Emotions arise from a combination of internal and external factors, such as personal events, environmental conditions, and social interactions, and they can substantially influence cognitive processes-including attention, decision-making-and subsequent behaviors. Recognizing how these affective states shape actions is particularly vital in high-stakes contexts like driving.

Appraisal has been widely adopted in emotion research, positing that people evaluate or “appraise” events to generate specific emotions (Smith and Lazarus, 1993; Roseman, 1984). Building on this, Russell's Emotion Circumplex model conceptualizes emotional states along two or three core dimensions-valence, arousal, and sometimes dominance (Russell, 1980).

These dimensions allow a spectrum view of emotion rather than focusing on discrete emotional labels. Accordingly, the Valence-Arousal-Dominance (VAD) framework has been applied to numerous settings, but the Valence-Arousal (VA) model often suffices for understanding essential affective processes (Juvina et al., 2018). Figure 1, adapted from Tsiourti et al. (2019), illustrates the conceptual placement of emotional states along these axes. Measurement approaches often rely on self-reported questionnaires to assess valence and arousal (Habibifar and Salmanzadeh, 2022). Researchers also induce emotions in participants through autobiographical recollection (Steinhauser et al., 2018), video clips (Zhang et al., 2016), or immersive simulations. This methodology has facilitated a substantial body of work examining how emotions, especially those characterized by high arousal, impact drivers' performance.

Numerous studies show that both valence and arousal strongly modulate driving outcomes. For instance, Du et al. (2020) found that positive emotions improved takeover performance in automated vehicles, while Zhang et al. (2022b) demonstrated that high-arousal negative emotions (e.g., anger) can lead to risky maneuvers. Similarly, background music with high-arousal characteristics enhances reaction times (Ünal et al., 2013), whereas heightened physiological arousal-often measured via heart rate (BPM)-can negatively affect social coping factors such as fear of negative evaluation (Sutherland et al., 2022).

Specifically, anger has been repeatedly associated with increased mean speed, greater speed variability, and more frequent errors compared to emotions such as fear (Zhang et al., 2022b; Roidl et al., 2014; Jeon et al., 2011; Abdu et al., 2012). Anxiety can undermine situational awareness, prompting failures to check mirrors or yield to oncoming traffic (Bernstein et al., 2019), while depression may lead to fatigue or lapses in concentration that elevate collision risks. Notably, around 16% of the general population reports moderate to severe driving anxiety-a high-arousal, negative-valence state that heightens concerns about road rage and lowers perceived driving safety (Taylor, 2018). Moreover, spatial anxiety, a domain-specific variation of anxiety, can exacerbate risky driving by leading to over-regulation and more frequent lapses behind the wheel (Traficante et al., 2024).

Collectively, these findings confirm that high-arousal negative emotions frequently result in more dangerous driving (e.g., speeding, erratic lane changes), whereas positive valence or low-arousal states can mitigate these risks (Hu et al., 2013). Building upon this literature, our study moves beyond simply linking emotions to driver performance by examining which stable individual differences-particularly personality traits-may contribute to these emotional responses and driving outcomes.

Personality traits have long been recognized as robust predictors of individual differences in both affective responses and behaviors (Taubman-Ben-Ari et al., 2004; Taubman-Ben-Ari and Yehiel, 2012; Kuppens et al., 2017; Steinbakk et al., 2019). In driving research, Eysenck's P-E-N theory (Psychoticism, Extraversion, Neuroticism) has been influential for explaining why some individuals are more susceptible to risky driving under emotional arousal (Eysenck, 1963; Tinella et al., 2021).

Driving behavior encompasses the broad range of actions drivers take, including speed choices, acceleration patterns, steering control, and decisions about lane-changing or following distances (Toledo et al., 2007, 2009). These outcomes can be linked to deeper psychological factors, such as individual risk tolerance, emotional states, and personality.

Research consistently shows that speed-related metrics (e.g., acceleration rates, average speed, speed variability) are predictive of traffic accidents (af Wåhlberg, 2004, 2006a,b). A driver's willingness to accelerate beyond set limits or fluctuate speeds significantly may indicate higher sensation-seeking or emotional reactivity (Wasielewski, 1984). For instance, individuals high in Extraversion or Neuroticism may be more susceptible to driving faster when emotionally aroused, either out of impulsivity (Extraversion) or anxiety-driven impatience (Neuroticism). Moreover, studies in bus driving contexts have shown that acceleration is positively correlated with accident likelihood (af Wåhlberg, 2000). Translating these findings to private vehicle contexts suggests that personality traits and emotional states could jointly influence how aggressively or smoothly a driver accelerates under stress or excitement (Li et al., 2022).

The driver's steering wheel performance is another essential aspect of driving behaviors, as it directly influences the vehicle's lateral movement (Zhang et al., 2022a; Pawar and Velaga, 2021). The measurement of steering control includes steering speed and steering swerve. Factors such as age, gender, and driving experience have been found to impact steering wheel performance significantly. For example, older non-professional drivers showed better steering performance (lower standard deviation of speed) in low traffic flow conditions but worse performance in high traffic conditions (Chen et al., 2019). Female drivers were found to have a higher steering reversal rate (SRR) than male drivers in car-following situations (Pawar and Velaga, 2021). Additionally, the same study reported that each year of driving experience led to a significant improvement in steering stability control (SSC).

