Drivers’ Risk and Emotional Intelligence in Safe Interactions with Vulnerable Road Users: Toward Sustainable Mobility

Abstract

Sustainable urban transportation relies on safe interactions between motor vehicles and vulnerable road users (VRUs) such as bicyclists and pedestrians. This study evaluates how drivers’ risk-taking and emotional intelligence (EI) influence their interactions with VRUs in urban environments. A driving simulator study with 40 participants examined nine bicycle-passing events and one pedestrian-crossing scenario. The results show that higher risk-taking is significantly associated with more hazardous behaviors: each unit increase in risk-taking predicted a 4.02 mph higher passing speed and a 60% lower likelihood of braking for pedestrians. Event context also shaped behavior: drivers reduced their speed by 2.52 mph when passing cyclists on the road and by 2.33 mph for groups of cyclists, compared to single cyclists in bike lanes. Across all risk categories, the participants expressed discomfort when sharing the road, preferring to pass bicyclists on sidewalks, although the ‘risk-avoidant’ group reported significant discomfort even in these scenarios. EI did not significantly predict driving outcomes, likely reflecting limited score variability rather than an absence of influence. These insights support sustainable urban mobility by informing risk-based driver training and safer infrastructure design. Improving driver–VRU interactions helps create safer streets for walking and cycling, an essential condition for reducing car dependence and advancing sustainable transportation systems.

1. Introduction

Encouraging and supporting multimodal transportation is crucial in addressing issues related to congestion, network capacity, access, air quality, human health, and overall quality of life. However, important modalities such as bicycling and walking are inherently vulnerable to injuries and crashes with motorized vehicles. In 2022 in the US, there were 1105 cyclist fatalities and 46,195 injuries, which represents a 13% and 11% increase from the year before, respectively [1]. Similarly, pedestrian injuries and fatalities in the US are on the rise, with 7522 killed and 67,336 injured in traffic crashes in 2022, representing increases of 0.7% and 11% from the previous year, respectively [2]. Hence, current methods are not adequately protecting these vulnerable road users, and there exists opportunity to improve infrastructure, policy, and education.

Interactions between bicyclists and vehicles have been previously explored in the literature, particularly for vehicle overtaking behaviors. Walker & Robinson [3] found that drivers tend to pass bicyclists closer laterally for situations when the cyclist is wearing a helmet. Ampe et al. [4] performed an on-road study comparing lateral clearance distance by motorized vehicles to bicyclists with a child versus without and bicyclists with a trailer versus a child bike seat; they found that lateral distance was greatest for cyclists transporting a child compared to no child, and greater when passing a child bike seat compared to a child bike trailer. Many jurisdictions in the US and across the globe have imposed minimum passing distance (MPD) laws for passing bicyclists. However, compliance to these laws remains low. A survey study on 3769 drivers in Queensland, Australia found that over 31% of respondents self-reported to at least most of the time not complying to the MPD, where drivers aged 18–39 were even more likely to be non-compliant [5].

Beyond individual driver characteristics, contextual roadway elements strongly influence overtaking behaviors. Prior studies have shown that the presence of parked vehicles [6] and opposing [7] or adjacent [8] traffic can reduce lateral distance that drivers provide when passing cyclists. Meanwhile, wider lanes and overall roadway width have been associated with greater lateral passing distance [9,10]. Pavement condition and lane markings also play a moderating role. For example, faded or absent lane markings can lower passing speeds [10], but may simultaneously reduce lateral clearance [7]. Importantly, the type of bicycle infrastructure significantly shapes driver behavior, where drivers tend to pass with greater lateral space on roads protected bicycle lanes compared to painted bicycle lanes [11]. These environmental factors can interact with individual driver characteristics, potentially amplifying or mitigating unsafe overtaking tendencies.

