean5.235.184.064.544.594.99
Median5.205.204.004.805.005.33
Standard deviation0.9941.071.571.441.371.57
Cronbach’s alpha0.840.900.890.950.950.93
Minimum1.201.001.001.001.001.00
Maximum7.007.007.007.007.007.00
Skewness−0.588−0.657−0.0569−0.526−0.510−0.753
Kurtosis0.6060.615−0.897−0.362−0.160−0.168

Safety evaluation did not require an answer to proceed in the survey, which lead to some missing data. All scales had a range of 1–7.

The average scores for all the scales leaned towards the higher end, with participants exhibiting a strong inclination to trust. Propensity to trust not only had the highest average but also the smallest variation. On the other hand, tech competence displayed the lowest average and a considerable variation in scores. The two dependent variables, subjective trust and intention to use shared autonomous vehicles, showed similar averages and variations in scores. Overall, people were positive to use a service like this as implied by the score of intention to use SAVs.

Propensity to trust, propensity to trust automation, subjective trust in SAVs, intention to use SAVs, and safety evaluation all exhibited a slight leftward skew. However, the skewness value for each scale remained within the conventional range deemed unproblematic, as per Field (2013). This suggests that participants scored higher on all these than a normal distribution would assume.

3.3 Assessing the effect of social norm intervention

One-way ANOVA was applied to test the effect of the manipulation predicting subjective trust in shared autonomous vehicles, testing hypothesis seven and eight. The assumptions for conducting an ANOVA were met: Levene’s test for equality of variances was not significant (p = 0.188), the scores on Subjective Trust in SAVs were approximately normally distributed, and no extreme outliers were detected. The results revealed a significant difference between the three groups on subjective trust in SAVs (F (2, 627) = 10.4, p < 0.001). However, the effect size was small (η *2* = 0.03), suggesting that other factors are likely to be important in determining subjective trust in shared autonomous vehicles. Tukey’s post hoc test was used to compare the mean differences between the groups, as presented in Table 3.

TABLE 3. Multiple comparisons of mean scores on subjective trust in SAVs for the three conditions (Tukey’s HSD test).

GroupNMeanSD95% CIPairwise comparisons
1 vs. 21 vs. 32 vs. 3
1. Control group2304.261.43[4.09, 4.44]−0.53**−0.50**0.03
2. Dynamic group2024.791.30[4.69, 4.98]
3. Static group1984.761.30[4.57, 4.95]

p < 0.001.

The post hoc test shows a systematically lower mean for the control group compared to both intervention groups of half a point, suggesting an effect of the experimental manipulation. The higher level of trust in the static group supports hypothesis seven, whilst the lack of significant difference between the intervention groups means that hypothesis eight is not supported.

An additional ANCOVA was conducted in which technical competence, age, gender, education, and prior exposure to SAVs were included as covariates. The differences between groups remained significant, F (2, 622) = 7.51, p < 0.001. Tukey’s post hoc comparisons indicated that Group 1 differed significantly from both Group 2 (mean difference = 0.44, p < 0.001), and from Group 3 (mean difference = 0.40, p < 0.005). Group 2 and Group 3 did not differ in mean score (mean difference = 0.04, p =0.942).

3.4 Correlations between the study variables

To examine the relationships between the independent, dependent, and control variables, as well as investigate hypothesis three, six and nine Pearson’s correlation coefficients were calculated. These are displayed in Table 4.

TABLE 4. Correlations for independent, dependent, and control variables.

Variable123456
1. Propensity to trust​​​​​
2. Technical competence−0.01​​​​
3. Prop. to trust automation0.29***0.29***​​​
4. Subjective trust in SAVs0.26***0.25***0.51***​​
5. Intention to use SAVs0.22***0.30***0.47***0.82***​
6. Safety evaluation0.26***0.26***0.48***0.88***0.87***

indicates p < 0.001.

