Correcting the Cognitive Bias for Commuting Time to Relieve the Driving Stress Level in Snow Weather Condition: A Naturalistic Driving Study in Harbin, China

As a negative emotion, professional drivers’ stress levels signifcantly afected driving behavior and thus were related to driving safety issues. Nevertheless, current evidence fell considerably short of explaining whether and why private drivers’ stress levels might be infuenced while commuting driving in a specifc scenario and how to relieve their stress levels. Tis study aimed to identify and analyze the contributing factors of the drivers’ stress levels while commuting driving in various scenarios (clear or snow weather conditions). On weekdays between 1 st October 2020 and 31 st January 2022, the questionnaire data from a sample of 985 private drivers were collected from six diferent locations of business districts in Harbin, China. Based on the naturalistic driving study (NDS) database, a 7-item questionnaire was designed for participants to self-report their driving stress levels in various scenarios, which was generated from the shortened and adapted version of the Perceived Stress Scale (PSS). Te results showed that participants’ stress levels had signifcantly increased in snow weather conditions, especially nervous and stressed feeling, and unable to control the arrival time, which indicated that participants’ highly increased cognitive bias for commuting time could be the critical reason. Te results of hierarchical linear regression models indicated that overall stress scores could be predicted through participants’ sociodemographic characteristics, driving experience, commuting driving, and cognitive bias for commuting time. Such an association was signifcantly strongest with commuting time gaps, especially in snow weather conditions. In addition, a recommendation was derived from these results that correcting the cognitive bias for commuting time could relieve participants’ stress levels. Te implication of the reminder message supported this recommendation. Te participants’ stress levels were reduced signifcantly after providing a reminder message every 10mins while commuting driving in clear weather conditions and every 5mins in snow weather conditions.