In addition to examining drivers' attitudes toward acceleration and speed in emotional states, it is essential to consider other driving behaviors that may indirectly be influenced by the driver's emotional state (Du et al., 2020; Zhang et al., 2022b). Factors such as head vehicle distance, fluctuations in lane maintenance, and eye fixation or gaze behavior play pivotal roles in shaping the overall driving experience and safety. Steinhauser et al. (2018) demonstrated that anger influenced both the average and the variability of the distance to the leading car. Nevertheless, in the angry condition, a positive impact on braking behavior was observed. Cai and Lin (2011) suggested that the arousal and valence levels have significantly influenced the eye's attention to the surrounding environment. Moreover, when drivers were in high arousal states, the lane deviation performance was more dynamic (Cai and Lin, 2011).

Collectively, past studies underscore the critical role of emotional valence and arousal in influencing how individuals drive. High-arousal negative emotions, like anger or anxiety, often correlate with aggressive or distracted behaviors (e.g., speeding, abrupt lane changes), while positive or low-arousal states can mitigate these tendencies. Personality traits-particularly Extraversion and Neuroticism-emerge as key moderators in this process, making some drivers more vulnerable to risk-taking under emotional strain. However, the interaction between (1) personality traits, (2) emotional states, and (3) demographic or cultural variables remains insufficiently explored. Most existing research is Western-centric, leaving gaps in understanding how these factors might play out in populations like Chinese drivers.

Addressing these research gaps is crucial for developing comprehensive models of emotional driving. Therefore, in the present study, we incorporate moderated moderation and mediated moderation analyses to examine how age, driving experience, and personality traits (Extraversion and Neuroticism) converge under different arousal states to shape critical driving behaviors (acceleration, speed stability, and steering). By tackling these understudied aspects in a Chinese driving context, we aim to deepen the field's understanding of cross-cultural factors and offer more robust insights into emotional driving phenomena.

Emotions are widely acknowledged to influence driving behavior directly. However, their impact on the interplay between other human factors, such as personality traits and driver profiles, remains underexplored (Gross, 2013; Purnamaningsih, 2017). Building on prior research, we propose that Extraversion and Neuroticism may act as underlying factors shaping driving behaviors, particularly in high-arousal states (Song et al., 2021).

Drawing from studies on personality and emotion regulation, prior findings suggest that high-arousal states contribute to risky driving behaviors (Töre et al., 2023). However, it remains unclear how Extraversion and Neuroticism influence driving behaviors under different arousal conditions. Furthermore, driver profiles during emotional driving states have been largely overlooked. From a social psychology perspective, the influence of personality variations on driving performance in emotional states warrants attention.

Our study aims to uncover the psychological components underlying driving behaviors in high- and low-arousal situations. Specifically, we examine the relationships among drivers' profiles (age, driving experience), personality traits (Extraversion, Neuroticism), and driving behaviors (acceleration, speed stability, and steering performance).

Since driving experience does not show a clear causal relationship with personality traits, we focus on examining its moderation effects instead of mediation. This study primarily involves young and middle-aged Chinese drivers.¹ Although gender is an important factor in traffic research, the sample's gender imbalance limits analysis of gender differences; therefore, it will be included as a covariate to reflect real-world traffic dynamics.

The data was collected from a multi-modal human emotions driving study described in Li et al. (2022). The dataset comprised 40 drivers from China, of which 31 were male and 9 were female. All participants' data was used in this paper. The mean age of the 40 participants was 28.1 years with a standard deviation of 9.47 years. Figure 3 provides details on the demographic characteristics. On average, the drivers had 8.93 years of driving experience. According to the Traffic Management Bureau of the Public Security Ministry, by the end of 2022, China had 481 million car drivers, of whom 162 million were female-indicating a male-to-female ratio of roughly 2:1 (Song et al., 2024). Although our sample shows a somewhat higher ratio of male to female participants (about 3.4:1), it still partially reflects the overall gender distribution of drivers in China.

To ensure safety and reliability in the driving tasks, all participants were screened for normal or corrected vision and hearing and reported their health status, including any medical conditions, prior to the experiment. Additionally, participants were instructed to maintain a regular 24-h schedule, abstain from alcohol and stimulating substances before the experiment, and meet the inclusion criteria of having at least one year of driving experience with a valid driver's license. Each person was rewarded 200 RMB after experiments.

The simulator is a fixed-based driving simulator and is designed to provide a 270° field of view for the driver, with a half-cab platform and an automatic transmission. It includes a simulated rear-view mirror that allows the driver to keep an eye on the traffic behind them. In addition, the simulator has speakers that produce the sounds of the engine and ambient noise, as well as a woofer that replicates the vibration of the vehicle. The dashboard of the simulator is displayed on an LCD screen with a high resolution of 1,920 × 720 and a refresh rate of 60Hz. This screen displays car state parameters such as the speed and gear position.

