Investigating the effects of open versus closed systems on trust in autonomous vehicles

1 Introduction

Although automated vehicles (AVs) have the potential to improve safety, benefit the environment, and reduce costs and travel time, these benefits can only be realized if drivers recognize the value of these vehicles and integrate them into their daily lives (Rahman et al., 2018). Consequently, driver acceptance is essential to the successful deployment of AV technologies (Rahman et al., 2017). Trust in automation is a critical prerequisite for such acceptance (Hutson, 2017; Kaur and Rampersad, 2018; Rahman et al., 2018; Deb et al., 2018; Choi and Ji, 2015). Distrust has been identified as one of the major barriers to AV adoption. As noted in A Matter of Trust, Jack Weast, chief systems architect of Intel’s autonomous driving group, stated, “we could have the safest car in the world, but if consumers do not want to put their kids into it, then there’s no market,” (Hutson, 2017, p. 1375). Public surveys reinforce this concern; for example, 78% of respondents in a AAA survey reported fear of riding in an autonomous vehicle.

To enable successful implementation, it is necessary to understand the factors that cultivate trust in AV technologies. Trust has been defined as “one’s willingness to place himself/herself in a vulnerable position, with respect to a technology, with a positive expectation of an outcome or a positive nature of future behavior” (Kaur and Rampersad, 2018; Rahman et al., 2018, p. 90). According to Choi and Ji (2015), trust mediates the human–automation relationship and is a major determinant of reliance and acceptance, alongside personal beliefs and intentions to use the technology. Trust in AVs has been conceptualized as comprising three dimensions: belief that the system is predictable and understandable, that it performs tasks accurately, and that it provides adequate assistance (Choi and Ji, 2015). Achieving these dimensions requires users to possess sufficient knowledge of the system’s functionality, benefits, and limitations.

Public education regarding the capabilities and limitations of autonomous vehicles is therefore essential (Hutson, 2017). Drivers and pedestrians must both understand and trust that AVs will perform as expected, while avoiding overestimation of system capabilities, which can create unsafe conditions. Overestimation is a well-documented issue; Walker et al. (2018) found that participants’ initial trust scores often reflected inflated expectations of AV performance. To address this, Hutson introduced the concept of calibrated trust, defined as trusting the vehicle where it performs well, while not trusting it beyond its capabilities (2017). Safe societal integration of AVs requires not just trust, but properly calibrated trust. Understanding system behavior also plays a critical role; Bertram Malle emphasized that “understanding why the vehicle is doing what it is doing will be a critical factor in trust and acceptance” (Hutson, 2017, p. 1376). Designers must therefore carefully account for system limitations to ensure user safety and increase trust (Deb et al., 2018).

Several methods exist to study receptivity and behavior toward automated vehicles, particularly among pedestrians. Diaz developed the Pedestrian Behavior Questionnaire (PBQ), a 16-item instrument assessing general factors influencing pedestrian behavior (Deb et al., 2018). However, the PBQ does not address attitudes specific to autonomous vehicles. To address this gap, Deb et al. (2018) developed the Pedestrian Receptivity Questionnaire for Fully Autonomous Vehicles (PRQF), which builds upon the PBQ framework. The PRQF assesses receptivity through items related to attitude, social norms, trust, compatibility, and system effectiveness (Deb et al., 2017). Higher PRQF scores indicate greater receptivity toward fully autonomous vehicles. Receptivity, defined as willingness to interact with the technology, is distinct from acceptance, which reflects willingness to use it (Deb et al., 2017). Both differ from trust, which entails a willingness to place oneself in a vulnerable position with positive expectations (Kaur and Rampersad, 2018). Receptivity therefore serves as a precursor to trust, making the PRQF an appropriate instrument for studying early stages of trust development in AVs.

Trust in autonomous vehicles is influenced by a range of factors, including vehicle design features, user characteristics, and environmental context. Anthropomorphic features, such as human-like vehicle faces, have been shown to foster trust (Haeuslschmid et al., 2017; Hutson, 2017). These features attempt to replace informal communication cues traditionally exchanged between drivers and pedestrians, such as eye contact and gestures (Deb et al., 2018). Other anthropomorphic elements, including assigning a name, gender, or voice to the vehicle, have similarly been found to increase liking and trust (Ruijten et al., 2018).