Propensity to trust was significantly correlated with all other scales except technical competence, with small correlations. This included propensity to trust automation, supporting hypothesis three. Technical competence had a moderate positive correlation with intention to use shared autonomous vehicles according to conventional interpretations (Cohen, 1988). Additionally, small positive correlations with propensity to trust automation, subjective trust in shared autonomous vehicles, and safety evaluation were found. Propensity to trust automation had a notably high correlation with subjective trust in shared vehicles, and moderate to high correlations with intention to use SAVs, safety evaluation. Subjective trust in SAVs had large correlations with intention to use SAVs and safety evaluation in addition to propensity to trust automation. This indicates a strong relationship between subjective trust in, and intention to use shared autonomous vehicles, supporting hypothesis nine. Safety evaluation had notably high correlations with subjective trust in SAVs and intention to use SAVs, supporting hypothesis six. The full table with confidence intervals can be found in Supplementary Appendix F.

A confirmatory factor analyses (CFA) with Maximum Likelihood estimation were performed to evaluate the convergent and discriminant validity of the six measured constructs. The results supported convergent validity, and despite high correlations among three of the constructs, the findings also provided evidence for the discriminant validity of the six-factor solution (see Supplementary Appendix G for details)”.

3.5 Examining moderation effects of propensity to trust

To investigate hypotheses one, two, four and five, we performed regression analyses that included interaction terms to test possible moderation effects. Hypothesis one and two, postulating a moderating effect and a direct effect of propensity to trust on subjective trust in SAVs is tested in the first model. The second model tests hypothesis four and five, postulating the same relationships for propensity to trust automation.

Table 5 presents the results of a regression analysis testing whether propensity to trust moderated the effect of the interventions on subjective trust in SAVs. The table is divided into two parts: the top part shows the main effects, while the bottom part presents the interaction effects.

TABLE 5. Moderation analysis of the effect of propensity to trust on the relationship between the normative framing and subjective trust in SAVs (N = 630).

VariablebSEtBETAp
Group (control = 0)
Dynamic0.470.133.690.16<0.001***
Static0.500.133.960.17<0.001***
Propensity to trust0.330.056.240.23<0.001***
Adjusted R20.087
Dynamic0.420.710.600.140.552
Static0.660.651.000.220.316
Propensity to trust0.340.084.100.24<0.001***
Dynamic × Prop. to trust0.0080.130.060.020.950
Static × Prop. to trust−0.030.12−0.24−0.050.812
Adjusted R20.081

indicates p < 0.001.

The main effects were all significant, signaling a direct effect of propensity to trust on subjective trust in SAVs. Hence, hypothesis two was supported. Interaction terms had no effect on subjective trust, thus indicating no support for a moderation effect of propensity to trust (hypothesis one). The overall model without the interaction term explained 9% of variance.

Table 6 presents the results of the second regression analysis examining the moderation effects of propensity to trust automation on subjective trust in SAVs based on intervention group. Like model 1, the top part of the table presents the main effects, whilst the latter the interaction terms.

TABLE 6. Moderation estimates of propensity to trust automation predicting subjective trust in SAVs based on normative framing (model 2, N = 630).

VariablebSEtBETAp
Group (control = 0)
Dynamic0.410.113.700.14<0.001***
Static0.380.113.460.13<0.001***
Propensity to trust automation0.670.0415.700.52<0.001***
Adjusted R20.302
Dynamic0.720.561.290.250.196
Static0.110.550.200.380.842
Propensity to trust automation0.680.089.030.53<0.001***
Dynamic × Prop. to trust auto.−0.060.11−0.56−0.110.575
Static × Prop. to trust auto.0.050.110.500.100.618
Adjusted R20.301

indicates p < 0.001.

As suggested by one of the reviewers, technical competence, age, gender and education were included as covariates in the analyses presented in Tables 5, 6. Their inclusion resulted in only trivial changes to the parameter estimates, and these adjusted results are therefore not shown in Table 5 or Table 6.

All main effects are significant. However, propensity to trust automation has a much higher estimate than propensity to trust. The interaction terms for both groups are included in the same manner, neither being significant. Thus, there is no support for a moderation effect of propensity to trust automation rejecting hypothesis five. The overall model fit explains around 30% of the variation, which is substantially larger than the model presented in Table 5.