Research has similarly explored interactions between pedestrians and vehicles, with an emphasis on drivers yielding to pedestrians at crossings. Video analysis of over 3400 pedestrian-motorist interactions at semi-controlled crosswalks (i.e., pavement markings and signs, but not traffic signal-controlled) found that the proportion of drivers that slowed down or stopped for pedestrians was 56.5% on one-way streets and 63.9% on two-way streets [12]. Similarly, Sucha et al. [13] performed an on-road study in an urban environment and found that a large proportion (36%) of drivers did not yield to pedestrians at marked but unsignalized crossings. Even more concerning, Goddard et al. [14] found that black pedestrians waiting to cross, compared to white pedestrians, were passed by twice as many vehicles and experienced 32% longer wait times. A further understanding of the underlying factors that influence drivers’ decisions to yield to pedestrians and passing behaviors to bicyclists can help develop more effective mechanisms to improve safety for these vulnerable road users.

One such approach is to investigate factors that lead to risky and dangerous driving. The association between personality traits and driving behaviors, for example, has been well explored in the literature. The five-factor model of personality, known as the Big Five, is a widely accepted measure of personality across five dimensions: neuroticism, extraversion, openness, agreeableness, and conscientiousness [15]. Broadly, higher levels of neuroticism have been shown to be a predictor of increased impulsivity and behavioral dysregulation, and decreased mindfulness and self-control [16]. Within transportation, neuroticism has been positively correlated with aggressive and risky driving [17,18,19,20,21]. Extraversion has had mixed findings, with positive associations with risky driving [21,22] but negative associations with aggressive driving [17,21]. Meanwhile, openness and agreeableness tend to be negatively associated with risky and aggressive driving [21]. Additionally, higher conscientiousness has shown to relate to lower mean speeds [23] and decreased likelihood of committing driving violations [19]. In a survey study specifically on bicyclists, O’Hern et al. [24] found that cyclists’ tendency to commit errors and violations was positively associated with extroversion and negatively with agreeableness and conscientiousness. While personality appears correlated with drivers’ behavior, these traits tend to be relatively stable and difficult to influence [25], and other factors may provide more actionable insights.

Alternatively, studies across several disciplines have shown that emotional intelligence (EI) can be improved through training [26,27]. This has potential in the transportation domain, as research suggests that EI and Big Five personality traits are correlated [28]. Specifically, EI correlates negatively with neuroticism [29] and positively with extraversion, openness, agreeableness, and conscientiousness [29,30]. Cavaness et al. [31] demonstrated that the Big Five personality traits can be enhanced by emotional intelligence. Additional research has even shown that EI can be more predictive than the Big Five for contextual performance [31] and compassion and self-compassion [32].

The relationship between EI and driving behaviors has been analyzed to some extent. Where higher EI is associated with decreased dangerous driving [33,34], decreased risky driving [35], and less anger while driving [36]. Research posits that lower emotional intelligence is associated with poor emotion control and regulation, which interferes with a drivers’ ability to make safe [35] and adaptive [36] behavioral decisions while driving.

Similarly, there is a relationship between the Big Five personality traits, driving behaviors, and risk propensity and tolerance. Joseph & Zhang [37] reported that extraversion, agreeableness, and conscientiousness were predictors of risk-taking, and that risk-takers tended to be extraverted, open to experiences, disagreeable, emotionally stable, and irresponsible. Similarly, Mathur & Nathani [38] concluded that agreeableness, neuroticism, and openness were strongly correlated with risk tolerance amongst investors. Several studies have evaluated the mediating role of risk on personality. In a study on construction workers, it was found that risk propensity mediated the relationship between extraversion, openness, and consciousness with unsafe behavioral intent [39]. Studies on risk in the financial sector demonstrate that risk-tolerance moderates Big Five traits in investment decisions [40,41]. Risk has also been applied in transportation, where high risk-takers are more likely to be involved in traffic crashes [42]. The link between emotion and risk in drivers shows that driving anger leads to risk-taking and traffic crashes [43]. Risk propensity research has also included other road users. Ahmed et al. [44] reported a relationship between likelihood to walk across railroad tracks and risk-taking. A survey study on 628 cyclists showed a significant correlation between cyclists’ anger, risky riding, traffic violations, and traffic errors [45].