Personal factors also play a role in shaping trust. Gender differences have been observed, with men more likely to accept autonomous vehicles than women (Kyriakidis et al., 2015). Age, system knowledge, and prior experience with automation further influence trust. Providing users with information about system capabilities and limitations has been shown to increase trust compared to providing no information (Khastgir et al., 2018). However, this presents a challenge, as most drivers have little to no direct experience with AVs (Haeuslschmid et al., 2017). Personality traits such as locus of control also affect trust perceptions. Individuals with an external locus of control—those who believe outcomes are determined by external forces—may be more inclined to trust AVs, as they tend to believe human drivers are more accident-prone (Choi and Ji, 2015).

Environmental factors play a similarly important role. Urban residents and frequent users of public transportation have demonstrated higher acceptance of AVs (Nordhoff et al., 2018). However, AVs are expected to be implemented initially in highway environments, which are viewed as more predictable than urban streets (Van Meldert and De Boeck, 2016). This highlights the need to understand trust formation in highway contexts. Sociodemographic status has also been examined; while low-income countries face greater transportation challenges, domain-specific attitudes appear more influential than demographics in predicting acceptance (Nordhoff et al., 2018).

Another key environmental distinction is between closed and open systems. According to Kaur and Rampersad (2018), early adoption of AVs is most likely in closed environments such as campuses, airports, and resorts. Research indicates that users are more willing to adopt AVs in closed settings, in public transportation with supervision, or when manual control remains an option (Kaur and Rampersad, 2018). Despite this, limited research has examined public willingness to use AVs for public transportation, even though this may be among the earliest large-scale applications.

Trust also varies by application domain. Autonomous systems are expected to be adopted sooner in logistics and controlled environments than in passenger transport due to reduced complexity, fewer regulations, and lower liability concerns (Van Meldert and De Boeck, 2016). These systems offer benefits such as reduced human error, improved reliability, and lower fuel consumption. However, it remains unclear whether trust developed in logistics contexts transfers to everyday traffic environments.

System performance is a critical determinant of trust. Trust declines when automation malfunctions, and negative experiences can lead to persistent negative expectations (Deb et al., 2018). Reflecting this concern, 54% of respondents in a Continental AG study doubted the reliability of AVs (Kyriakidis et al., 2015). Successful market adoption depends not only on technical capability but also on consumer trust (Deb et al., 2018). Systems must operate effectively and be compatible with their environments to earn user trust.

Autonomous features are increasingly becoming standard in U.S. vehicles (Rahman et al., 2017). However, widespread benefits depend on public trust and acceptance. Fear and distrust persist among many potential users (Kyriakidis et al., 2015), limiting societal gains from AV technology. Continued research is therefore needed to identify situations in which users are most willing to adopt autonomous vehicles. Understanding these scenarios can inform early implementation strategies and facilitate smoother transitions to automation (Kaur and Rampersad, 2018).

Recent research emphasizes the importance of calibrated trust supported by system transparency and communication. Explanation strategies influence trust differently depending on perceived risk; under high-risk conditions, explanations may reduce trust, while under low-risk conditions they may enhance it (Zhang et al., 2024). Explainability has also been shown to directly and indirectly influence trust through perceptions of safety, usefulness, and intelligence, suggesting that users perceive AVs as intelligent agents rather than passive tools (Khan, 2024). Interaction design further affects trust, with proactive communication improving trust in some contexts but undermining it in others when perceived as excessive (Li et al., 2024). Trust has also been shown to interact with perceived risk to shape adoption intentions, with performance trust emerging as particularly influential (Panagiotopoulos and Dimitrakopoulos, 2018).

Trust extends beyond vehicle occupants to pedestrians and other road users. Pedestrian trust influences crossing behavior, with higher trust increasing crossing likelihood and higher perceived risk decreasing it (Liu et al., 2025). Methodologically, trust research has been criticized for over-reliance on short-term experimental studies. Recent reviews advocate for interdisciplinary and longitudinal approaches that better capture socially embedded trust dynamics (Hind, 2024; Broo et al., 2024). These perspectives underscore the complexity of trust in autonomous vehicles and the need for comprehensive, context-sensitive research frameworks.