Introduction
As one of the traditional adverse weather conditions, the snow weather condition is a signifcant cause of increased trafc accidents and compromised trafc fow in northern Europe and northern America [1].Also, the Trafc Administration Bureau of the Ministry of Public Security of the People's Republic of China (2021) proposed that between 2015 and 2020, snow weather conditions in northern China contributed to an average of 1085 fatalities, 3800 injuries, and 25% of total trafc accidents annually.Primarily during morning peak hours, snow weather conditions have long been known to contribute to the higher frequency of severe trafc accidents, due to the reduced visibility, the slick surface conditions, and the increased interaction between vehicles and pedestrians [2].
Moreover, the surge in trafc accidents reduced the transportation network's safety, mobility, and reliability, a signifcant priority of state departments of transportation and other transportation agencies in China and around the world [3,4].In practical terms, meta-analytical evidence suggests that most trafc accidents are preventable and caused by driving errors and trafc violations [5,6].Terefore, driving behavior as the most prevalent factor contributing to trafc accidents is needed to understand how the snow weather condition contributes to this driving behavior during the morning peak hour, why this driving behavior occurs, and how to reduce this driving behavior, which further reduces the snow weather-related trafc accidents during morning peak hours [7][8][9].
Some previous studies have concentrated on the links between negative emotions and unsafe/risky driving behavior.For example, anger and pleasure afected risky driving behaviors positively by enhancing the relationship between selfreported driving style (SDBS) and actual risky driving (ARD) behaviors, while surprise and fear weakened this relationship to afect risky driving behaviors negatively [10].Contempt afected risky driving behaviors positively by enhancing the relationship between self-reported sensation seeking (SSS) and ARD, while helplessness and relief weakened this relationship to afect risky driving behaviors negatively [11][12][13].Anxiety positively afected risky driving behaviors by synchronously enhancing the relationship between SDBS and ARD and the relationship between SSS and ARD [14].Furthermore, anger, anxiety, depression, contempt, fatigue, and other harmful and robust emotions led to stronger acceleration and higher speeds [15,16], breaking driving rules [17,18], crossing yellow trafc lights (e.g., [18]), and harder braking (e.g., [16]).Other previous studies have further emphasized the diferentiating roles of some specifc negative emotions on particular unsafe/risky driving behavior.For example, anger and hostility were related to aggressive driving [19], which may have been more sensitive to anger-based violations [20], including higher driving speed [15,21] and aberrant lane position [22].Anxiety and fear could negatively afect a driver's adjustment to changes in the driving environment, which can lead to a deterioration in driving behavior [23,24], including an increased risk of collision [25], more driving errors [23,26], increased reaction times in braking tasks [27], and greater likelihood of speeding [28].Moreover, fatigue decreases the driving behaviors of commercial motor vehicle drivers, resulting in the increased risk of crashes.Restricting the driving hours could mediate the causal path from fatigue to performance shortfalls to crashes [29].On the other hand, several studies have provided evidence that anxiety and fear due to experiencing near-misses or crashes can result in various problematic driving behaviors, such as slowing for green lights, driving far below the speed limit [30], increased speed compliance [31], avoiding nonessential journeys, and even avoidance typical of phobia [32].An alternative perspective that linked negative emotions and unsafe/risky driving behavior by considering and comparing the changing driving context, such as adverse weather conditions (e.g., rain, fog, and snow), driving time (e.g., daylight and nighttime, morning/evening peak hour or leisure time, and workdays or weekends), and driving purpose (e.g., work, shopping, and school), has received very little attention in the empirical literature [5].Tese specifc changing driving contexts represented the situational factors ignored by the drivers in planning behavior, which induced negative emotions and further afected the driving behavior [4].
In the changing driving context, experimental studies have shown that driving under stressful conditions led to adverse changes in physiological parameters, including increased arterial blood pressure [33], reduced heart rate variability (HRV) [34], and increased salivary stress hormones' concentrations (i.e., salivary cortisol levels) [35], which could be a predictor of unsafe/risky driving behavior [22].Moreover, several recent studies have documented stress as one of the traditional negative emotions, which could be a symptom of potential accident risk [6,36,37].Based on the transactional framework for driver's stress, these studies linked stress with unsafe/risky driving behavior through psychophysiological mechanisms (stress reactions) [6,38].Tis theoretical approach defned unsafe/risky driving behavior as transactional outcomes generated by interactions between drivers and the changing driving context [6].Stress processes were generated when the changing driving context exceeded the driver's coping capability [39][40][41].
From this perspective, most studies concentrated on the association pattern between personality traits and stress vulnerability [42][43][44].Drivers with positive personality traits (including optimism, enthusiasm, assertiveness, planning, and problem-solving orientation) were less vulnerable to stress and stress-related reactions and might be more able to quickly adapt to changing driving contexts without compromising driving behavior [45,46].On the contrary, drivers with negative experiences (e.g., neuroticism and negative afectivity) were more vulnerable to stress and more likely to react negatively to stressors (physically, emotionally, and behaviorally) [44,47,48].Tese fndings were usually derived from the occupational studies addressing the case of professional drivers (e.g., cargo/freight drivers, taxi drivers, and public transport drivers) [6,49,50].In consideration of their specifc task-related conditions, some specifc personality traits ft better with driving stress, which could be applied to predict unsafe/risky driving behavior in professional drivers accurately [51,52], such as physiological hyper-responsiveness to stress [53], exaggerated defensive engagement [54], and attentional biases [55].
Apart from professional drivers, stressful driving conditions also strongly afect private drivers' driving behavior [56].In particular, faced with snowfalls during morning peak hours, private drivers, as the major group of drivers, are required to continuously monitor surrounding trafc conditions and the route progress to make efcient and safe driving decisions [57,58].Teir driving tasks (e.g., arriving at the workplace on time) are engaging in nature, and the changed roadway-surface conditions combined with trafc congestion would easily make private drivers irritable [59,60] and stressful [61,62].Nevertheless, the case of private drivers' stress levels has remained unattended mainly, especially in view of this changing driving context.Tis study addressed this gap in the literature by examining how snow weather conditions infuenced private drivers' stress levels while commuting driving.Tis is one of the critical contributions of this study.
Self-reporting was deemed as the most common method for collecting the data on drivers' stress levels.However, by only using the self-reported questionnaires, drivers tend to forget their stress levels after some time.Terefore, naturalistic driving studies (NDSs) through the use of global positioning systems (GPSs), video cameras, accelerometers, and other in-vehicle technologies have made signifcant strides in capturing and recording drivers' behavior in the real world, which could solve this question.Drivers could drive as they normally would (i.e., without specifc experimental or operational protocols and not in a simulator or test track).Te period of observation can vary from several weeks to a year or more.In the last decade, NDS, such as the 100-car naturalistic driving study [63], LongROAD study [64], and the SHRP2 study [65], has presented an unparalleled opportunity for a greater understanding of driving behavior.Several studies have used the naturalistic driving studies to explain nuances of driving behavior, such as distracted driving behavior and collisions [66][67][68].Te second essential contribution of this study is to create a naturalistic driving study (NDS) database for getting back the participants' natural driving behaviors.By watching the commuting driving videos in various scenarios (clear or snow weather conditions) in the latest month, participants were asked to report their stress levels at that moment, which ensured the resulting self-reported data of stress levels were more reliable, valid, and accurate.
Although previous studies explored a broader range of the association pattern between personality traits and stressed vulnerability, it was unclear whether and why drivers felt more stressed or nervous while commuting driving in a specifc scenario (i.e., snow weather conditions and morning peak hour).In particular, relatively little is known about how drivers' psychometric properties were associated with their stress levels.Unlike personality traits, psychometric properties (such as the cognitive bias for commuting time) are related to the drivers' predisposition to perceive and react to the changing weather, road surface, and trafc conditions.In reality, considering all personality traits to conduct safety education for drivers is not feasible.Terefore, focusing on the targeted psychometric property, the evidence collected from this study (e.g., an application of a reminder message) can be benefcial for relieving drivers' stress levels (e.g., by correcting their cognitive bias for commuting time) while commuting or driving in various scenarios.It has further been considered as the contributing factor to reducing stress-related risky driving behavior, which is the third vital contribution of this study.