4 Discussion

To summarize the findings in relation to the hypotheses proposed, it was found that static norm interventions positively impacted subjective trust in shared autonomous vehicles (H7), while the dynamic norm intervention did not have a larger effect compared to the static norm intervention (H8). Contrary to expectations, propensity to interpersonal trust and propensity to trust automation did not enhance the intervention effect as moderator variables (H1 and H4). However, both types of propensity exhibited direct positive effects on subjective trust in shared autonomous vehicles (H5 and H2). Furthermore, a positive correlation was found between safety evaluations and subjective trust in SAVs (H6). Subjective trust in SAVs predicted intention to use shared autonomous vehicles (H9), and finally, a positive correlation between propensity to interpersonal trust and propensity to trust automation was observed (H3).

4.1 The impact of social norms

This study investigated the level of subjective trust in shared autonomous vehicles when primed with a statically framed normative vignette or a dynamically framed normative vignette compared to no normative statement. The degree of trust among participants who were presented with a normatively loaded statement, regardless of static or dynamic, was higher than those presented with a control text, supporting hypothesis seven. The results from the two ANOVA tests strengthen the hypothesis that static norm intervention would increase subjective trust in shared autonomous vehicles. The dynamic group scored a bit higher on trust, however, it was a non-significant difference and may simply be a coincidence.

Even if there was no significant effect separating the dynamic intervention from the static, the fact that there was such a similar effect of both normative vignettes compared to the control-group strengthens the notion that trust in shared autonomous vehicles is impacted by what others believe about them. They can be viewed as two types of normative influence, and in this way, may bolster the overall effect. This is also in line with the analogical trust formation (Lee and See, 2004) where the testimony of others is an important part in making quick assessments of an actor or trustee. The lack of knowledge and hands-on experience should increase this effect, as the less information available, the more heuristic judgement is necessary to make an evaluation (Bruhn, 2019). Autonomous public transport is a very novel service. Subjective norms and leaning on others experience and their perceived evaluations should therefore be prevalent in the attribution process (Hoff and Bashir, 2015), as displayed in this study.

There could be several reasons for the lack of significant effects between the static and dynamic norm intervention. First, the effect may simply be too weak. Some could argue that even with a high sample size, the effect of simply reading a normative statement may not be sufficient to imprint an actual change in perception of trust. Previous research has failed to find effects on perceptions of SAVs using online surveys (Aasvik, 2025; Aasvik et al., 2025). However, this explanation seems incomplete here, as there are clear effects of both normative interventions as compared to the control group. Second, there is, to the best of the authors knowledge, little-to-no literature examining if a dynamic norm influence has a significant effect on trust attribution, at least in this context. The effects found related to, for example, more environmentally positive behavior (Sparkman and Walton, 2019) may not translate to trusting behavior and attribution, and deserves more thorough examination. The context in which trust is being exhibited can also impact the degree of effect, as people may be more wary of changing norms regarding some attitudes and not others. For example, how “futuristic” or forward leaning one wishes to identify as, may change the degree of conformity to such norms. Other values may directly counteract the effect in this context, such as the fear that more automation may lead to less workplaces, and simply does not have anything to do with the trust in the technology itself. Other factors may also be more important than normative influence, simply because the individual may assign higher priority to these factors in that particular moment. Lastly, while this study hypothesized a stronger effect for dynamic norms based on their ability to signal ongoing change and foster anticipated conformity, the lack of significant differences may be attributed to the framing of the dynamic norm. Unlike prior studies that emphasized quantifiable shifts (e.g., ‘30% of people have changed their behavior’; Sparkman and Walton, 2019), the dynamic norm in this study may have been too subtle to distinguish itself from the static norm. Furthermore, participants’ limited familiarity with SAVs may have reduced the salience of normative framing, suggesting that the context of trust formation plays a critical role in moderating these effects.

4.2 Propensity: direct effects and the absence of moderation

Hypotheses one, two, four and five postulated effects of propensity as moderators and direct effects of propensity. No moderation effects were found in either model. Given this, hypothesis one and four is rejected. There was, however, a direct effect of propensity and propensity to trust automation, supporting hypothesis five and two.