While previous research has identified correlations between personality, EI, risk, and driving behaviors, much of the focus has been on general driving behaviors (e.g., speeding, vehicle-vehicle interactions) and predominately constrained to survey studies (i.e., self-reported measures). This paper seeks to address these gaps by directly examining how drivers’ EI and risk-taking tendencies influence their interactions with vulnerable road users, specifically bicyclists and pedestrians, within a controlled driving simulator environment. The following research questions are addressed in this paper: (1) how drivers’ risk-taking and emotional intelligence predict overtaking speed and lateral clearance when passing cyclists; (2) how cyclist location and group size affect drivers’ overtaking speed and lateral clearance; and (3) how drivers’ risk-taking and emotional intelligence influence their likelihood of yielding to pedestrians. Findings from this work provide evidence to guide practical safety interventions, such as driver education and licensing, risk-based training, and roadway design improvements. These safety improvements are central to sustainable urban development by encouraging active transportation, reducing motor vehicle dependence, and improving public health and environmental outcomes.

2. Materials and Methods

A driving simulator study was conducted. The study had approval from the Colorado State University Institutional Review Board (IRB), protocol #4265. Informed consent was obtained from each participant prior to data collection.

2.1. Participants

The sample comprised 40 participants who were recruited using flyers posted around Colorado State University. Inclusion criteria required participants to be at least 18 years of age and possess a valid US driver’s license.

There were 31 males, 8 females, and 1 prefer not to answer, with an average age of 29.8 years (min = 18, max = 52, SD = 8.1). The average length of time for having a driver’s license among the sample was 10.1 years (min = 0.75, max = 35, SD = 7.9). The highest level of education completed by the participants was as follows: less than high school (n = 1), high school (n = 1), some college (n = 5), bachelor’s degree (n = 12), and post-graduate degree (n = 21). Participants also reported their frequency of bicycling: 11 (27.5%) cycled daily, 6 (15%) weekly, 10 (25%) a few times per month, 12 (30%) a few times per year, and 1 (2.5%) never cycled. This information provides demographic context and insight into participants’ familiarity with cycling but was not included as a covariate in the models due to the limited sample size.

2.2. Driving Simulator

A fixed-based miniSim quarter cab (Driving Safety Research Institute, Iowa City, IA, USA) was used for the experiment (see Figure 1). The simulator provides 140 degrees of horizontal field of view via three 48-inch monitors. Participants first completed a practice drive in the simulator and then completed the 10 min study drive.

The simulation environment was built on the urban section of the Driving Safety Research Institute’s (Iowa City, IA, USA) Springfield Road Network and was customized to include added vehicular traffic, pedestrians, bicyclists, adjusted traffic signal timings, a midblock crossing, and on-screen navigation cues. Because only limited portions of the Springfield network feature bicycle lanes, we selected the urban route with dedicated bicycle lanes and sidewalks as the primary study segment and incorporated a turnoff onto an adjacent urban street without bicycle lanes. The resulting scenario represented a typical urban road environment with several traffic signals, light vehicle traffic, occasional pedestrians and bicyclists, and a posted speed limit of 40 miles per hour (mph). Participants first drove along a 4-lane (2-lane in each direction) road with dedicated bicycle lanes and sidewalks on both sides, followed by a 2-lane (1-lane each direction) road with sidewalks but no bicycle lanes. Weather conditions were programmed to remain clear and dry for the duration of the drive.

Participants received written navigation instructions displayed on the simulator screen about when to turn. They also received a written message saying, “Speeding Alert: Please Slow Down!” whenever they exceeded the speed limit by 10 mph.

During the drive, participants experienced nine bicycle passing events and one pedestrian crossing event. Specifically, each participant passed a single bicyclist riding in the bicycle lane three times, a single bicyclist riding in the right lane (no bicycle lane present) three times, and a group of six bicyclists riding in the bicycle lane three times. The order of these events was fixed rather than randomized, ensuring that each participant experienced identical scenarios for statistical comparability. In the data, each bicycle passing event was coded as 4 s before through 4 s after the vehicle passed the bicycle. Shortly after each interaction with a bicycle, a slow-moving vehicle was positioned in the left lane, such that participants would return to the right lane (i.e., adjacent to the bicycle lane) if they had changed lanes to pass the cyclists. There was never a vehicle immediately to the left of the participants while they were passing a cyclist, such that they could provide lateral space between their vehicle and the bicycle if they wanted.