2.1.
Participants.For this study, a sample of 985 private drivers in the city of Harbin (Heilongjiang Province, China) was gathered.A random sampling method was employed for the participant selection.Te following inclusion criteria were used: (a) to be 18-55 years of age (people could apply for the driving license after 18 years of age and retire after 55 years of age); (b) to have a valid driver's license at the time of the study; (c) to own a vehicle; (d) to have a dashcam in the vehicle; (e) to drive to the workplace during the morning peak hour in both clear weather and snow weather conditions; and (f ) to have a fxed arrival time in the morning (compensation deductions for late arrivals).Te exclusion criteria included chronic or acute mental health disorders and/or physical diseases assessed using an authordeveloped questionnaire.All participants were anonymous and volunteered, and all responses were confdential.Any reward or compensation was not ofered.
Given that all participants have experienced snowfalls during their commuting driving, it can be concluded that snowfalls are the common experiences among residents who drive to/from the workplace in Harbin.From 1 st October 2020 to 31 st January 2022, the most recent snowfall during one's commute was relatively deep.For example, from 18 th November 2020 to 19 th November 2020, the average snowfall in 24 hours reached 25.5 millimeters.In the wintertime of 2021, the average snowfall reached 38.9 millimeters.Te questionnaire data suggested that in all participants, the most recent snowfalls often occurred on the way to the workplace, not on the return home journey.Terefore, in this study, we set the observation period to be the morning peak hour (07:00-09:00 AM) on weekdays from 1 st October 2020 to 31 st January 2022.

Study Variables and Measurement Instruments
2.2.1.Basic Information Questionnaire.From 1 st October 2020 to 31 st January 2022, a questionnaire survey was conducted on 985 participants who drove to the workplace during morning peak hours in both clear and snow weather conditions.In the survey administration, six diferent locations of business districts (Qiulin, Huizhan, Central Street, Aijian, Westred Square, and Wangfujing Department Store) were selected as survey location sites in the city of Harbin (latitude 44 °04ʹN-46 °40ʹN and longitude 125 °42ʹE-130 °10ʹE).Figure 1 shows the geographical location of the research area and the survey sites in Harbin.Te locations were selected based on areas with a large concentration of ofce workers with fxed working hours.Moreover, the locations' proximity to the workplace ends of commuting driving would increase the participation rate of the survey.
Te questionnaire was designed to identify the participants' sociodemographic characteristics (represented as gender and age), driving experience (represented as years licensed), and commuting driving (represented as drive frequency and commuting trip distance).At the start of the questionnaire, specifc attention was directed toward their fexibility to adjust arrival time at the workplace.Terefore, questions about whether employers penalized participants for arriving late were also included.Te participants with the fxed arrival time or compensation deductions for late arrivals were selected to continue the questionnaire.

Driving Stress Level Measurement.
To evaluate the participants' stress levels while commuting driving in various scenarios (clear or snow weather conditions), the shortened and adapted version of the Perceived Stress Scale (PSS) [69][70][71] was used.Tese items were combined with the suggested items from a panel of 50 experts (including 30 professionals in trafc safety and 20 trafc police ofcers).Trough focus groups with these experts and an extensive literature review, a 7-item questionnaire was designed.Moreover, the original PSS item was used to measure the degree to which situations in one's life were appraised as stressful [72].Te modifcation of the selected 7-item was carried out to consider the specifc driving condition.For example, an item such as "In the last month, how often have you been upset because of something that happened unexpectedly" was modifed to "When driving, how often have you been upset because of some unexpected driving events?"Tese modifed items were used to identify and measure how often the participants felt stressed and nervous while commuting driving in various scenarios, respectively.Te comparative table between the original 14-item of the Perceived Stress Scale (PSS) [72] and the modifed 7-item used for driving stress level measurement in this study is available in Table 1.
Te exclusive use of self-report measure increases the risk of biased results, since it just provides the participants' overall perception of their stress levels while commuting driving.However, it could not refect the participants' stress characteristics that fuctuate day-to-day.To overcome this limitation, naturalistic driving study (NDS) records the participants' driving behavior over the course of several days through the years, producing a large amount of data with a nested or multilevel data structure (i.e., multiple commuting driving trips within each participant).Trough observing the real driving behavior data during each commuting trip, participants could self-report their stress levels from both subjective and objective aspects to form a large amount of selfreported stress data for multiple commuting driving trips, which could further provide valuable information on variability in participants' stress levels while commuting driving in various scenarios (clear or snow weather conditions, with and without reminding the arrival time).No.