Because propensity is such an integral part of the trust process (Mayer et al., 1995), it is surprising that neither of the two types proved to moderate the effect of the intervention, even when only a small effect was expected. Being more inclined to trust others was believed to increase the impact of the intervention vignettes. There could be several possible explanations for a lack of effect. First, it may entail that the propensity to trust does not increase the impact of others’ opinions when making a trust attribution, because the propensity has already “made its mark” on the attribution. Because both types of propensity are believed to be a deeply ingrained and learned part of a person’s disposition, it seems that the influence of others may prove more of an additional source of data, on top of the baseline evaluation. Second, there might not be grounds for “using” or tapping into ones’ disposition when attributing trust in this context, as the variable assessing this assumes that the individual can clearly imagine what the transport service will be like and how they personally will use and interact with it. The testimony of others may be informative enough on its own. This further emphasizes the direct effects of propensity and is also supporting the notion that context-specific measures are better at predicting an outcome regarding trust-attribution. For example, the propensity to trust automation explained much more of the variance in the moderation models.

This also emphasizes that these two concepts are distinct enough that they demand different measures and must be studied independently. Moreover, propensity to trust automation had a higher effect than the intervention when investigating the main effects, indicating that a solid baseline-trust in automated systems can have a similar impact on shaping people’s perceptions as norm-based interventions. Thus, it is of importance to understand and address both individual dispositions and automation-specific factors.

4.3 Interconnected constructs: associations with intention to use

Except for technical competence and propensity to trust, all other variables exhibited significant correlations with one another. A modest correlation was observed between the propensity to trust automation and the propensity for interpersonal trust, supporting hypothesis three. This may entail that they exhibit a degree of interdependence, while still being clearly separate constructs. Propensity to trust automation also had moderate to high correlations with subjective trust in SAVs, intention to use SAVs and safety evaluation. Intriguingly, propensity to trust others had a similar pattern, but the correlations were weaker overall. Perhaps the two constructs share some common elements, or that propensity to trust automations is functioning as a context specific propensity measure of the elsewise same tendency. Propensity to trust others may be general enough to affect trusting behavior regardless of context, while propensity to trust automation exhibits a high validity in the specific context of autonomous vehicles.

Additionally, technical competence correlated with propensity to trust automation, but not with propensity to trust others. This may indicate that more technically competent people have a higher tendency to trust automation, or that people who trust automation easier will have more exposure to technology and hence get increased competence through experience. However, the correlation was low.

Not surprisingly, subjective trust and intention to use shared autonomous vehicles correlated highly, supporting hypothesis nine. The measure of intention is a strong indicator of real-world behavior (Ajzen, 1985), and the fact that subjective trust is closely related validates the measure as an important decider in trusting autonomous vehicles.

4.4 Limitations

Both intervention groups had a higher tech competence score than the control (although only the mean of the static-norm group was statistically significantly different), which may suggest that more technologically competent respondents were better able to answer the comprehension check correctly. While this may create an artificial association between group assignment and technical competence, it is unlikely to have influenced the results, as the intervention was administered after measuring technical competence. The intended measure is an aggregate of tech-savviness, ability to interact with technology, and interest in technology. Hence, it is reasonable that individuals who score slightly lower on this measure would have a marginally increased risk of failing such a check, although this aspect should not be overly emphasized, and would need testing in a follow up study. Overall, it has most likely not affected the analysis to a large extent but should not be overlooked.

A notable limitation of this study lies in the operationalization of the static and dynamic norms. The static norm presented a high majority influence (70%), which may have elicited different psychological mechanisms than the 30% static norm used in Sparkman and Walton’s work, where minority influence played a key role. Majority norms may generate conformity but lack the motivational force associated with shifting or emerging social trends.

Furthermore, the phrasing of the dynamic norm, while highlighting change, did not specify the direction or magnitude of the change (e.g., percentage of people adopting a more favorable opinion). This vagueness may have reduced the salience of the dynamic norm intervention compared to the static norm, leading to the lack of significant differences observed between the two. Future studies should consider refining the phrasing of dynamic norms to explicitly state the extent and direction of change, ensuring a clearer distinction between the two normative framings.

Another potential limitation is the use of a rumination prompt, which was included for participants in both treatment conditions but not for those in the control group. While the intent was to encourage active engagement with the normative statements, it is possible that the prompt itself influenced subjective trust in SAVs by fostering reflection on the technology. This additional engagement may have contributed to the higher levels of trust observed in the treatment groups, independent of the normative intervention. Future studies should consider controlling for this potential confounding variable by including similar reflective tasks across all experimental conditions. The degree of rumination can also be of importance, as the control group did not get a prompt to stir rumination the same way as the manipulation group. However, they were showed the text reminder in the same vein, and we believe this is of minimal impact.