The driving scenario consisted of a two-lane, two-way highway route with a total length of 3 km. Participants were instructed to drive in the right lane without changing lanes while maintaining a constant speed of 80 km/h. The oncoming road had a density of 3 cars per kilometer, and there were an average of 2 buildings per kilometer along the traffic route. The experiment employed a 20-inch central display simulator with a resolution of 1,280 × 1,024 and a refresh rate of 60Hz to present audio-video stimuli. The stimuli consisted of 7 videos, namely neutral, surprise, disgust, sadness, happiness, fear, and anger. Previous studies have confirmed the effectiveness of these videos in eliciting emotional arousal among drivers (Li et al., 2022). For example, the fear stimulus video depicted a car driving along the edge of a cliff in darkness, with its wheels stuck in the mud while the passengers made frightening noises after the car slowly began to move forward. All emotional stimulus video clips are shown in Figure 4. For each emotion stimulus, there is only one video clip. The procedure required drivers to first watch a video in its entirety and then commence their driving task. During the driving, the same video segment continued to play on a loop to maintain a consistent level of stimulation.

These emotions can highly reflect humans' emotional spectrum. In Eysenck's theory (Eysenck, 1983), neutral emotion is characterized by a lack of strong feelings or emotional intensity. It typically falls in the middle of the valence and arousal spectrum. Surprise is characterized by its high level of arousal. It can be either a positive or negative emotion, depending on the context. It often involves a sudden and intense reaction to unexpected events. Disgust is moderately arousing and is consistently negative in valence. It involves a sense of revulsion or repulsion, often triggered by something offensive or unpleasant. Sadness can vary in arousal from mild to moderate and is characterized by negative valence. It involves feelings of unhappiness, sorrow, and a lack of energy. Happiness can range from low to high arousal levels, and it is consistently positive in valence. It encompasses feelings of joy, contentment, and well-being. Fear is highly arousing and has a negative valence. It involves a strong emotional reaction to perceived threats or danger. Anger is highly arousing and carries a negative valence. It involves strong feelings of frustration and irritation. These abundant emotions help us to select high arousal state data.

The participants' emotional state was evaluated based on two dimensions: valence and arousal. The Self-Assessment Manikin (SAM) no-verb questionnaire (Bradley and Lang, 1994) was employed for this purpose. SAM is frequently used in conjunction with measurements of driving behaviors or physiological indicators to obtain a comprehensive understanding of emotional responses. The Self-Assessment Manikin (SAM) scale comprises three sets of figures, each representing a continuum along one of three affective dimensions: Valence, Arousal, and Dominance. In our analysis, we primarily focused on the Valence and Arousal dimensions, as these are more commonly used axes, while the Dominance dimension was not included. The Valence scale is depicted by a range of figures expressing emotions from smiling and happy to frowning and unhappy. Similarly, the Arousal scale is illustrated with figures ranging from excited and wide-eyed to relaxed and sleepy. As illustrated in Figure 5, the valence range spans from 1 (extremely unpleasant) to 9 (extremely pleasant), while the arousal range ranges from 1 (extremely excited) to 9 (extremely unexcited). Participants indicate their emotional state in response to a stimulus by selecting the figure that best represents their experience on each dimension. SAM's visual and intuitive format makes it particularly useful for studies involving participants with varying levels of literacy or language proficiency (Khosrowabadi et al., 2010; Koch et al., 2017).

Both qualitative and quantitative measures were recorded. After experiencing different driving conditions in a random order of a total of 7 videos, the participants selected their level of valence and arousal based on their subjective feelings. SAM was utilized to measure the participants' level of arousal and valence, with a scale ranging from 1 to 9. Moreover, the Eysenck Personality Questionnaire (EPQ) is used to measure drivers' personality traits (Eysenck and Eysenck, 1975). The EPQ used in this experiment includes 88 questions (Li et al., 2022) to access 4 dimensions of personality: E-Extraversion, N-Neuroticism, P-Psychoticism, and L-Lie. The questionnaire consists of a series of yes-no questions, each aimed at measuring specific aspects of these personality traits. Extraversion evaluates sociability and impulsiveness. Neuroticism is characterized by elevated levels of anxiety and depression, while Psychoticism is often associated with tough-mindedness and anger. Additionally, insincerity (i.e., L-Lie) is used to evaluate the validity and accuracy of the results. Our research primarily investigates the trait of Extraversion, often linked with positive emotions. Additionally, we examine Neuroticism, a trait typically associated with emotional stability and the propensity for anxiety. Psychoticism is not included in our analysis. The EPQ is renowned for its psychometric properties, including reliability and validity, making it a popular choice for personality research in various fields (Taubman-Ben-Ari et al., 2004; Du et al., 2018).

In addition, the simulator recorded multiple time series of car parameters, including acceleration, lateral acceleration, gas pedal position degree, brake pedal force, steering wheel position, and speed. Consistent with prior research, we focused on acceleration (af Wåhlberg, 2000), speed (Toledo et al., 2009), and steering wheel (Chen et al., 2019) as key indicators to represent the driving behavior and attitude. To represent acceleration behavior, we used the 95th percentile resulting acceleration, denoted as ACC_95th. Furthermore, we calculated the standard deviation of speed (Speed_std) as an indicator of speed stability, and the standard deviation of steering wheel speed (Steering_std) as an indicator of steering performance. At last, we applied log to all the driving behavior data. More details are provided in Section 3.5.