Taken together, prior research demonstrates that trust in autonomous vehicles is influenced by a broad range of personal, technological, and environmental factors. However, much of the existing literature has examined these factors in isolation or within highly controlled experimental settings, with limited attention to how specific deployment environments shape trust and receptivity among non-users, particularly pedestrians. Although closed-system environments such as campuses, airports, and industrial sites are widely acknowledged as likely early adoption contexts for autonomous vehicles, empirical research directly comparing trust and receptivity across open and closed systems remains limited. This gap is especially important, as closed systems reduce environmental complexity and assume a greater level of user familiarity or training, which may fundamentally alter trust formation.

Building on these limitations, the present study focuses specifically on the environmental distinction between open and closed systems to provide a more targeted understanding of how deployment context influences pedestrian receptivity and trust in autonomous vehicles. By isolating system accessibility as a defining feature of the environment, this study advances existing literature by examining whether trust is more readily established in controlled, closed systems compared to broadly accessible open systems. In doing so, the study contributes empirical evidence to support early-stage implementation strategies and informs the design and placement of autonomous vehicle technologies in environments where public trust is most likely to emerge.

2 Research question

The goal of this study was to investigate factors that influence trust in autonomous vehicles. Though all of the factors mentioned above could be studied in reference to their effects on autonomous vehicle perceptions, this study focused on one factor in order to have a well-defined scope. The environmental factor of open versus closed systems was evaluated because it poses a significant contribution to current research. For the purpose of this study, a system was defined in terms of its accessibility. An open system is easily accessible to a broad group of people. In contrast, a closed system is only accessible to certain groups of people and vehicles. A closed system eliminates many variables that are present in an open system, one of those being training or experience level. There is a level of training that is assumed to be involved with a closed system as opposed to a broad, open one. Examples of closed systems where autonomous vehicles could be incorporated include university settings, golf courses, airports, or company warehouses. Open systems can refer to highways or roadways. Thus, the research question of this study is: Are pedestrians more receptive and trusting of autonomous vehicles when the autonomous vehicles are operating in a closed system?

3 Methods

3.1 Experimental design

The independent variable for this study was system type (general, open, or closed). The dependent variable was trust. Trust was measured in three ways: scenario question one, scenario question two, and the pedestrian receptivity questionnaire. Scenario question two was the most important trust measure, because it explicitly asked participants to rate their level of trust in the vehicle. However, this question alone was not adequate to measure trust because some participants may not know how to accurately report their level of trust using a single rating. Therefore, another scenario question was asked to determine how a participant would respond to the vehicle. As trust elicits certain behaviors, we would expect varying trust levels to become apparent in varying reported responses. The behavior question (scenario question one) was used to provide an indirect measurement of trust. Lastly, the receptivity questionnaire measures receptivity to autonomous vehicles. It was an important addition to this study because receptivity indicates a participant’s willingness to interact with the technology. A participant must be willing to interact with autonomous vehicles before they can develop trust in them. Therefore, receptivity was a pertinent factor that was used to examine trust in this study.