Most of the scales used in the study were partly or completely constructed from the ground up. While the measures had a good reliability, some should be explored further, and validated to assure their precision.

Additionally, the respondents were asked to evaluate something that few or none of them had prior knowledge about. This may be a threat to the validity of the measures, as it may be difficult to accurately assess something without firsthand experience. Still, a central goal of this study was to examine if normative influence could affect a novel or unknown agent. The results may have been different if the sample was more experienced or exposed to shared autonomous vehicles.

The variable that measures exposure may not be the most accurate way to examine the effect of knowledge related to SAVs. The way the item was constructed excluded all forms of self-driving that was not directly related to the pilots mentioned in the study. Additionally, it did not measure experience, but rather knowledge. It may have been more informative to measure any hands-on experience in addition.

5 Generalizability, future research and conclusion

This study underscores the critical impact social norms have on the acceptance of shared autonomous public transport, pinpointing the detrimental consequences a poorly executed pilot project may have on implementation. The formation of trust is largely hierarchical, suggesting that negative experiences can swiftly sour public perception and hinder the adoption of such technology. Findings from this study may also be generalized to other autonomous methods of transport or inform further studies and experiments, including larger vehicles and other use cases. Even if the small shuttle design introduces some unique interactions and caveats that may not be present in other modes of transport, trust can have a spillover-effect, especially when attributing similar services and trustors.

Future research should delve into the roles of static and dynamic social norms, aiming to further validate the observed effects on trust behaviors. The rapid pace of technological advancements in autonomous vehicles further underscores the urgent need for ongoing research, particularly focusing on the psychological elements of technology adoption.

The study indicated that the propensity to trust others, and the propensity to trust automation influences a person’s attribution of shared autonomous vehicles. Propensity to trust automation had a larger effect, suggesting that propensity is context specific. Neither type of propensity moderated the influence of social norms, contradictory to expectations. Safety evaluation seems to be of great importance, but the effect of social normative influence persists even when controlling for this factor. Conclusively, safety concerns are one of the most important barriers to trusting an autonomous shuttle, however, the way we perceive that others attribute trust in autonomous public transport can also affect our own opinion and attributions, especially in novel trusting situations.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. Declaration of generative AI in scientific writing Statement: During the preparation of this work the authors used ChatGPT 4.0/4O in order to proofread and polish. After using this tool/service, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

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.2026.1687363/full#supplementary-material

Supplementary material Supplementary material at the source

Glossary

Autonomous Vehicles (AVs) Self-driving vehicles that operate without human intervention, using sensors, cameras, and artificial intelligence for navigation and control Shared Autonomous Vehicles (SAVs) A subset of AVs designed for shared, public use, often in a ride-sharing or public transportation context Subjective Trust The personal perception or belief in the reliability and safety of an autonomous system or vehicle, influenced by individual or social factors Propensity to Trust An individual’s general disposition or tendency to trust others, including technology and automation, in the absence of specific information Static Norms Social norms that reflect a stable and unchanging standard, used to influence attitudes by presenting fixed information about others’ behaviors Dynamic Norms Norms that highlight ongoing changes in behavior or attitudes over time, often used to suggest increasing trends in trust or acceptance Trust in Automation The level of confidence users have in automated systems, based on their performance, safety, and reliability Multi-Level Model on Automated Vehicle Acceptance (MAVA) A model that measures several factors influencing the acceptance of autonomous vehicles, such as safety, performance, and ease of use Social Normative Influence The impact that the behavior or attitudes of others have on an individual’s own behaviors and beliefs, especially in contexts involving new technologies Intention to Use A psychological measure that predicts whether an individual plans to use a new technology, such as SAVs, based on their perceptions of trust, safety, and usefulness

Footnotes

Edited by: Juneyoung Park, Hanyang University, Republic of Korea

Reviewed by: Pavel Loskot, The Zhejiang University-University of Illinois at Urbana-Champaign Institute, United States

Dragan Pamucar, University of Belgrade, Serbia

Statements

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Funding

The author(s) declared that financial support was not received for this work and/or its publication.