In this experiment, all participants were involved, and they were given instructions to maintain a regular 24-h schedule and avoid consuming alcohol or stimulating drugs before the experiment. Upon arriving at the testing place, each participant completed a demographic form and a health form to check their health state and to ensure they were fit for the experiment. The health form included requirements about health history, medication use, mental status, etc. After being equipped with the EEG device,² the participants were given a 10-minute training session to get used to the simulator environment and the signal recording process. The training took place on an 8 km stretch of highway with four lanes in both directions, where the drivers were asked to comply with the traffic laws. After the training, the participants completed a Simulator Sickness Questionnaire (SSQ) to ensure that the participants had no mental comfort like being fatigued or a physical illness (Singla et al., 2021). Following the training and information collection session, participants completed 7 emotional driving experiments in random order: neutral, surprise, disgust, sadness, happiness, fear, and anger. The driving scenario was a 3 km straight section of road with two lanes and oncoming traffic at a rate of 3 vehicles per kilometer. The road included 2 buildings in the field of vision every kilometer, and the weather was sunny with good visibility. Participants were instructed to drive at a speed of approximately 80 km/h in the right lane with no vehicles in front or behind. The 7 emotional induction materials were selected based on their efficiency as reported in Li et al. (2022). Participants were instructed to watch video and audio clips displayed on the screen in a random manner while maintaining emotional control as if they were driving.

After watching the videos, participants began driving, and data recording started. Following each emotional induction and driving session, participants completed the Differential Emotion Scale (DES) and the Self-Assessment Manikin (SAM). The DES, a multi-dimensional self-report questionnaire designed to assess individuals' emotional experiences (Izard and Izard, 1977), was employed to confirm the elicitation and intensity of specific discrete emotions (e.g., anger, sadness, surprise). The recorded data were considered valid only when the emotions reported on the DES aligned with those intended by the video segments. The SAM, on the other hand, was used to measure participants' Valence and Arousal levels-two core dimensions of emotional states that are widely studied in traffic psychology due to their direct influence on cognitive processes and driving behaviors (Du et al., 2020; Ünal et al., 2013). The Dominance dimension of SAM was excluded from this study, as it is less commonly included in driving-related research and was not directly relevant to the study's objectives.

Following each driving task, participants were given a 3-min pause to relax and regain a neutral state before encountering the next emotional stimulus. At the end of the experiment, participants removed the devices, left the simulator, and completed the EPQ survey. The experimental process is depicted in Figure 6. Two hundred and forty emotional driving experiments were validated (i.e., this included all 40 participants recordings). Forty emotional driving experiments were deleted because based on the results of DES, certain emotions were not shown as expected. The discrete emotion data from DES were not directly linked to the Valence and Arousal dimensions of SAM and were therefore not used in this study's analysis.

First, the participants' levels of arousal and valence were measured using the Self-Assessment Manikin (SAM) scale, which ranges from 1 to 9. The arousal states were categorized as high and low. We categorized instances where the arousal level exceeded the mean arousal of the samples (i.e., mean arousal = 5.3) as high arousal states, while the others were considered low arousal states. This approach not only aligns with prior studies such as Du et al. (2020), but it also simplifies the interpretation of emotional states, making the distinction between high-intensity (e.g., happy, angry) and low-intensity (e.g., sad, boring) emotions clearer and more actionable for analysis (see Figure 7).

The high arousal samples consist of 132 instances and the number of low arousal samples is 108 (i.e., from all male and female drivers), each containing questionnaire data, and multiple time series driving data collected during one experiment. As previously mentioned, the demographic characteristics used for analysis include age, and driving experience, which were obtained from the demographic questionnaire. The personality characteristics used in the analysis are Extraversion (E) and Neuroticism (N), as accessed by the EPQ. Driving behavior data was captured using the simulation equipment, with the original recordings containing parameters such as acceleration, speed, and steering wheel position. To stabilize the variance and handle heteroscedasticity, we employed log transformation on the output variables of acceleration, velocity stability, and steering stability. The logACC_95th (acceleration choice) was directly calculated using Equation 1. For each experiment, logSpeed_SD (speed stability) was calculated using Equation 2 from the driving data. logSteering_SD (steering performance) was obtained from the steering wheel position using Equations 3, 4. In the formulas, i represents the i_th moment in driving, and the Timeinterval represents the sampling interval of the simulator: 0.33 seconds.

To investigate the complex relationships between personality traits, driving experience, arousal states, and driving behaviors, we employed moderated mediation and moderated moderation models using Hayes's PROCESS macro.

The moderated mediation models with Extraversion as a mediator were tested using Hayes's PROCESS macro (model 14), as shown in Figures 7A–C. These models explored the relationships between Age, Extraversion, arousal states, and driving behaviors, alongside the covariate effects of Gender.

The results supported H1.1, demonstrating that Age positively predicted Extraversion (β = 0.113, p < 0.001), indicating that older drivers tend to exhibit higher levels of Extraversion. Furthermore, Gender significantly influenced Extraversion (β = −3.872, p < 0.001), with male drivers (coded as 1) scoring higher than female drivers (coded as 2). Together, Age and Gender accounted for 20.6% of the variance in Extraversion (R² = 0.206, F = 30.722, p < 0.001).

In support of H2.1, Age negatively predicted acceleration (log ACC₉₅ : β = −0.014, p < 0.001) and speed variability (log Speed_std : β = −0.006, p < 0.001), while positively predicting steering variability (log Steering_std : β = 0.009, p < 0.001). These results suggest that older drivers tend to drive more cautiously, with reduced variability in acceleration and speed but slightly increased variability in steering. Gender also influences driving behaviors. Male drivers exhibited greater acceleration (log ACC₉₅ : β = −0.075, p < 0.05), although no significant effects of Gender were observed for speed or steering variability (log Speed_std : β = −0.003, p = 0.932; log Steering_std : β = −0.019, p = 0.542).