Original 14-item
No. In this paper, naturalistic driving data collected by the dash cams in participants' vehicles were used to conduct the naturalistic driving study (NDS).All participants had the dash cams (as the driving recorders) mounted on the windshield of their vehicles.As a video camera, these dash cams were used to monitor the commuting driving environment, track and record participants' natural driving behavior, and analyze their interaction behavior with the surroundings in various scenarios.All participants voluntarily provided the commuting driving videos in various scenarios in the latest month.All videos were assured of anonymity and confdentiality.Furthermore, more than 12000 commuting driving videos comprised this NDS database.
Tree expert analysts manually coded and inspected all videos in various scenarios (clear or snow weather conditions, with and without reminding the arrival time).All videos were grouped and named for each participant.Tese grouped videos were arranged based on random order, which might avoid the fxed and same viewing orders affecting the participant's judgment.Te participants received instructions on the frst day of training.Based on the NDS database, all participants watched their commuting driving videos in various scenarios and self-reported the 7-item questionnaire on a 5-point Likert scale where 0 � never, 1 � almost never, 2 � sometimes, 3 � fairly often, and 4 � very often.Each participant was assigned three videos in each scenario to self-report, and then a discussion was held to ensure that all items in the questionnaire were understood.A random sample of eight videos was processed by each participant in order to initially evaluate the consistency of participants' self-reported stress levels in various scenarios.Ten, the modifed 7-item of Perceived Stress Scale (PSS) was examined by viewing, coding, and analyzing video recordings.All items were averaged to obtain an overall stress score.Overall, it took three weeks to complete this selfreported questionnaire process.

Commuting Time Gap Measurement.
Trough the unstructured interviews (i.e., "Why you felt more nervous and stressed when driving to the workplace during the morning peak hour in the snow weather condition?"),we uncovered "hidden" information if and what might afect the participants' stress levels while commuting driving in snow weather conditions.Most participants were more worried that they did not know how long the commuting time will be increased in the snow weather conditions, which may induce compensation deductions for late arrivals.Tey all felt more stressed to rush to the workplace and were motivated to drive as quickly as possible.Terefore, worries about the inaccurately estimated commuting time would be associated with higher stress levels, which further induced a dysfunctional concentration in drivers' own thoughts.Tis distraction may, in turn, heighten drivers' risk and accident propensity.Te inaccurately estimated commuting time could be defned as the cognitive bias for commuting time, which was the commuting time gap between the perceived and actual commuting time.In the unstructured interviews, there was a section to identify the participants' cognitive bias for commuting time by measuring the commuting time gap.Trough watching their commuting driving videos in various scenarios in the latest month, these 985 participants were questioned (i.e., "When you were driving, how long did you think it would take you to get to workplace?") to assess their commuting time (as the perceived commuting time (PCT)) during each commuting trip.Furthermore, each commuting driving video was observed and analyzed by the expert analysts to record and calculate the participants' real commuting time (as the observed commuting time (OCT)) during each commuting trip.Ten, diferentiating between PCTand OCT with the commuting time gap could be used to express the participants' cognitive bias for commuting time.
To reduce most participants' stress levels, it was essential to examine the association between the cognitive bias for commuting time and stress level in snow weather conditions.

Driving Message Reminding the Arrival Time.
Various previous studies have proposed that a driving assistance system could improve the driving behavior and assist in avoiding safety-critical events, while it has been still vague whether and how this system could afect the drivers' stress levels.To investigate this, the reminder message, as a type of driving aid, was applied to adjust the baseline and comparative driving scenarios.Each participant received a reminder message reminding them of his/her arrival time while commuting driving in both adjusted baseline driving scenarios (clear weather conditions with a reminder of the arrival time) and adjusted comparative driving scenarios (snow weather conditions with a reminder of the arrival time).Furthermore, in each adjusted scenario, multiple reminding intervals (including 2 mins, 5 mins, 10 mins, and 15 mins) were set, which were used to explore what was the best time interval to disseminate a reminder message.Te navigation system in Baidu Map, a mobile phone-based app, captured the location of vehicles and sent a reminder message with a deep sound to each participant.Moreover, the dash cams in participants' vehicles collected the commuting driving videos in all adjusted scenarios.Te participants were questioned to report the 7-item scale by watching these commuting driving videos.Terefore, this driving assistance reminded the arrival time cloud to be used to understand whether the participants efectively relieved their stress levels when their cognitive bias for the commuting time was corrected.
2.3.Procedure.All participants were asked to voluntarily complete the questionnaires, which contained questions about their sociodemographic characteristics (represented as gender and age), driving experience (represented as years licensed), and commuting driving (represented as drive frequency and commuting trip distance).Te participants were informed of their rights and the protection of their personal information through an informed consent form.All participants were anonymous and volunteered, and all responses were confdential.

6
Journal of Advanced Transportation Various (clear or snow weather conditions) scenarios were defned to capture the stress level variations of participants while commuting driving.Tese scenarios involved a baseline driving scenario (clear weather condition) and a comparative driving scenario (snow weather condition).All participants were asked to voluntarily provide the commuting driving videos in these scenarios in the latest month.All videos were assured of anonymity and confdentiality.
As the naturalistic driving data, these commuting driving videos were collected by the dash cams in participants' vehicles.A 7-item questionnaire was designed to evaluate participants' stress levels while commuting driving in various scenarios.Trough watching their commuting driving videos in various scenarios in the latest month, participants were questioned about their stress-related emotional state by reporting these seven items on a 5-point Likert scale (0 � never; 4 � very often), and they also provided feedback about their perceived commuting time (PCT).
Te driving message reminded the arrival time with multiple reminding intervals (including 2 mins, 5 mins, 10 mins, and 15 mins) which were applied for the participants while commuting driving in four adjusted baselines (clear weather condition with reminding the arrival time) and four comparative (snow weather condition with reminding the arrival time) driving scenarios, respectively.Tese participants were questioned to report the 7-item scale by watching these commuting driving videos in all adjusted scenarios.