During the portion of the drive along the segment of the 2-lane road, participants encountered a midblock crosswalk with two pedestrians waiting to cross and another pedestrian already three-quarters of the way crossed, on the other side of the street. Hence, they could continue through without hitting a pedestrian or stop and yield to the pedestrians waiting to cross.

2.3. Questionnaires

Prior to driving in the simulator, participants completed an emotional intelligence (EI) survey, which had 30 questions using a 7-point Likert scale. The survey was an adaptation of the Trait Emotional Intelligence Questionnaire—Short Form, TEIQue-SF [46,47] for use in the driving context, where this adaptation has been validated in previous work [33,48]. Participants were not told that the survey measured EI, but rather asking about their typical driving experiences. After completing the drive, participants completed one more survey, which collected information on demographics, risk propensity using the General Risk Propensity Scale, GRiPS [49], and about their general driving experiences, focusing on interactions with bicyclists.

2.4. Procedure

Participants first completed the EI survey on a computer. Then, they were told how the driving simulator worked, performed the practice drive, and then completed the study drive. Participants were told to drive the posted speed limit [of 40 mph], to drive in the right lane whenever possible, and that navigation instructions would appear on the screen. They were told the purpose of the study was to collect data on how people drive through a city environment. After the drive, they filled out the second survey [on risk and driving experience] on the computer.

2.5. Analytical Methods

Data analysis and cleaning were performed using Python (version 3.6.12) and RStudio (R version 4.3.2). Generalized linear models were used to predict driving behaviors based on risk-taking and EI. Specifically, linear mixed models were used to predict driving speed variables for the bicycle interaction events. A random intercept for participants was included in the mixed models to account for within-subject correlation and reduce measurement error, since each participant experienced repeated trials. Additionally, a binary logistic regression model was used to predict braking during the pedestrian interaction event. Lastly, chi-squared tests were used to compare differences in survey responses across risk groups.

2.5.1. Dependent Variables

A linear mixed model was fit on each of the following three dependent variables: average speed during bicycle passing event; speed at passing instance; distance to bicycle at passing instance.

• Average Speed During Bicycle Passing Event (mph): Average driving speed during the 8 s time interval during the passing event.
• Speed at Passing Instance (mph): Driving speed at the exact time when the vehicle passed the bicycle. This was extracted from the data based on the observation for which the minimum distance to the bicycle was the smallest.
• Distance to Bicycle at Passing Instance (inches): Minimum lateral distance of the vehicle to the bicycle(s) during the passing instance.

One binary logistic model was fit on the following one dependent variable: applied brakes for pedestrians.

• Applied Brakes for Pedestrians (yes/no): Whether the participant applied the brakes and slowed down to less than 25 mph when the crosswalk became visible.

2.5.2. Independent Variables

The following three variables were included as fixed effects in the models: risk-taking; total EI; and bicycle pasting event.

• Risk-Taking: Average score across risk questions (GRiPS survey), where a larger value indicates more risk-seeking tendencies.
• Total EI: Average score across EI questions (Driver Emotional Intelligence Scale, DEIS, survey), where a larger value indicates higher EI.
• Bicycle Passing Event: Three each of—bicyclist in bike lane; bicyclist on road; group of bicyclists in bike lane.

3. Results

3.1. Participant Risk and EI Characteristics

Risk-taking scores from the GRiPS survey were computed based on their average across the eight questions, which yields one total risk-taking value for each participant, with possible ranges from 1 (risk-adverse) to 5 (risk-seeking). In our sample, we observe total risk-taking scores ranging from 1 (n = 2) to 4.5 (n = 3), and a mean of 2.9 (SD = 1.1).