The results further supported H2.3 and H3.1, as Extraversion mediated the relationship between Age and driving behaviors. Specifically, Extraversion was positively associated with acceleration (log ACC₉₅ : β = 0.013, p < 0.001) and speed variability (log Speed_std : β = 0.010, p < 0.001), aligning with theories that Extraverted individuals are more stimulation-seeking and risk-prone. However, Extraversion did not significantly affect steering variability (log Steering_std : β = 0.003, p = 0.278).

Moderated mediation analysis revealed that arousal states amplified the indirect effects of Extraversion on driving behaviors, supporting H4. In high-arousal states, the interaction between Extraversion and arousal significantly increased the indirect effect of Age on acceleration [ΔR2=0.014,F(1, 234)=4.431,p<0.05] and speed variability [ΔR2=0.020,F(1, 234)=5.261,p<0.05]. For log ACC₉₅, the pre-bootstrap conditional indirect effect was 0.002 (SE = 0.0006, 95% CI = [0.0007, 0.0033]), with post-bootstrap results refining this to a 95% CI of [0.0011, 0.0031]. Similarly, for log Speed_std, the pre-bootstrap conditional indirect effect was 0.0017 (SE = 0.0005, 95% CI = [0.0009, 0.0025]), while the bootstrapped results narrowed this to [0.0007, 0.0029].

These findings highlight the critical role of emotional states, particularly high arousal, in amplifying the influence of Extraversion on driving behaviors. High-arousal conditions were found to significantly increase variability in both acceleration and speed, emphasizing the dynamic interplay between personality traits and situational factors in shaping driving patterns. The detailed regression results (models 1–3), which showcase these interactions, are presented in Table 1. Summaries of the indirect effects for models 1 and 2 are provided in Table 2. Furthermore, a schematic diagram illustrating the moderated mediation relationships is shown in Figures 9, 10A offers a comprehensive visualization of the interaction among age, Extraversion, arousal states, and driving behaviors under various conditions.

The moderated mediation models with Neuroticism as a mediator were tested using Hayes's PROCESS macro (model 14), as shown in Figures 8D–F.

The models results showed that both H1.2 and H3.2 were not supported. While Age positively predicted Neuroticism (β = 0.046, t = 1.138, p = 0.256), this effect was not statistically significant. In contrast, Gender significantly predicted Neuroticism (β = 2.524, t = 2.642, p < 0.01), with female drivers (coded as 2) scoring higher than male drivers (coded as 1). Together, Age and Gender explained 2.9% of the variance in Neuroticism (R² = 0.029, F = 3.540, p < 0.05). Consequently, since the path from Age to Neuroticism was not significant, Neuroticism could not act as a mediator, thereby invalidating its role in supporting H3.2.

In line with H2.1 and results in Section 5.1, age negatively predicted acceleration (log ACC₉₅ : β = −0.012, t = −8.020, p < 0.001) and speed variability (log Speed_std : β = −0.005, t = −3.433, p < 0.001), while positively predicting steering variability (log Steering_std : β = 0.009, t = 6.690, p < 0.001). Gender negatively predicted acceleration (log ACC₉₅ : β = −0.111, t = −3.057, p < 0.01), indicating less stable acceleration patterns among male drivers.

H2.4 was partially supported. Neuroticism was positively associated with speed variability (log Speed_std : β = 0.005, t = 2.296, p < 0.05), suggesting that individuals with higher levels of Neuroticism experience greater speed fluctuations. However, Neuroticism did not significantly affect acceleration (log ACC₉₅ : β = −0.002, t = −0.894, p = 0.372) or steering variability (log Steering_std : β = 0.002, t = 1.024, p = 0.307).

H4 was not supported when neuroticsim played a mediated role. Arousal states did not significantly moderate the relationship between Neuroticism and driving behaviors. The interaction term between Neuroticism and arousal states was not significant for acceleration (β = −0.006, t = −1.154, p = 0.250), speed variability (β = 0.002, t = 0.457, p = 0.648), or steering variability (β = −0.006, t = −1.386, p = 0.167).

These findings highlight that, unlike Extraversion, Neuroticism plays a more limited role in influencing driving behaviors. However, its positive association with speed variability suggests its relevance for understanding individual differences in driving performance. The detailed regression results are presented in Table 3.

The moderated moderation models were established as shown in Figures 8G–L, which were tested using Hayes's PROCESS macro (model 3). These models explored the relationships between Driving Experience, Extraversion, Neuroticism, arousal states, and driving behaviors, alongside the covariate effects of Gender. However, no significant moderation effects of arousal states were observed in any of the models. Consequently, the focus shifted from the moderated moderation effects to identifying the moderation effects, which revealed a significant moderating role of Neuroticism in the relationship between Driving Experience and driving behaviors.

The results showed that H3.3 was not supported. Specifically, the interaction term between Driving Experience and Extraversion (X*M) did not significantly predict acceleration (log ACC₉₅ : β = 0.001, t = 0.359, p = 0.720), speed variability (log Speed_std : β = 0.002, t = 1.21, p = 0.227), or steering variability (log Steering_std : β = −0.000, t = −0.006, p = 0.995). Similarly, H4, which hypothesized a moderated moderation effect, was not supported, as the three-way interaction term (X*M*Z) was non-significant across all driving behavior outcomes.