Statistical Analysis.
Descriptive statistics were obtained for the participants' stress levels and commuting time gap in various scenarios (clear or snow weather conditions).Internal consistency was estimated with composite reliability indexes (CRIs) to overcome the limitations of Cronbach's alpha.Te comparisons of participants' stress levels and commuting time gap among various scenarios were performed using paired sample t-tests, and p < 0.05 was considered statistically signifcant.
Hierarchical linear multiple regression analysis was constructed to evaluate the diferent validity of the adapted PSS for participants in various scenarios.In each driving scenario (including the baseline driving scenario in clear weather condition and the comparative driving scenario in snow weather condition), we designed the hierarchical linear multiple regression models to explore if the cognitive bias for commuting time (represented as the commuting time gap) could statistically and signifcantly explain the variances in overall stress scores beyond that explained by sociodemographic characteristics (represented as age and gender), driving experience (represented as years licensed), and commuting driving (represented as drive frequency and commuting trip distance) in both baseline (clear weather condition) and comparative (snow weather condition) driving scenarios.Terefore, we used the overall stress scores as criterion variables, and the participants' age, gender, years licensed, drive frequency, commuting trip distance, and commuting time gap as predictors.Four regression models were designed, respectively, following the same criteria: sociodemographic characteristics (represented as age and gender) were entered in the frst step, with driving experience (represented as years licensed) entered in the second step.In the third step, commuting driving (represented as drive frequency and commuting trip distance) were entered.Finally, in the fourth step, the cognitive bias for commuting time (represented as the commuting time gap) was entered.

Result
3.1.Characteristics of the Participants.For this study, a total of 985 participants were selected and gathered.Te minimum sample size (N � 315) was estimated using the statistical power analysis, with an anticipated efect size of 0.2, a statistical power level of 0.8, and a probability level of 0.05.Terefore, to ensure adequate statistical power for this study, we tripled the minimum sample size up to N � 985.
Among the sample, there were 537 males (54.5%) and 448 females (45.5%), and the age ranged from 20 to 53 years (M � 32.16, SD � 9.89).Regarding their driving experience, the participants had a valid driving license between 2 and 30 years (M � 9.58, SD � 8.49).Furthermore, regarding their commuting driving, the participants drove to/from the workplace between 1 and 20 times per week (M � 6.30, SD � 9.19), and commuting trip distance was between 20 and 280 km per week (M � 75.20,SD � 346.87).Te participants had no history of neurological or psychiatric disorders.

Self-Reported Stress Scores.
Table 2 provides the descriptive statistics of participants' self-reported stress levels, which showed the means and standard deviations of each item and the overall stress scores in various scenarios.When looking at the raw data on the Likert scale from 0 (never) to 4 (very often), the higher the items' scores of a participant, the more frequently he/she felt stressed when commuting driving.Cronbach's alpha coefcient values of the items ranged from 0.85 to 0.92 in the baseline driving scenario and from 0.83 to 0.90 in the comparative driving scenario.Tese overall stress scores had an alpha reliability of 0.88 in the baseline driving scenario and 0.85 in the comparative driving scenario.Te composite reliability index (CRI) of the overall stress scores was 0.86 and 0.85 in the baseline and comparative driving scenarios, respectively.Tese all indicated adequate internal consistency.Te diferences in overall stress scores were statistically signifcant between baseline and comparative driving scenarios at the p < 0.01 level.
Overall, in the baseline driving scenario (clear weather condition), participants had an overall stress score of 1.392, indicating a moderately low stress level for commuting driving.Te most frequently reported items were from unexpected driving events (no. 1, M � 2.397) while commuting driving.Compared with the baseline driving scenario, participants had signifcantly higher stress levels while commuting driving in a comparative driving scenario (snow weather condition).Tese results captured the efect of snow ) mins/km, p < 0.001) induced the higher commuting time gap (as the diferentiation between PCT and OCT) in the comparative driving scenario.Tese results suggested that the commuting time gap was prone to the snow weather condition, resulting in the increased participants' cognitive bias for commuting time.Moreover, the content analysis of interviews revealed that the work commitments, as manifested in fxed arrival time and compensation deductions for late arrivals, signifcantly impacted the likelihood of participants rushing to the workplace, especially in snow weather conditions.However, this impact was less favorable.Te highest increased "nervous and stressed feelings" and "unable to control the arrival time" caused most participants to overestimate the commuting time in snow weather conditions.In this situation, it was not surprising that these participants experienced elevated stress levels while commuting driving in the comparative driving scenario.4 summarizes the results of these hierarchical linear multiple regression models for predicting the overall stress scores in various scenarios.Te signifcant increase of R 2 (△R 2 ) in each step was analyzed, as well as the Akaike Information Criteria (AIC) [73], whereas R 2 has the advantage of being a standardized value and having a signifcance test associated, and the AIC has the advantage of considering both the contribution of independent variables and the complexity of the model [74].Both in the baseline and comparative driving scenarios, the AIC decreased in every step concerning the previous one, and the lowest AIC value was attained in the fourth step.Te lower the AIC value, the better the model ft.