EI scores were computed as a composite score (“Total EI”) for each participant, based on the average of their responses across the EI questions, with possible ranges from 1 (lower EI) to 7 (higher EI). Observed values in this study range from 3.7 to 5.3 (mean = 4.7, SD = 0.4).

3.2. Braking for Pedestrians

A binary logistic model was fit to predict braking for pedestrians waiting to cross at a midblock crosswalk based on total EI and risk-taking (see

Table 1. Summary predicting braking for pedestrians.

Variable	Estimate	Std Error	t-Value	p-Value
Intercept	8.60	6.22	1.38	0.167
Risk-Taking	−0.92	0.45	−2.06	0.039
Total EI	−1.01	1.19	−0.86	0.393
Model Fit	LL	DF	Chi-Sq	p-value
Model	−20.7	3	5.7	0.05
Null	23.5	1	--	--

). Only fixed effects are included in this model, as there is only one pedestrian event per participant (i.e., no repeated measures). The model suggests that there is a significant effect of risk-taking, where participants are less likely to break (i.e., not yield) for increased risk-taking scores. Specifically, the odds ratio (exponent of β) indicates that for each one unit increase in risk-taking, the odds of braking decrease by a factor of 0.4 (95% confidence interval of 0.14 to 0.87).

3.3. Speeds When Passing Bicycles

Each participant passed a cyclist on the road three times, a cyclist in the bike lane three times, and a group of cyclists in the bike lane three times. These passing events were analyzed for the 8 s interval and the exact moment of passing for each event.

3.3.1. Average Speed During Bicycle Passing Event

A linear mixed model was fit to predict average driving speed during the passing event based on total EI, risk-taking, and passing event type (see

Table 2. Summary predicting average speed while passing.

Variable	Estimate	Std Error	t-Value	p-Value
Intercept	30.49	27.58	1.11	0.270
Risk-Taking	4.02	1.86	2.16	0.037
Total EI	1.09	5.59	0.19	0.847
Passing Event (reference level = bike in bike lane)
Bike on Road	−2.52	0.96	−2.62	0.009
Group of Bicyclists	−2.33	0.96	−2.42	0.016
Model Fit	AIC	LL	L Ratio	p-value
Model	2591.6	−1288.8	24.6	<0.001
Null	2608.2	−1301.1	--	--

). The model shows a significant effect of risk-taking on speeding, where each unit increase in the risk-taking score is associated with an increase of 4.02 mph during the passing event. There is also a significant effect of passing event type, where participants drive, on average, slower for a cyclist on the road (2.52 mph slower) and slower for a group of cyclists in the bike lane (2.33 mph slower), as compared to passing a cyclist in the bike lane. There is no effect of EI on average speed.

3.3.2. Speed at Bicycle Passing Instance

A linear mixed model was fit to predict driving speed at the exact time of passing based on total EI, risk-taking, and passing event type (see

Table 3. Summary predicting speed at passing instance.

Variable	Estimate	Std Error	t-Value	p-Value
Intercept	31.32	27.73	1.13	0.260
Risk-Taking	3.78	1.87	2.02	0.050
Total EI	1.61	5.62	0.29	0.776
Passing Event (reference level = bike in bike lane)
Bike on Road	−3.98	1.21	−3.30	0.001
Group of Bicyclists	−2.68	1.21	−2.22	0.027
Model Fit	AIC	LL	L Ratio	p-value
Model	2736.5	−1361.2	27.7	<0.001
Null	2756.2	−1375.1	--	--

). This model output is consistent with the previous model on average speed. Specifically, participants that are more likely to take risks are also more likely to drive faster at passing (3.78 mph increase for each unit increase in risk-taking score). Additionally, participants drive slowest when passing a single bicyclist on the road (3.98 mph slower) and slower when passing a group of bicyclists in the bike lane (2.68 mph slower) as compared to a single bicyclist in the bike lane. There is no significant effect of EI.

3.4. Distance When Passing Bicycles

A linear mixed model was fit to predict minimum distance to the bicycle(s) during each passing instance. In addition to the fixed effects of EI and risk-taking, speed at the passing instance is included as an independent variable to account for possible compensatory behaviors (see

Table 4. Summary predicting minimum distance at passing instance.