Despite the lack of significant interactions, Driving Experience (X) consistently predicted key driving behaviors. In support of H2.2, Driving Experience negatively predicted acceleration (log ACC₉₅ : β = −0.019, t = −3.942, p < 0.001) and positively predicted steering variability (log Steering_std : β = 0.017, t = 4.243, p < 0.001). However, its effect on speed variability was not significant (log Speed_std : β = −0.010, t = −1.784, p = 0.076).

Extraversion (M) had a significant positive effect on acceleration (log ACC₉₅ : β = 0.011, t = 2.676, p < 0.01) and speed variability (log Speed_std : β = 0.011, t = 2.463, p < 0.05), supporting its association with greater fluctuations in driving behavior, as hypothesized in H2.3. These findings align with the conclusions of the previous sections, further highlighting Extraversion's role in driving patterns characterized by higher variability in acceleration and speed. Also, consistent with earlier analyses, Extraversion did not significantly affect steering variability (log Steering_std : β = 0.004, t = 1.159, p = 0.248).

Gender, as a covariate, had a significant effect on acceleration (log ACC₉₅ : β = −0.076, t = −2.009, p < 0.05), with male drivers (coded as 1) exhibiting less stable acceleration patterns compared to female drivers (coded as 2). However, Gender did not significantly predict speed variability (log Speed_std : β = −0.022, t = −0.667, p = 0.505) or steering variability (log Steering_std : β = −0.003, t = −0.108, p = 0.914).

These results suggest that while Driving Experience and Extraversion independently influence certain driving behaviors, their interaction effects and the moderating role of arousal states are not significant. The detailed regression results are presented in Table 4.

The statistic results of Hayes's PROCESS macro (model 3) were shown in Table 5. Since the moderation effect of Neuroticism was observed. The interaction term between Driving Experience and Neuroticism (X*M) significantly predicted acceleration (log ACC₉₅ : β = 0.002, t = 2.762, p < 0.01), speed variability (log Speed_std : β = 0.002, t = 2.491, p < 0.05).

Then, we reverted our analysis model to a moderation analysis of Neuroticism without the arousal states as variables. We used PROCESS model 1 to continue the test. The results of this analysis can be found in Table 6.

In support of H2.2, driving experience negatively predicted acceleration (log ACC₉₅ : β = −0.019, t = −7.596, p < 0.001) and speed variability (log Speed_std : β = −0.009, t = −3.891, p < 0.001). This indicates that experienced drivers exhibit greater stability in both acceleration and speed. Additionally, driving experience positively predicted steering variability (log Steering_std : β = 0.016, t = 7.080, p < 0.001), more experienced drivers exhibit greater fluctuations in steering behavior, which may reflect increased responsiveness or instability depending on the driving context. Collectively, driving experience explained a substantial portion of variance across these behaviors, with R² = 0.205, R² = 0.317, and R² = 0.506 for acceleration, speed, and steering variability, respectively.

The results also supported the moderation role of Neuroticism (H3.4) in the relationships between driving experience and both acceleration and speed variability. Neuroticism moderated the effect of driving experience on acceleration [ΔR2=0.040,F(1, 235)=11.809,p<0.001] and speed variability [ΔR2=0.057,F(1, 235)=14.973,p<0.001]. For steering variability, the interaction term was not significant [ΔR2=0.003,F(1, 235)=0.904,p=0.342], suggesting that Neuroticism had a limited moderating role in this domain.

From the bootstrapping results presented in Table 7, the moderation effect of Neuroticism revealed nuanced patterns. For acceleration (log ACC₉₅), the effect of driving experience was strongest under low levels of Neuroticism (β = −0.029, 95% CI = [−0.0383, −0.0204]), decreased at mean Neuroticism levels (β = −0.019, 95% CI = [−0.0243, −0.0143]), and further diminished under high Neuroticism (β = −0.009, 95% CI = [−0.0153, −0.0033]). A similar trend was observed for speed variability (log Speed_std), where the effect was significant under low (β = −0.018, 95% CI = [−0.0260, −0.0105]) and mean levels of Neuroticism (β = −0.009, 95% CI = [−0.0128, −0.0042]), but not significant under high Neuroticism (β = 0.001, 95% CI = [−0.0040, 0.0063]).

Gender, included as a covariate, had a significant negative effect on acceleration (log ACC₉₅ : β = −0.100, t = −2.760, p < 0.01), indicating that male drivers (coded as 1) exhibited less stable acceleration patterns compared to female drivers (coded as 2). However, gender did not significantly predict speed variability (log Speed_std : β = −0.041, t = −1.325, p = 0.186) or steering variability (log Steering_std : β = −0.022, t = −0.708, p = 0.479). Theses results align with the previous analysis.

These findings are visualized in Figure 11, which highlights the moderation effects of Neuroticism on driving behaviors, with a detailed scheme provided in Figure 12. The regression results, presented in Table 6, indicate that driving experience negatively predicted both acceleration and speed variability, suggesting that greater experience contributes to more stable driving behaviors. Neuroticism, however, significantly moderated these relationships, amplifying the effects of driving experience on both acceleration and speed stability, as shown in the bootstrapped results summarized in Table 7. Notably, drivers with lower levels of Neuroticism exhibited more pronounced improvements in stability with increased driving experience, while those with higher Neuroticism showed less stable patterns. However, Neuroticism's moderating influence on steering variability was minimal, reflecting its limited role in shaping this aspect of driving behavior. These results underscore the nuanced interplay between personality traits and experience in shaping driving performance.