3.2 Survey

A survey was administered to participants via Amazon Mechanical Turk. Participants indicated their consent to participate in the survey after reading a description of its purpose and choosing to participate. The survey began with seven demographics questions to determine gender, age, education, race, and other applicable background questions. The PRQF survey was adapted from a study by Deb et al. (2017) and was used to elicit information about participants’ beliefs about autonomous vehicles. The first section consisted of a 20-item pedestrian behavior questionnaire. The next section contained 16 items to assess pedestrian receptivity to fully autonomous vehicles and four items to assess personal innovativeness. These questions were followed by a set of scenarios and accompanying questions (see Table 1). After completing the first scenario, the participants completed the second scenario. The second scenario consisted of an introduction to autonomous vehicles in the applied setting (either open or closed) followed by a pedestrian receptivity questionnaire. Then, participants read either an open or closed scenario description, followed by the same questions that were asked of the general scenario (one for response to the vehicle and one for trust in the vehicle). The open setting included the following introduction “Consider a fully autonomous vehicle operating in an open system. Examples of open systems could be a downtown city street or the streets of your hometown. The environment is open to all vehicles and pedestrians. The driverless vehicle is interacting with other drivers, pedestrians, bicycles, and is operating under existing traffic laws. Answer the following questions based on this type of environment.” After completing the pedestrian receptivity questionnaire, participants read the open scenario, “You are walking downtown. On your way, you need to cross multiple crosswalks, both signalized and unsignalized. As you prepare to cross at an unsignalized crosswalk, you find that a driverless vehicle is approaching the crosswalk (with no one sitting in the driver seat). Based on the scenario, please select the choice that best reflects your behavior for each qu estion.” Participants then answered the scenario questions. The closed system began with the following introduction, “Consider a driverless vehicle operating in a closed system. A closed system is an environment with limited access. An example of a closed system could be a company campus where only certain people can access roads and vehicles must have permission to drive on them. Answer the following questions based on this type of environment.” Participants then took the pedestrian receptivity questionnaire for the closed scenario. After this, participants read the following description for the closed scenario, “You are walking across your company campus. On your way, you need to cross multiple crosswalks, both signalized and unsignalized. As you prepare to cross at an unsignalized crosswalk, you find that a driverless vehicle is approaching the crosswalk (with no one sitting in the driver seat). Based on the scenario, please select the choice that best reflects your behavior for each question.” Each participant only answered questions for one of the scenarios: either open or closed. The chosen scenario was randomized for the participants, ensuring that approximately half of the participants completed the open scenario and the other half completed the closed scenario. After completing the survey, participants were given an Amazon Mechanical Turk code to use for compensation for their participation in the survey.

TABLE 1. Survey scenarios.

3.3 Hypotheses

Hypothesis 1 Trust will be higher for closed systems. Supported by research from Kaur and Rampersad (2018), it is expected that participants will be more trusting of an autonomous vehicle if it is being operated in a controlled environment as opposed to a non-controlled environment. Furthermore, a closed system will elicit higher trust scores because it contains fewer variables than an open system, and more training is required to access the system.

Hypothesis 2 Trust will increase as personal innovativeness increases. The more experience a participant has with technology, the more likely they are to accept and trust autonomous vehicle technology. This concept is supported by research from Khastgir et al. (2018) stating that knowledge of the system increases trust in autonomous vehicles. Applied to this study, increased personal innovativeness scores indicate deeper knowledge of technology. This technological knowledge will subsequently transfer to autonomous vehicle technology to result in increased trust scores.

Hypothesis 3 Trust will be higher for participants who have a high general pedestrian receptivity score. Participants who indicate a high receptivity to autonomous vehicles will be more likely to trust the vehicles described in the scenarios in the survey. Receptivity is the willingness to interact with a technology (Deb et al., 2017), while trust is the willingness to place oneself in a vulnerable position with respect to technology with a positive expectation of the outcome (Kaur and Rampersad, 2018). Therefore, receptivity is a necessary precursor to trust. If participants indicate receptivity to autonomous vehicles, they may be more likely to trust the vehicles as well.

3.4 Participants

This study was approved by Mississippi State University Institutional Review Board (#19–072) prior to beginning. The survey was assembled on Qualtrics and posted on Amazon’s Mechanical Turk service as a work task. In order to complete the survey, participants were required to be U.S. residents aged 18 and older who are members of Amazon Mechanical Turk. There were no other inclusion or exclusion criteria for survey participation. Participants indicated consent to complete the study after reading a description and choosing to participate in the survey. Participants were compensated for completing the survey. A total of 442 people participated in the survey. A check question was included in the survey to confirm that participants were reading the questions and providing accurate responses. Forty-three of those responses were not included in the analysis because they failed an attention check question. Therefore, 399 responses were used in subsequent analyses. Of the total sample size of 399 participants, 201 responded to the open scenario and 198 responded to the closed scenario.