Variable	Estimate	Std Error	t-Value	p-Value
Intercept	16.70	12.55	1.33	0.184
Risk-Taking	0.53	0.86	0.62	0.537
Total EI	0.73	2.53	0.29	0.775
Speed at Passing	−0.10	0.04	−2.22	0.027
Passing Event (reference level = bike in bike lane)
Bike on Road	−1.53	1.15	−1.32	0.187
Group of Bicyclists	1.39	1.15	1.22	0.225
Model Fit	AIC	LL	L Ratio	p-value
Model	2646.2	−1315.1	15.6	0.008
Null	2651.8	−1322.9	--	--

). The model shows a significant effect of speed at passing on distance to the bicycle. Specifically, as driver speed increases, their passing distance to the bicycle decreases. This suggests that people drive faster when they are laterally closer to the bicyclists, compared to farther away [laterally].

3.5. Perceived Comfort Passing Bicycles

The associations between risk scores and stated preferences were further investigated based on the significance of risk on observed behaviors in the driving simulator. Specifically, we categorized participants into three distinct risk groups based on quartiles of the sample’s risk-taking scores: (1) Risk Avoidant (n = 14), average score below 25th percentile (i.e., ≤2); (2) Moderate (n = 14), average score between 25th and 75th percentiles (i.e., 2.01 to 3.99); and (3) Risk Taker (n = 12), average scores above 75th percentile (i.e., ≥4).

Participants were asked, in general, to rate their comfort level as a driver when passing a bicyclist in three distinct situations: on the road (sharing the lane), in the bike lane, and on the sidewalk. Figure 2 illustrates the differences in these perceived comfort levels across the three risk-taking groups.

These proportions were analyzed using chi-squared tests. For the risk-takers, there is a significant effect of scenario on their comfort level; where they are significantly more comfortable passing bicyclists on the sidewalk (χ² (2, n = 12) = 6.89, p = 0.032). Similarly, the moderates are also significantly more comfortable passing bicyclists on the sidewalk compared to the other passing scenarios (χ² (2, n = 14) = 13.73, p = 0.001). However, there was no relationship between passing scenario and comfort level for the risk-avoiders, suggesting that they were equally uncomfortable passing bicyclists across all scenarios (χ² (2, n = 14) = 3.61, p = 0.164).

3.6. Correlations Between Driving Behaviors

A correlation analysis was conducted on the driving variables (see Figure 3). The analysis was conducted individually within each of the three bicycle events. The variables for each were average speed during passing, speed at passing instance, distance at passing instance, and percentage of entire drive spent speeding. Across all interaction events, there is a significant and strong positive correlation between the speed variables (i.e., average speed during the event, speed at passing instance, and percentage of total drive spent speeding). This consistency indicates that drivers tend to maintain their speed as they approach and pass bicyclists, regardless of the type of interaction.

4. Discussion

This study examined how drivers’ risk-taking and emotional intelligence influence their interactions with vulnerable road users, specifically cyclists and pedestrians, under controlled urban driving conditions. The analyses revealed three main findings aligned with the research questions. First, higher risk-taking scores were associated with more hazardous behavior: each one-unit increase in risk-taking corresponded to a 4.02 mph increase in passing speed (p = 0.037) and a 60% decrease in the odds of braking for pedestrians (odds ratio = 0.40, p = 0.039). Second, the nature of the passing event significantly affected driver speed: participants drove 3.98 mph slower (p = 0.001) when passing a cyclist directly on the road and 2.68 mph slower (p = 0.027) when overtaking a group of cyclists compared to passing a single cyclist in a bike lane. Third, EI did not significantly predict passing speed or braking behavior, likely due to limited score variability within the sample. These findings support the study’s goal of understanding how individual risk profiles and roadway context influence driver safety, informing targeted interventions such as risk-based driver education, licensing, and infrastructure design. Safer interactions between drivers, bicycles, and pedestrians are a prerequisite for mode shift to cycling and walking, which lowers greenhouse gas emissions and congestion. Thus, risk-informed driver training and protected bike lanes are not only safety measures but also enablers of sustainable mobility and climate-resilient urban design.