Figures 11, 12 illustrate the interaction effects explored in this study. Below, we provide detailed interpretations of the visualized trends and their alignment with the study's hypotheses.

To summarize our analysis, we established several models to find the interactions between age, driving experience, personality, and driving behaviors under different arousal states. The statistical results suggested that our Hypothesis 1 was partly supported that the age of drivers was positively associated with Extraversion (H1.1 was supported), however, the relation between age and Neuroticism was not found (H1.2 not supported). As for Hypothesis 2, H2.1, H2.2, H2.3, and H2.4 were partly supported. Driver's age and driving experience predicted lower acceleration (lower logACC₉₅), higher speed stability (lower logSpeed_std), but larger steering instability (higher logSteering_std). Additionally, Extraversion was a positive predictor of higher acceleration (higher logACC₉₅) and higher speed instability (higher logSpeed_std). Neuroticism was found to be positively associated with higher speed instability (higher logSpeed_std). Moreover, The mediation role of Extraversion (H3.1 was supported) and the moderation role of Neuroticism were found (H3.4 was supported). At last, from H4, we found significant evidence that the arousal states moderated the mediation effect of Extraversion on the relation between age and driving behaviors (logACC₉₅ and logSpeed_std). In this way, H4 was suggested and Extraversion was sensitive to the variation of arousal states.

Our findings supported H1.1, demonstrating that age positively predicts Extraversion. This aligns with previous research suggesting that Extraversion, particularly its social dominance component, increases with age due to life experiences fostering independence, self-confidence, and assertiveness (Roberts et al., 2006; Ashton and Lee, 2016). These findings are consistent with the idea that older individuals exhibit greater social confidence and reduced social anxiety, as highlighted by Bleidorn et al. (2019).

However, H1.2 was not supported, as no significant relationship was observed between age and Neuroticism. This discrepancy with prior research, which often shows a decline in Neuroticism with age (Roberts et al., 2006; Stieger et al., 2021), may reflect uneven age distribution in our sample or cultural differences influencing emotional stability (Yik et al., 2002; Kuppens et al., 2017). These results suggest that further research is needed to explore this relationship using more balanced samples and culturally sensitive measures.

Age and driving experience emerged as pivotal factors influencing driving behaviors, reflecting distinct yet interconnected effects. Supporting H2.1, higher age was associated with reduced acceleration variability and improved speed stability, aligning with previous research that highlights older drivers' better emotional regulation and risk assessment skills (Harré et al., 2000; Taubman-Ben-Ari and Skvirsky, 2016). However, increased steering variability with age suggests adaptive adjustments rather than instability. Older drivers may employ more nuanced steering strategies to manage complex driving scenarios effectively, consistent with findings by Pawar and Velaga (2021) and Macdonald and Hoffmann (1980). These results reinforce the view that age contributes to safer driving behaviors but also indicates the complexity of its influence on different driving dimensions. While age significantly influences personality traits, particularly Extraversion, its role in shaping driving behaviors highlights both direct and indirect pathways through emotional and behavioral adaptations.

Supporting H2.2, greater driving experience was associated with reduced acceleration variability and enhanced speed stability, reflecting the benefits of experience in fostering smoother and more controlled driving behaviors (Crundall et al., 1999; Evans et al., 2022). Experienced drivers' ability to anticipate hazards and maintain consistent driving patterns underscores their superior adaptability in diverse traffic situations. However, the observed increase in steering variability with experience, similar to the effects of age, suggests that these adjustments may reflect deliberate and adaptive strategies rather than instability.

Extraversion and Neuroticism further modulated these effects, providing insights into how personality traits influence driving behaviors. Supporting H2.3, Extraversion was associated with higher acceleration and poorer speed stability, consistent with prior research linking Extraversion to stimulation-seeking and risk-prone behaviors (Eysenck and Eysenck, 1975). Drivers with high Extraversion levels may engage in behaviors that compromise stability, particularly in dynamic driving conditions. Conversely, H2.4 was partially supported: Neuroticism was associated with higher acceleration variability, suggesting a link between emotional instability and less controlled driving behaviors. However, its impact on speed and steering stability was less pronounced, highlighting the nuanced role of Neuroticism in shaping driving patterns directly.

In summary, the interplay between age, driving experience, and personality traits shapes driving behaviors in multifaceted ways. While age and experience generally promote safer driving styles, the influence of Extraversion and Neuroticism underscores the importance of considering individual differences in personality when evaluating driving performance. These findings contribute to a more nuanced understanding of the factors underpinning driving stability and risk, emphasizing the need for personalized approaches in driver training and interventions.

Extraversion played a significant mediating role in driving behaviors, supporting H3.1. The findings revealed that age indirectly influenced driving behaviors through Extraversion. Older drivers tended to exhibit higher levels of Extraversion, which in turn increased their acceleration and speed variability. This result aligns with Eysenck and Eysenck (1975)'s theory, suggesting that Extraverts are more stimulation-seeking and prone to risk-taking behaviors. Moreover, the moderated mediation effect highlights the interplay between Extraversion and emotional regulation (supporting H4), particularly under high-arousal conditions. In such states, the effects of Extraversion on driving behaviors were amplified, leading to heightened acceleration and speed instability. These findings are consistent with Brummer et al. (2014), who emphasized the significant role of Extraversion in shaping emotion regulation strategies.