Data availability

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.

Ethics approval

The studies involving humans were approved by the Department of Psychology’ internal research ethics committee (UiO). The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

Author contributions

SV: Investigation, Conceptualization, Validation, Writing – review and editing, Methodology, Formal Analysis, Resources, Data curation, Visualization, Writing – original draft, Project administration. OA: Validation, Supervision, Conceptualization, Writing – review and editing, Methodology. PU: Supervision, Writing – review and editing, Methodology, Validation.

References

  1. Aasvik O. (2025). PCI RR - SAVA. doi:10.17605/OSF.IO/RGCW4
  2. Aasvik O., Hagenzieker M., Ulleberg P., Bjørnskau T. (2024). How testing impacts willingness to use and share autonomous shuttles with strangers: the mediating effects of trust and optimism. Int. J. Human–Computer Interact. 41, 1 – 16. doi:10.1080/10447318.2024.2352220
  3. Aasvik O., Ulleberg P., Hagenzieker M. (2025). Exploring the general acceptance factor for shared automated vehicles: the impact of personality traits and experimentally altered information. Front. Psychol. 16, 1531386. doi:10.3389/fpsyg.2025.1531386
  4. Acheampong R. A., Cugurullo F. (2019). Capturing the behavioural determinants behind the adoption of autonomous vehicles: conceptual frameworks and measurement models to predict public transport, sharing and ownership trends of self-driving cars. Transp. Res. Part F Traffic Psychol. Behav. 62, 349 – 375. doi:10.1016/j.trf.2019.01.009
  5. Ajzen I. (1985). “ From intentions to actions: a theory of planned behavior,” in Action control. Editors Kuhl J., Beckmann J. (Springer Berlin Heidelberg), 11 – 39. doi:10.1007/978-3-642-69746-3_2
  6. Bicchieri C., Muldoon R., Sontuoso A. (2018). “ Social Norms ”. in In The Stanford Encyclopedia of Philosophy (Winter 2018 Edition). Available online at:.
  7. Bruhn A. (2019). Relying on the heuristic of trust: a case study. Account. and Finance 59 (S1), 333 – 357. doi:10.1111/acfi.12346
  8. Choi J., Ji Y. G. (2015). Investigating the importance of trust on adopting an autonomous vehicle. International Journal of Human-Computer Interaction 31, 150709133142005. doi:10.1080/10447318.2015.1070549
  9. Cohen J. (1988). Statistical power analysis for the behavioral sciences. Routledge. doi:10.4324/9780203771587
  10. Cronbach L. J., Meehl P. E. (1955). Construct validity in psychological tests. Psychol. Bull. 52 (4), 281 – 302. doi:10.1037/h0040957
  11. Faul F., Erdfelder E., Buchner A., Lang A.-G. (2009). Statistical power analyses using G*Power 3.1: tests for correlation and regression analyses. Behav. Res. Methods 41 (4), 1149 – 1160. doi:10.3758/BRM.41.4.1149
  12. Field A. P. (2013). Discovering statistics using IBM SPSS statistics: And sex and drugs and rock “n” rols. 4th Edn. London, United Kingdom: SAGE Publications Ltd.
  13. Frazier M. L., Johnson P. D., Fainshmidt S. (2013). Development and validation of a propensity to trust scale. J. Trust Res. 3 (2), 76 – 97. doi:10.1080/21515581.2013.820026
  14. Fröhlich P., Baldauf M., Meneweger T., Erickson I., Tscheligi M., Gable T. (2019). “ Everyday automation experience: non-expert users encountering ubiquitous automated systems ” in PExtended Abstracts of the 2019 CHI Conference on Human Factors in Computing Systems, 1 – 8. doi:10.1145/3290607.3299013
  15. Goldbach C., Sickmann J., Pitz T., Zimasa T. (2022). Towards autonomous public transportation: attitudes and intentions of the local population. Transp. Res. Interdiscip. Perspect. 13, 100504. doi:10.1016/j.trip.2021.100504
  16. Greifenstein M. (2024). Factors influencing the user behaviour of shared autonomous vehicles (SAVs): a systematic literature review. Transp. Res. Part F Traffic Psychol. Behav. 100, 323 – 345. doi:10.1016/j.trf.2023.10.027
  17. Harrison McKnight D., Chervany N. L. (2001). “ Trust and distrust definitions: one bite at a time,” in Trust in cyber-societies. Editors Falcone R., Singh M., Tan Y.-H. (Springer Berlin Heidelberg), 2246, 27 – 54. doi:10.1007/3-540-45547-7_3
  18. Hoff K. A., Bashir M. (2015). Trust in automation: integrating empirical evidence on factors that influence trust. Hum. Factors J. Hum. Factors Ergonomics Soc. 57 (3), 407 – 434. doi:10.1177/0018720814547570
  19. Hult R., Campos G. R., Steinmetz E., Hammarstrand L., Falcone P., Wymeersch H. (2016). Coordination of cooperative autonomous vehicles: toward safer and more efficient road transportation. IEEE Signal Process. Mag. 33 (6), 74 – 84. doi:10.1109/MSP.2016.2602005
  20. Jessup S. A., Schneider T. R., Alarcon G. M., Ryan T. J., Capiola A. (2019). “ The measurement of the propensity to trust automation,” in Virtual, augmented and mixed reality. Editors Chen J. Y. C., Fragomeni G. (Springer International Publishing), 11575, 476 – 489. doi:10.1007/978-3-030-21565-1_32