The analysis, employing a binary logistic model, revealed a significant association between higher levels of risk-taking and decreased likelihood of braking for pedestrians. This aligns with Martha & Delhomme [50], who noted that high-risk takers often possess skewed perceptions of driving risks, potentially overlooking pedestrian-related dangers due to narrowed attentional focus. This behavior might stem from cognitive biases such as overconfidence in one’s driving abilities or underestimation of the unpredictability of pedestrian movements.

Recent advances in driving behavior modeling have leveraged machine learning and computer vision approaches to better detect and predict high-risk driving tendencies. For example, Liu et al. [51] developed a maneuver indicator and ensemble learning-based framework capable of recognizing risky drivers in complex highway merging areas with high accuracy, offering a data-driven method for identifying hazardous behaviors in real time. Similarly, Chen et al. [52] introduced a MetaFormer-based monocular metric depth estimation model for distance measurement in dynamic environments, improving environmental perception accuracy for automated systems. While our study adopts a behavioral modeling approach through psychological constructs rather than sensor-based detection, integrating such computational models with psychological profiling could provide a more comprehensive framework for understanding and mitigating driving risk.

In this study, we utilized linear mixed models to examine the dynamics between risk-taking behavior, types of bicycle passing events, and driving speeds. Our analysis identified a significant influence of risk propensity on driving speeds. Specifically, we observed that an uptick in the risk-taking score corresponded with an increase in driving speed, both during the overall interaction and specifically at the moment of passing bicyclists. This trend not only highlights the direct impact of risk-taking tendencies on speed regulation but also corroborates existing literature which suggests that individuals with higher risk-taking profiles are predisposed to engage in speeding [53]. This finding reinforces the notion that risk-taking is a key determinant in aggressive driving behaviors, potentially escalating the likelihood of crashes during interactions with vulnerable road users.

The presence of dedicated bike lanes appears to play a crucial role in shaping these behaviors. When cyclists are physically separated from motor vehicle traffic, drivers likely perceive a lower risk of collision, leading to reduced anxiety and a safer interaction dynamic. This hypothesis is supported by the work of Reynolds et al. [54], who noted that purpose-built bicycle-specific infrastructure, like bike lanes, significantly decreases the likelihood of crashes and injuries by providing a dedicated and predictable space for cyclists, thereby enhancing mutual road user predictability. Similar results were also shown regarding increased bicycle passing distance based on traffic lane width [55] and bicycle position from curb edge [56].

Further supporting this observation, Deliali et al. [57] found that drivers are less attentive to cyclists in scenarios where protected bike lanes separate them, likely due to the perceived security provided by the physical barriers. This could lead to a paradox where drivers, feeling overly secure due to the infrastructure, might pay less attention to potential crossing points or emerging cyclists. However, the overall impact of such infrastructure on road safety is undeniably positive, as evidenced by reduced crash rates and improved safety metrics. Moreover, a roadway design survey conducted by Sanders & Judelman [58] reinforces the notion that the presence of separated bike lanes not only alters driver behavior but also enhances their perceptions of safety. This perceived safety likely contributes to more cautious driving behaviors when drivers approach these well-defined and separated spaces. These findings underscore the importance of intentional urban and transportation planning that incorporates dedicated cycling infrastructure. Such infrastructure not only provides physical safety benefits but also psychologically impacts drivers, encouraging them to adopt more cautious driving behaviors near cyclists. This dual benefit is crucial for developing holistic road safety strategies that accommodate the growth of cycling in urban centers and ensure the safety of all road users.

Our findings also indicate that as the number of cyclists in a passing event increase, drivers tend to decrease their speed, suggesting a perceived higher risk when navigating around groups of cyclists. This behavioral adjustment may stem from drivers allocating more attention to the multiple dynamic factors present when multiple cyclists are involved. The presence of a group requires drivers to anticipate a variety of possible actions from different cyclists, necessitating a more cautious driving approach. Additionally, the variability in cyclists’ behavior within these groups may compel drivers to adapt their driving strategies more conservatively, effectively reducing their passing speed to accommodate the unpredictable nature of the group’s movement. This adaptation is crucial in complex traffic environments where the likelihood of sudden changes in the road scene increases.