4 Data collection and results

4.1 Data preparation

The responses from the survey were exported to Excel to begin analysis. First, the data was cleaned and coded. The responses that failed a check question were removed. Some of the items were reverse-coded to maintain consistency in the results.

After coding the responses, the scores were calculated for the various questionnaires included in the survey. The methods used to calculate the scores for this survey were drawn from a study performed by Deb et al. (2017). To calculate the PBQ score, the responses were divided into subscales. The subscales are indicated by their headings in the survey, with questions one to four indicating violations, questions five to eight indicating errors, questions 9–12 indicating lapses, questions 13–16 indicating aggressive behavior, and questions 17–20 indicating positive behavior. These subscale scores were calculated by averaging the items in each category to create a score for each subscale. These items were then added together to create the total PBQ score. There were five subscales for the PBQ, so possible scores ranged from 5 to 30. For the pedestrian receptivity questionnaire, there were three subscales: safety, interaction, and compatibility. The safety category included questions 1, 2, 8, and 11. The interaction category included questions 3, 4, 5, 6, 7, 12, 13, 14. The compatibility category included questions 15, 16, 17. These subscales scores were calculated by averaging the items in each category to create a score for each subscale, just like what was done for the PBQ. These items were then added together to create the total PRQ score. Possible scores for the PRQ ranged from three to 21. These steps were completed three times to calculate the general PRQ, open PRQ, and closed PRQ scores. In order to calculate the personal innovativeness score, the responses for the four items in the personal innovativeness questionnaire were averaged together to create a total score. The possible scores for the personal innovativeness questionnaire ranged from one to 7.

After calculating the scores, the data was reduced to 17 variables. These variables included the demographics questions, total scores for the PBQ, receptivity (PRQ), and innovativeness, as well as the scenario responses. There was missing data for the scenario responses since half of the participants answered each scenario, so these responses were consolidated into a single variable, known as “second.” The second item represented the second scenario participants viewed in the survey which was either open or closed. Lastly, a variable was added to distinguish these “second” responses between open and closed. The list of variables is described in detail in Table 2.

TABLE 2. Final variables and descriptions.

4.2 Results

4.2.1 Descriptive summaries

After cleaning and coding the responses and calculating the scores, the data was imported to SPSS to begin analysis. First, a set of descriptive summaries were obtained. In terms of the participants of the study (n = 399), the mean age was 34 with a standard deviation of 10 years. The majority of participants (67.9%) were male. Most of the participants had an associate’s or bachelor’s degree (52.6%). Participants were asked how often they walk in a day. The most frequently reported answer (41.6%) was that participants walk often, or two to four times a day. The second most reported answer (34.1%) was that participants walk rarely, or 0–2 times a day. Of this walking frequency, 38.1% of participants stated that they typically walk 15–30 min a day. The majority of respondents (78.7%) reported that they live in an urban area, which was defined as a place with an overall population density of at least 500 people per square mile.

In Table 3, the means and standard deviations were listed for the PBQ, personal innovativeness questionnaire, and the three PRQ’s. As discussed previously, a high PBQ score indicated adverse pedestrian behavior. The average score for the respondents was in the middle of the scale (M = 11.66, SD = 5.04), indicating standard pedestrian risk-taking behavior. The personal innovativeness scores were fairly high (M = 5.08, SD = 1.17), which shows that the participants were innovative. The results for the baseline PRQF (M = 13.69, SD = 3.74) and the open PRQ (M = 13.70, SD = 3.90) were not much different. However, the closed PRQ average score had the highest numerical value, indicating high receptivity to autonomous vehicles (M = 14.11, SD = 3.78).

TABLE 3. Descriptive statistics for questionnaire scores (PBQ, PRQF, innovativeness).