The mediation pathway from age to driving behaviors via Extraversion underscores the complex influence of personality maturation on driving outcomes. Older drivers, characterized by higher social dominance and stimulation-seeking tendencies, may exhibit driving behaviors that reflect both cautious decision-making and occasional risk-taking. This duality aligns with the findings of Roberts et al. (2006), who noted an increase in Extraversion with age, particularly in traits like social dominance and self-confidence. Thus, while age positively contributes to personality development, the behavioral outcomes mediated through Extraversion may sometimes counteract the stabilizing effects of age on driving.

However, H3.3 was not supported, as the interaction between Extraversion and driving experience did not significantly influence driving behaviors. This suggests that the stimulation-seeking tendencies associated with Extraversion may not directly interact with the learning and adaptive processes facilitated by driving experience. According to Zuckerman (1995, 1991), structured environments such as driving may impose external constraints that limit the expression of Extraversion-related tendencies.

Conversely, Neuroticism did not significantly mediate the relationship between age and driving behaviors, contradicting H3.2. Although Neuroticism is typically linked to emotional instability and heightened stress sensitivity (Eysenck and Eysenck, 1975), its indirect influence was not evident in our data. This discrepancy may reflect cultural or methodological factors, as suggested by Stieger et al. (2021), who emphasized the role of context in shaping Neuroticism-related outcomes.

In contrast, H3.4 was supported, as Neuroticism moderated the relationship between driving experience and driving behaviors. Specifically, Neuroticism attenuated the positive effects of driving experience on acceleration and speed stability. Drivers with lower levels of Neuroticism benefited more from increased driving experience, exhibiting greater improvements in stability. Conversely, those with higher Neuroticism showed diminished benefits, reflecting the disruptive influence of emotional instability on learning and adaptation. These results align with Castillo-Gualda et al. (2019), who noted that Neuroticism is associated with emotional exhaustion, potentially hindering drivers' ability to effectively leverage their experience.

Together, these findings illustrate the dual influence of age and personality traits on driving behaviors. While age promotes the development of Extraversion, leading to riskier tendencies under high-arousal conditions, Neuroticism disrupts the stabilizing effects of driving experience. These insights underscore the importance of considering both age-related personality maturation and the dynamic role of emotional traits in shaping driving performance.

This study's focus on Chinese drivers offers valuable insights into how cultural contexts may influence the relationships among personality traits, emotional states, and driving behaviors. While our findings regarding Extraversion are consistent with global research, they also suggest potential cultural influences that require further investigation. For instance, comparative studies of drivers in Japan, China, and Vietnam have revealed differences in aberrant driving behaviors, with Vietnamese drivers exhibiting higher levels of such behaviors and Japanese drivers displaying fewer aggressive violations. These findings underline how cultural norms and societal values can shape driving behaviors in distinct ways (Hussain et al., 2019). Similarly, research comparing Chinese and Pakistani drivers has highlighted differences in dangerous and positive driving styles, emphasizing the importance of cultural contexts in understanding driving patterns (Yousaf and Wu, 2024).

Our observations of cautious driving tendencies among female drivers, though limited by sample size, align with broader studies linking gender norms and stereotypes to driving behaviors. For example, research has shown that adherence to male stereotypes is associated with increased risk-taking, whereas female drivers are often characterized as more cautious (Granié, 2024; Castro-Nuno and Lopez-Valpuesta, 2023). Although our findings cannot provide definitive conclusions, they highlight potential cultural and gender-related factors that merit further exploration.

Future research should adopt a cross-cultural approach to examine these dynamics in greater detail. Investigating cultural dimensions, such as individualism vs. collectivism, could shed light on how personality traits like Extraversion interact with cultural contexts to influence driving behaviors. Additionally, designing culturally specific interventions based on these insights could help enhance road safety while respecting local norms and values. Such efforts could bridge the gap between theoretical research and practical applications, contributing to safer driving practices globally.

Our findings offer valuable practical implications for improving traffic safety and advancing our understanding of the interplay between personality traits, emotional regulation, and driving behaviors. The observed moderated mediation effects underscore the importance of tailoring training programs to address individual differences in personality and emotional responses. Specifically, Extraverted drivers, who exhibit heightened risky behaviors under high-arousal conditions, would benefit from targeted interventions focusing on emotional regulation. Similarly, addressing Neuroticism in driving interventions could amplify the stabilizing effects of driving experience for drivers with high emotional instability.

From a theoretical perspective, these findings challenge traditional assumptions about emotional instability and driving behavior. While Neuroticism is typically associated with emotional instability, its limited role in our study suggests that cultural or contextual factors may shape the expression of personality traits in driving contexts. Conversely, the pronounced influence of Extraversion highlights the dynamic interaction between personality and situational factors, offering new insights into the complexity of driving behavior.

To enhance road safety, we propose two specific recommendations aimed at applying our research findings effectively:

By implementing these targeted recommendations, driver training programs can better prepare individuals to handle emotional challenges on the road, while advanced ADAS can enhance the safety features of next-generation vehicles. Together, these strategies offer a comprehensive approach to mitigating risky driving behaviors and fostering safer road environments.