Predicted Overall Stress Scores. Table
During Step 1, the results of each model showed that gender (1 � male and 2 � female) had positive and signifcant associations.Tese results indicated that female participants were more likely to feel stressed or nervous while commuting driving in both baseline and comparative driving scenarios.During Step 2, years of license were added to each model.Drive frequency and commuting trip distance were entered following age, gender, and years of license during Step 3. Furthermore, overall stress scores were found to be afected by both of these two added variables in both baseline and comparative driving scenarios.Specifcally, drive frequency signifcantly predicted the participants' stress levels while commuting driving, such that their overall stress scores decreased statistically and signifcantly as drive frequency increased.During step 4, the added commuting time gap induced R 2 to increase by 0.22 (△R 2 � 0.22 in the baseline driving scenario) or 0.31 (△R 2 � 0.31 in the comparative driving scenario), which suggested that the commuting time gap was the most crucial variable to statistically and signifcantly explain the variances in the overall stress scores.Te commuting time gap remained the most signifcant and positive association with overall stress scores in all models.Tis fnding indicated that participants with a more signifcant commuting time gap were most likely to feel stressed or nervous while commuting driving in both baseline and comparative driving scenarios.Terefore, the commuting time gap was shown to be the strongest predictor of overall stress scores, which confrmed that the participants' stress levels were statistically and signifcantly afected by their cognitive bias for commuting time.
By comparing the results of hierarchical linear multiple regression models in the comparative driving scenario with that in the baseline driving scenario, the considerable differentiation might detect whether and why the participants' stress levels were much higher while commuting driving in snow weather conditions.Tese results implied that participants were more sensitive to feeling stressed or nervous while commuting driving in the comparative driving scenario.Tis was because the efects of variables positively associated with overall stress scores increased more dramatically in all models than those negatively associated with Journal of Advanced Transportation overall stress scores, except drive frequency.In particular, the mean value of the commuting time gap, which had the most signifcant and positive association with overall stress scores during Step 4, increased by 1.38 times (1.1 mins/km) in the comparative driving scenario.After controlling other variables, only this increased commuting time gap was expected to increase the overall stress scores by almost four times in snow weather conditions.Terefore, the participants' highly increased stress levels while commuting driving in snow weather conditions seemed to be shaped by their increased cognitive bias for commuting time, meaning that the perceived commuting time tends to be increased more signifcantly than the actual commuting time because of snowfalls.

Corrected Cognitive Bias for Commuting Time.
Tese results ofered concrete evidence that the participants' cognitive bias for commuting time had a statistically signifcant impact on their stress levels in both baseline (clear weather condition) and comparative (snow weather condition) driving scenarios.After applying the reminder messages with multiple reminding intervals (including 2 mins, 5 mins, 10 mins, and 15 mins), Tables 5 and 6 provide the descriptive statistics of participants' self-reported stress levels in adjusted baseline and comparative driving scenarios, respectively, which showed the mean scores and changes (compared with the baseline and comparative driving scenarios) of each item and the overall stress scores.Cronbach's alpha coefcient values of these items ranged from 0.83 to 0.91 in four adjusted baseline driving scenarios and from 0.85 to 0.93 in four adjusted comparative driving scenarios.Tese overall stress scores had an alpha reliability from 0.86 to 0.89 in four adjusted baseline driving scenarios and from 0.87 to 0.90 in four adjusted comparative driving scenarios.Te composite reliability index (CRI) of the overall stress scores were from 0.80 to 0.86 in four adjusted baseline driving scenarios and from 0.81 and 0.85 in four adjusted comparative driving scenarios, respectively.Te overall stress scores were all statistically signifcant among the four adjusted baseline and comparative driving scenarios below the p < 0.01 level.
Results in Tables 5 and 6 suggested that the mean scores of all seven items and the overall stress scores were reduced in each adjusted driving scenario, which revealed that the participants with the reminder message had lower stress levels while commuting driving in both clear and snow weather conditions, compared to that without the reminder message.Terefore, the reminder message, which provided the arrival time information to participants, was expected to help participants correct their cognitive bias for commuting time, thereby relieving their stress levels.More specifcally, it was anticipated that the efect of reminder messages was more prominent in adjusted comparative driving scenarios.Te mean scores of all seven items (range: 0.18%-61.64%reduced) and the overall stress scores (range: 16.22%-24.80%reduced) were reduced more signifcantly in each adjusted comparative driving scenario, compared to those in each adjusted baseline driving scenario.In the routine driving situations, participants mainly relied on their rich experiences to accurately estimate the arrival time while commuting driving in clear weather conditions.Te commuting time gap was smaller and induced participants' lower overall stress levels.Terefore, the efect of reminder messages on relieving stress levels was less signifcant.
However, in snow weather condition, the uncertainty of road surface and trafc conditions might exceed participants' capabilities, resulting in a highly increased commuting time gap, which was signifcantly associated with increased stress levels.Terefore, the reminder message of arrival time could help participants minimize their commuting time gap while commuting driving in snow weather conditions, leading to signifcantly reduced stress levels.Such fndings were consistent with hierarchical linear multiple regression models during Step 4, in which the commuting time gap had the most signifcant and positive association with overall stress scores in the comparative driving scenario.
Due to the diferent reminding intervals in each adjusted driving scenario, the mean scores of all seven items and the overall stress scores were reduced diferently.For example, in both adjusted baseline and comparative driving scenarios, the mean scores of all seven items and the overall stress scores had the lowest reduction rate with 2 and 20 mins reminding intervals.Tis fnding refected that the participants might ignore the reminder message when the time interval was so long (20 mins).In comparison, the mental workload of participants might increase if the reminder message was provided too often (2 mins).On the other hand, the mean scores of all seven items and the overall stress scores maintained the highest reduction rate with 10 mins reminding interval in the adjusted baseline driving scenario and with 5 mins reminding interval in the adjusted comparative driving scenario.Many participants commented that this reminder message, which reminded them of the arrival time every 10 mins, was more comfortable and suitable for relieving stress levels while commuting driving in clear weather condition.Participants said that they felt more stressed and nervous while commuting driving in snow weather condition, thereby utilizing the same reminder message but more frequently (5 mins) to relieve their signifcantly increased stress levels.Journal of Advanced Transportation