  21. Jin H., Roy A. (2024). Waymo arson in San Francisco sparks new debate on self-driving cars. Reuters: New York. Available online at: (Accessed November 24, 2025).
  22. Körber M. (2019). “ Theoretical considerations and development of a questionnaire to measure trust in automation,” in Proceedings of the 20th congress of the international ergonomics association (IEA 2018). Editors Bagnara S., Tartaglia R., Albolino S., Alexander T., Fujita Y. (Springer International Publishing), 823, 13 – 30. doi:10.1007/978-3-319-96074-6_2
  23. Lee J. D., See K. A. (2004). Trust in automation: designing for appropriate reliance. Hum. Factors J. Hum. Factors Ergonomics Soc. 46 (1), 50 – 80. doi:10.1518/hfes.46.1.50_30392
  24. Lee M., Turban E. (2001). A trust model for consumer internet shopping. International Journal of Electronic Commerce/Fall 6, 75 – 91. doi:10.1016/j.trf.2025.05.032
  25. Lewis M., Sycara K., Walker P. (2018). “ The role of trust in human-robot interaction,” in Foundations of trusted autonomy. Editors Abbass H. A., Scholz J., Reid D. J. (Springer International Publishing), 117, 135 – 159. doi:10.1007/978-3-319-64816-3_8
  26. Mayer R. C., Davis J. H., Schoorman F. D. (1995). An integrative model of organizational trust. Acad. Manag. Rev. 20 (3), 709. doi:10.2307/258792
  27. McKnight D. H., Choudhury V., Kacmar C. (2002). Developing and validating trust measures for e-Commerce: an integrative typology. Inf. Syst. Res. 13 (3), 334 – 359. doi:10.1287/isre.13.3.334.81
  28. Mengel A. K., Nastjuk I., Lechte H., Kolbe L. M. (2025). How can i trust you? The development of trust in autonomous vehicles. Transp. Res. Part F Traffic Psychol. Behav. 114, 354 – 375. doi:10.1016/j.trf.2025.05.032
  29. Nordhoff S., Kyriakidis M., van Arem B., Happee R. (2019). A multi-level model on automated vehicle acceptance (MAVA): a review-based study. Theor. Issues Ergonomics Sci. 20 (6), 682 – 710. doi:10.1080/1463922X.2019.1621406
  30. Nordhoff S., Hagenzieker M. P., Lehtonen E., Oehl M., Wilbrink M., Öztürk İ. (2023). Towards the development of an instrument for the assessment of road user automated vehicle acceptance: a pyramid of user needs of automated vehicles. arXiv. doi:10.13140/RG.2.2.25055.02728
  31. Paddeu D., Parkhurst G., Shergold I. (2020). Passenger comfort and trust on first-time use of a shared autonomous shuttle vehicle. Transp. Res. Part C Emerg. Technol. 115, 102604. doi:10.1016/j.trc.2020.02.026
  32. Rotter J. B. (1967). A new scale for the measurement of interpersonal trust1. J. Personality 35 (4), 651 – 665. doi:10.1111/j.1467-6494.1967.tb01454.x
  33. Ruter (2019). The oslo study – how autonomous cars may change transport in cities. Available online at: (Accessed March 23, 2023).
  34. Schoorman F. D., Mayer R. C., Davis J. H. (2007). An integrative model of organizational trust: past, present, and future. Acad. Manag. Rev. 32 (2), 344 – 354. doi:10.5465/amr.2007.24348410
  35. Sherif M. (1936). The psychology of social norms. Harper. xii, 210.
  36. Sparkman G., Walton G. M. (2019). Witnessing change: dynamic norms help resolve diverse barriers to personal change. J. Exp. Soc. Psychol. 82, 238 – 252. doi:10.1016/j.jesp.2019.01.007
  37. Tavakol M., Dennick R. (2011). Making sense of Cronbach’s alpha. Int. J. Med. Educ. 2, 53 – 55. doi:10.5116/ijme.4dfb.8dfd
  38. Venkatesh V., Bala H. (2008). Technology acceptance model 3 and a research agenda on interventions. Decis. Sci. 39 (2), 273 – 315. doi:10.1111/j.1540-5915.2008.00192.x
  39. Venkatesh V., Thong J., Xu X. Hong Kong University of Science and Technology The Hong Kong Polytechnic University (2016). Unified theory of acceptance and use of technology: a synthesis and the road ahead. J. Assoc. Inf. Syst. 17 (5), 328 – 376. doi:10.17705/1jais.00428
  40. Xu Z., Zhang K., Min H., Wang Z., Zhao X., Liu P. (2018). What drives people to accept automated vehicles? Findings from a field experiment. Transp. Res. Part C Emerg. Technol. 95, 320 – 334. doi:10.1016/j.trc.2018.07.024
  41. Yuen K. F., Huyen D. T. K., Wang X., Qi G. (2020). Factors influencing the adoption of shared autonomous vehicles. Int. J. Environ. Res. Public Health 17 (13), 4868. doi:10.3390/ijerph17134868
  42. Zhang Z., Tian R., Duffy V. G. (2023). “ Trust in automated vehicle: a meta-analysis,” in Human-automation interaction. Editors Duffy V. G., Landry S. J., Lee J. D., Stanton N. (Springer International Publishing), 11, 221 – 234. doi:10.1007/978-3-031-10784-9_13
  43. Zhanhang Y., Jingyang W. (2024). N wuhan, robotaxis put china’s self-driving ambitions to the test. Sixth Tone. Available online at: (Accessed November 24, 2025).