Table 4 describes the frequencies in percentages of the answer choices for the behavior question for all three scenarios. In all three scenarios, when participants were asked about their response at the crosswalk with the FAV approaching, the most reported answer was “I will make sure that the driverless vehicle stops before I start crossing” (Baseline: 43.9%, Open: 33.8%, Closed: 44.4%). The least reported answer for all three scenarios was “I will stand in front of the vehicle because I know it will not hurt me” (Baseline: 0.8%, Open: 1.0%, Closed: 0.5%). This answer choice was included to test for over-trust in the autonomous vehicle. Most of the participants of this study were not over-trusting of the autonomous vehicle described in the scenario, as they would not stand in front of it. The mean responses were the same for the open and closed scenarios, and they both had slightly higher values than the baseline scenario. The higher values indicated more trusting behavior around autonomous vehicles. Therefore, this indicates that participant behaviors were slightly more trusting for the open and closed scenarios as opposed to the baseline scenario, possibly due to an order effect or the additional detail included in the second scenario.

TABLE 4. Percentages of responses to scenario question 1: “What will your response at the crosswalk be, with the FAV approaching?”

Table 5 describes the frequencies in percentages of the answer choices for the trust question for all three scenarios. When participants were asked about their trust in the vehicle described in the scenario, the most reported response for the baseline scenario was “somewhat agree” (28.1%). However, for the open and closed scenarios, the most reported response was the same for both scenarios, “moderately agree” (Open: 24.9%, Closed: 24.7%). This indicated an increase in trust for the open and closed scenarios as opposed to the baseline scenario. The least reported response was the same for all three scenarios, “moderately disagree” (Baseline: 8.0%, Open: 5.5%, Closed: 7.6%). For the closed scenario, the least reported response was tied between “moderately disagree” and “somewhat disagree,” with both responses showing a frequency of 7.6%. The mean response for the closed scenario was numerically greater than that of the open and baseline scenarios. Because a greater mean value indicates more trust, this finding can be used to conclude that participants were generally more trusting of the autonomous vehicle in the closed system.

TABLE 5. Percentages of responses to scenario question 2: “Please rate your level of agreement with the following statement: I will trust the vehicle in the scenario.”

4.2.2 Inferential analysis results

Repeated measures ANOVA was used to compare the results for the baseline scenario to that of the open and closed scenarios. Scenario type was set as the within subjects variable (baseline scenario or second scenario) and system type was set as the between subjects variable (open or closed). Repeated measures ANOVA was used to evaluate differences in responses to behavior, trust, and PRQ scores.

For behavior, there was a significant difference between the baseline scenario (M = 2.77, SD = 1.20) and the second scenario participants viewed (M = 2.90, SD = 1.25), indicating a significant main effect (F(1, 1) = 6.85, p = 0.009) with participants rating question one higher on the second scenario. A higher rating for the behavior question indicated more trusting behavior, as the responses increased numerically from the least trusting behavior, 1: “I will not cross the road at the crosswalk to avoid crossing in front of the FAV,” to the most trusting behavior, 6: “I will stand in front of the vehicle because I know it will not hurt me.” However, there was not a significant difference between the baseline scenario and the open/closed scenario (F(1,1) = 0.01, p = 0.942). Therefore, there was a difference between the baseline scenario and the non-baseline scenario, but that difference was not affected by the open/closed factor.

For trust, there was not a significant main effect difference between the baseline scenario and the second scenario (F(1,1) = 1.21, p = 0.272). However, there was a significant interaction effect between the baseline scenario and the open/closed scenario (F(1,1) = 2.88, p = 0.090), with the change in closed scenario scores being greater than the change in open scenario scores (Closed Δ x ̄ = 0.12, Open Δ x ̄ = -0.03). This indicated a significant change in how participants responded based on whether they viewed the open or closed scenario. Similarly to behavior, a higher response for trust also indicated more trusting behavior, as the responses increased numerically from the least trusting response, 1: “moderately disagree,” to the most trusting response, 7: “strongly agree.”

There was a significant difference between the baseline receptivity score (M = 13.69, SD = 3.74) and the second receptivity score (M = 13.90, SD = 3.84), indicating a significant main effect (F(1,1) = 7.95, p = 0.005), with second receptivity scores being higher than baseline receptivity scores. Higher receptivity scores indicated that participants were more willing to interact with the autonomous vehicles. There was also a significant interaction difference between the baseline receptivity score and whether the participants took the open or closed receptivity questionnaire (F(1,1) = 3.56, p = 0.060), with the change in closed receptivity scores being greater than change in open receptivity scores (Closed Δ x ̄ = 0.36, Open Δ x ̄ = 0.07). This means the open or closed scenario had an effect on the receptivity responses.