Discussion
Several studies proposed the negative association between neuroticism and driving performance [43]; that is, the high neuroticism diminishes one's driving capacity by decreasing cognitive capacities and diverting attention to personal concerns rather than the driving environment [75].As known as cognitive interference, personal thoughts are likely to disrupt driving-related behaviors and may disable one's adjustment to the current driving environment [76].Te fndings of this study could support and extend this argument, which indicated that the cognitive bias for commuting time is the most signifcant contributing factor infuencing the drivers' stress levels while commuting driving, especially in snow weather condition.Drivers with higher cognitive bias for commuting time are disturbed by thoughts of potential compensation deductions for late arrivals and are stressed while commuting driving.In particular, the higher cognitive bias for commuting time induced the drivers to overestimate the commuting time more seriously in snow weather condition.Te efect of cognitive bias has a more permanent action on negative emotional activation (such as stress), which would have a more immediate and ephemeral impact [77].Repeated or chronic exposure to stressful situations leads to fatigue [78] and driving risk behaviors [79].
From this perspective, the evidence on the signifcant and positive relationships between cognitive bias for commuting time and stress levels in participants while commuting driving in various scenarios can be used in the design of safety interventions.Despite the growing evidence on the association between stress and driving risk behaviors, there are very few intervention studies focused on the cognitive bias and stress [80].Tis study developed a safety intervention (the reminder message, combined with the driver assistance system) to understand and manage the specifc stressors of drivers while commuting driving in various scenarios.Tis combined driver assistance system can efectively reduce the drivers' stress levels, which further optimizes their response while commuting driving.Especially, the safety implications of such a combined driver assistance system may be more evident in snow weather conditions.It can also be leveraged in some educational campaigns to be organized to release the drivers' stress by correcting their cognitive bias for commuting time.More attention should be paid to female drivers, with shorter driving years or less frequently to/from the workplace, who were more likely to feel stressed or nervous while commuting driving in snow weather conditions.In practical terms, these fndings support the design of organizational and individual safety interventions focused on correcting cognitive bias for commuting time to further reduce drivers' stress.As stress level and commuting time gap were measured and assessed using generic measurements and the driving environment is similar for professional drivers in the ground transportation industry of Harbin, this combined driver assistance system could be extended to design the occupational safety and health intervention for the professional drivers, focused on job strain management and WTC prevention.Moreover, future research could also be devoted to add this combined driver assistance system into the communication technologies, such as the vehicle-to-infrastructure or intervehicle communication systems, which can be leveraged in conditionally or fully autonomous vehicles.