Citation

Vassanyi, S., et al. (2026). Driven by trust? Social norms increase trust in shared autonomous public transport. Frontiers in Future Transportation, Volume 7. https://doi.org/
Received
Accepted
Published
Volume
7 - 2026

Copyright

Copyright © 2026 Vassanyi, Aasvik and Ulleberg.

License: https://creativecommons.org/licenses/by/4.0/

*Correspondence: Sander Vassanyi

Article metrics

Views
0
Downloads
0
Citations
0
Shares
0

Related articles

Original ResearchA semi-quantitative decision matrix for Polycyclic aromatic hydrocarbons source apportionment to support sustainable urban transport and air quality managementCarla Sofía Stadler, Manuel Peña Moules, Brenda Estefanía Badura, Daniela Silvana Giuliani, Erica Yanina Sanchez, Atilio Andrés Porta, Jorge Esteban Colman LernerOriginal ResearchSemi-supervised zero-knowledge federated learning for intelligent transportation systemsMirabela-Melinda Medvei, Miruna-Mihaela Modiga, Iuliani Aciobăniței, Nicolae ŢăpuşOriginal ResearchAdaptive calibration and spatio-temporal filtering for robust highway wrong-way driving detectionDian Liu, Xiaomei Yuan, Zening Zhang, Lambert Zhang, Haoxiang Wang, Yang Liu