To determine whether participant’s stated behavior was different based on if they viewed open or closed scenarios, a Chi-Square test was completed on the behavior question. According to this analysis, there was not a significant difference between participants’ behavior based on whether the scenario was open or closed (χ 2(6, N = 399) = 5.71, p = 0.335).

A Chi-Square test was also completed to determine whether participants’ trust changed based on the open and closed scenarios, using the responses for the trust question. According to this analysis, there was not a significant difference between participant’s trust scores based on whether the scenario was open or closed (χ 2(5, N = 399) = 2.92, p = 0.819).

A bivariate Spearman correlation with a two-tailed significance level was completed on several variables. The first compared the personal innovativeness total score to the baseline trust question. This was done to test the hypothesis that trust would increase as personal innovativeness increased. The results from this analysis showed that the two variables had a weak positive correlation (r s(397) = 0.282, p < 0.01), with higher personal innovativeness scores relating to higher trust scores. A scatter plot comparing the two variables was included in Figure 1.

A high pedestrian receptivity score corresponded to an increased willingness to interact with autonomous vehicles, whereas a low pedestrian receptivity score indicated a decreased willingness to interact with autonomous vehicles. A bivariate Spearman correlation was completed to compare the baseline pedestrian receptivity total score to the baseline trust score. The results indicated that the two variables had a strong positive correlation (r s(397) = 0.772, p < 0.01), with higher PRQ scores relating to higher trust scores. The relationship between the two variables was represented in the scatter plot in Figure 2.

The PBQ total score and the baseline trust question had a weak positive correlation (r s(397) = 0.197, p < 0.01), with higher PBQ scores relating to higher trust scores. A high PBQ score represented adverse pedestrian behavior. Therefore, adverse pedestrian behavior corresponded to an increased level of trust in autonomous vehicles. The relationship between the two variables was represented in Figure 3.

The only demographic variable that showed a significant correlation to trust was participant age. Age and baseline trust had a weak negative correlation (r s(397) = −0.178, p < 0.01), with higher-aged participants showing a lower level of trust. The relationship between age and trust was represented in Figure 4.

5 Discussion

The purpose of this study was to investigate the effects of open and closed systems on trust in autonomous vehicles. The first hypothesis was that participants would be more trusting of the autonomous vehicle in the closed system. Although the results from the repeated measures ANOVA did not indicate a significant difference in the responses to the trust question for the baseline scenario and the second scenario participants viewed, there was a significant change in scores based on the open and closed variable. Specifically, there was a positive change for the closed scenario, representing an increase in trust for FAVs in closed systems.

The second hypothesis was that trust would increase as personal innovativeness increased. The results from this analysis showed that the two variables were positively correlated, meaning that a higher personal innovativeness score was associated with higher trust scores. A high personal innovativeness score represented an increase in innovativeness. This supports the hypothesis that the more technologically innovative a person was, the more likely they were to trust the technological innovation of autonomous vehicles. This finding is supported by research stating that trust in autonomous vehicles is increased for those that have more knowledge about the system’s abilities and limitations (Khastgir et al., 2018; Hutson, 2017). This suggests that participants who were more technologically innovative also had more knowledge about autonomous vehicle technology than those that did not frequently use new technologies.

The third hypothesis was that trust would be increased for participants who had a high baseline pedestrian receptivity score. Results indicated that a higher baseline receptivity score was associated with higher trust scores. This proved the hypothesis that people who indicated a high receptivity to autonomous vehicles were more likely to trust the autonomous vehicle described in the survey. The results from this study agreed with the expected findings. There was a significant positive change in PRQ scores for the closed system. Average scores for the trust question were also increased for the closed system. This can be used to conclude that trust and receptivity to autonomous vehicles increase in closed systems. This finding suggests it can be applied to autonomous vehicles in these settings.