Notably, most participants in our sample reported at least occasional bicycling, with over 40% riding weekly or daily. This relatively high level of personal cycling experience may have contributed to generally cautious and empathetic behavior toward cyclists observed in the simulator. Prior work suggests that drivers who also cycle are more attuned to cyclists’ vulnerabilities and may adapt their driving accordingly [59]. Such familiarity could partly explain the observed reductions in passing speed when drivers encountered cyclists on the road or in groups. Future research should examine cycling experience as a potential moderator of driver behavior.

Our findings should be considered alongside contextual factors such as traffic volume and roadway design. Research shows that wider lanes, dedicated or protected bike lanes, salient lane markings, and lower traffic volumes promote safer overtaking [6,7,8,9,10,11]. Although our simulator maintained moderate traffic and standardized lane configurations to control experimental variability, real-world conditions can amplify these effects.

Our study leverages the controlled environment of a driving simulator, which provides a high degree of control over experimental variables. However, there are still noteworthy limitations to our study design. While the driving simulator provided strong experimental control over confounding variables and safety risks, it cannot fully replicate real-world sensory and traffic dynamics. Combining simulator data with naturalistic driving or on-road tests would help confirm whether the observed behaviors persist under authentic traffic conditions. The experimental instructions, such as maintaining a specific speed limit and lane discipline, may also restrict natural driving behaviors that occur in less controlled environments. Additionally, the limited variability observed in emotional intelligence scores among participants may have contributed to the non-significant findings regarding its influence on driving behavior; however, future research should consider employing alternative emotional intelligence measures or broader sampling strategies to capture a wider range of EI levels and further clarify its potential role. Similarly, the study’s sample was relatively small (n = 40) and demographically skewed male with limited age and educational variability. These factors constrain generalizability. Future studies should include a larger and more demographically diverse sample to enhance external validity and account for potential confounding demographic variables. Lastly, the nine bicycle passing events followed a fixed order rather than randomized order across participants, which may have introduced sequence effects that influenced driver adaptation or expectancy over time. Randomizing event order in future studies would strengthen experimental rigor and help ensure that observed effects reflect true behavioral differences rather than presentation sequence.

5. Conclusions

Our study offers valuable insights into how risk-taking behavior influences driver interactions with bicyclists and pedestrians under urban driving conditions. Utilizing a driving simulator experiment, we establish a clear link between risk-taking tendencies and specific driving behaviors. It is notably apparent that drivers with higher risk-taking scores are less inclined to brake for pedestrians and exhibit increased speeds while passing bicyclists. These behaviors suggest a potential disregard for established safety norms and underscore the importance of addressing such tendencies in road safety programs. Moreover, our findings reveal that the nature of bicycle passing events, whether passing a cyclist on the road or a group in a bike lane, significantly affects driving speed. This indicates that drivers adjust their behavior based on the perceived complexity and risk of the traffic environment, showcasing a nuanced understanding of situational driving dynamics.

Although emotional intelligence did not significantly influence driving behaviors in this study, our findings highlight the value of targeted, evidence-based road safety interventions. Integrating knowledge of individual risk profiles with roadway design can help policymakers and safety advocates implement measures that directly reduce conflicts with vulnerable road users. In particular, risk propensity assessments could be incorporated into driver education and licensing to identify individuals prone to unsafe interactions with vulnerable road users, enabling early interventions for individuals predisposed to unsafe interactions with vulnerable road users. At the infrastructure level, expanding protected bicycle lanes and improving roadway design can create safer operating spaces, while tailored safety campaigns can discourage high-risk driving behaviors. Together, these strategies offer practical pathways to reduce crash risk and support safer multimodal transportation in urban environments. Such measures are key to advancing sustainable cities and communities.