Another noteworthy finding was the response to the behavior question. Although there was a significant change in responses from the baseline scenario to the second scenario, there was not a difference between the open and closed scenarios. Additionally, in Table 3, the frequencies for the responses to this question were listed. It was interesting to note that few participants chose the answer choice that stated, “I will cross the road with full confidence that the driverless vehicle will stop for me,” (Baseline: 5%, Open: 8%, Closed: 6.1%). One of the possible explanations for this could be based on the findings from Deb et al. (2018) that indicated that the lack of ability to interact with autonomous vehicles through informal communication methods like eye contact and hand gestures could make it difficult for pedestrians to trust these vehicles.

Furthermore, it was discovered that the participants’ ages and responses to the trust question were negatively correlated. This means that older participants indicated less trust of the vehicles described in the scenarios, while younger participants tended to be more trusting of these vehicles. This finding was not surprising, as younger generations typically have more experience with technology and can, therefore, be more trusting of it. There were no other demographic variables that were found to have a significant impact on trust. Although there has been research stating that city residents are more likely to accept autonomous vehicles (Nordhoff et al., 2018), this finding was not supported by this study, as trust was not significantly affected by whether the participants lived in an urban or rural area (χ 2(6, N = 399) = 5.30, p = 0.506).

There was also a significant correlation between PBQ scores and trust, with the two variables being positively correlated. This was an interesting finding because high PBQ scores represented adverse pedestrian behavior. Therefore, participants who exhibited adverse pedestrian behavior were more trusting of the autonomous vehicle in this study. This could indicate over-trust in autonomous vehicles, as some of the questions from the PBQ were items like, “I cross even if vehicles are coming because I think they will stop for me,” and “I forget to look before crossing because I am thinking about something else.” It can be concluded that the participants who had higher PBQ scores and higher trust scores were already trusting of vehicles and were therefore more trusting of autonomous vehicles than those that indicated being more cautious around vehicles in general.

In conclusion, a difference was observed in participants’ identifications of trust based on autonomous vehicle settings. Participants of this study indicated higher trust and receptivity to autonomous vehicles in closed settings, as indicated from the increase in receptivity scores for the closed setting and the higher mean for responses to the trust question. This finding can be used to further research the topic of trust in autonomous vehicles and to develop scenarios for autonomous vehicle use that will elicit maximum trust from drivers and pedestrians. Autonomous vehicles are currently on their way to being introduced in the general market, but there is still a need to improve user perceptions and trust of these vehicles before this technology can truly be successful.

Although there were interesting findings from this study, there were also some limitations. It would have been helpful to include a question in the demographics section about how much exposure the participant had to autonomous vehicles prior to completing the study. This could have been useful to determine if previous knowledge and exposure impacted trust in autonomous vehicles. It also might have been beneficial to include more detail in the open and closed scenarios in order to gain a better understanding of participants’ feelings towards the vehicles in these settings. For instance, the scenario description could have been more detailed so participants could have a deeper understanding of the differences between an open and closed setting. There also could have been multiple scenarios for each setting. These additions would have created a more thorough evaluation of the differences in participants’ responses to autonomous vehicles in open and closed settings.

6 Conclusion

This work establishes the importance of system context in shaping trust and receptivity toward autonomous vehicles. By isolating the distinction between open and closed environments, the study shows that deployment setting plays a meaningful role in pedestrian perception of autonomous systems. These findings provide a clear understanding of how trust develops in early adoption scenarios and highlight the influence of environmental structure on user response. The results also emphasize the role of individual characteristics, indicating that user readiness and familiarity with technology remain important factors in trust development. Together, these insights suggest that controlled environments may serve as effective starting points for building trust in autonomous vehicles before broader deployment.

Future work should extend this analysis by incorporating more diverse scenarios, examining changes in trust over time, and evaluating how trust formed in controlled environments translates to more complex settings. Advancing this line of research will support more informed deployment strategies and improve the integration of autonomous vehicles into public environments.

Generative AI statement

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

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

Publisher’s note

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

Footnotes

Edited by: Wael Alhajyaseen, Qatar University, Qatar

Reviewed by: Tufail Ahmed, Qatar University, Qatar

Nora Krizsik, Institute for Transport Sciences, Hungary