Conclusion
As a traditional adverse weather condition, the snow weather condition severely impacts the safety, mobility, and reliability of transportation networks, especially during the morning peak hour.For example, multiple contributing factors such as roadway characteristics, environmental characteristics, crash characteristics, temporal characteristics, vehicle characteristics, and driver characteristics signifcantly infuence adverse weather-related crash injury outcomes [81].Several infuential factors such as the skidding vehicles, high-speed roadways, high engine capacities of vehicles, tree-related collisions, and pedestrian involvement have consistent efects on accident injury severities across various adverse weather conditions [82].Among these factors mentioned above, recent studies revealed that the drivers' stress levels, as a critical negative emotion, signifcantly afected their driving behaviors and were related to driving safety issues, such as trafc fow, visibility, and speed levels change correspondingly [83].Terefore, measuring the impact of snow weather conditions on private drivers' stress levels accurately was critical for improving trafc safety.Despite the growing evidence on the association between job stress and professional drivers' driving behaviors [84,85], very few studies have focused on private drivers' stress levels while commuting driving in snow weather conditions.In particular, no previous study has resolved whether and why private drivers felt more stressed and nervous in a specifc driving scenario (i.e., snow weather conditions and morning peak hour).Tis study aimed to shed more light on this question, and four main conclusions could be derived from this study.
First, the 7-item questionnaire was designed to evaluate the participants' stress levels while commuting driving in various scenarios (clear or snow weather conditions) after applying the shortened and adapted form of the Perceived Stress Scale (PSS).Based on the naturalistic driving study (NDS) database, the participants self-reported the scores of all seven items and overall stress scores in various scenarios, respectively.Tese results (i) identifed if the participants' stress levels difered in various scenarios and (ii) quantifed those diferences.In comparing the mean scores of all seven items and overall stress levels, statistical and signifcant increases were detected in the comparative driving scenario.Consistent with the previous literature [86][87][88], these results confrmed that the snow weather conditions afected the participants' stress levels while commuting driving in snow weather conditions.In particular, the fndings emphasized that some specifc items, involving nervous and stressed feelings (no. 3) and unable to control the arrival time (no.6), had a signifcant impact.
Second, the unstructured interviews were conducted to understand the reason for participants' statistical and 14 Journal of Advanced Transportation signifcantly increased stress levels in the comparative driving scenario.Te fndings implied that all participants' (N � 985) increased nervous and stressed feelings were due to their subjective perceptions of highly increased commuting time while commuting driving in snow weather conditions.Compared with the observed commuting time (OCT), the signifcantly increased perceived commuting time (PCT) induced a higher commuting time gap in the comparative driving scenario.Furthermore, the higher commuting time gap refected a highly increased cognitive bias for commuting time.Tese fndings supported the suggestion of other authors [89,90] that the work commitments (as manifested in fxed arrival time and compensation deductions for late arrivals) imposed a signifcant time constraint on participants, which directly triggered the greater cognitive bias for commuting time while commuting driving in snow weather conditions, thereby generating the increased stress levels' outcomes.Tird, hierarchical linear regression models were developed to capture the variables to predict overall stress scores and their relative importance to determine the most infuential variable in various scenarios.Te estimation results supported the suggestion of other authors [91,92] that age and years licensed were negatively associated with overall stress scores, while commuting trip distance was positively associated with overall stress scores in both baseline and comparative driving scenarios.Moreover, such associations were more signifcant in gender (with a positive association) and drove frequency (with a negative association).
Terefore, female participants with shorter driving years, who drove less frequently to/from the workplace, were more likely to feel stressed or nervous.Such feelings were stronger while commuting driving.In practical terms, these fndings supported the design of some organizational and individual safety interventions focused on specifc groups, which could relieve their stress levels, enhance their coping strategies, and potentially prevent risky driving behaviors [47,78].In particular, the commuting time gap was the most signifcant and positive predictor of overall stress scores, especially in the comparative driving scenario.As the most infuential variable, the participants' highly increased commuting time gap could signifcantly increase their stress levels while commuting driving in snow weather conditions.Terefore, some intervention strategies focused on correcting participants' cognitive bias for commuting time could efectively reduce their stress levels.
Fourth, considering the commuting time gap as the most infuential variable in predicting participants' stress levels, it could be critical to address and design efective intervention strategies to correct participants' cognitive bias for commuting time, further relieving their stress levels.It was known that the driving assistance system could help drivers achieve optimal or at least acceptable driving behaviors [93,94].Our fndings highlighted that the reminder message as a type of driving aid had an exciting and signifcant efect on reducing participants' stress levels.Te mean scores of all seven items and the overall stress scores were reduced in both adjusted baseline and comparative driving scenarios, compared to that without the reminder message.Specifcally, we found that, with the implication of diferent reminding intervals, the mean scores of all seven items and the overall stress scores were reduced to a greater extent in adjusted comparative driving scenarios.In addition, it was worth noting that the most signifcant efect of the reminder message was reminding the arrival time every 10 mins in adjusted baseline driving scenarios and every 5 mins in adjusted comparative driving scenarios.It was possible that these reminding intervals were more comfortable and suitable to help drivers correct cognitive bias for commuting time, thereby relieving their stress levels while commuting driving in both clear and snow weather conditions.
Finally, the current study has some limitations.Selfreported questionnaires were designed to evaluate participants' stress levels and perceived commuting time (PCT), which were difcult to observe and analyze objectively as it might result in social desirability bias.However, as anonymity and confdentiality were guaranteed, we expected low social desirability bias in our study.Tis study reassured the participants that all data would be anonymized and treated confdentially.Te questionnaire was designed to combine various instruments, a proven method of reducing social desirability bias [95][96][97].However, in terms of precision and reliability, the physiological measurements (i.e., heart rate, blood pressure, and brain wave) could be an efcient complementary and objective measurement for the self-reported questionnaire [98].Regarding future research, it is worth highlighting the use of both objective and subjective measures for assessing drivers' stress levels, which could minimize the social desirability bias typically from the self-reported questionnaires [99].Furthermore, the naturalistic driving study (NDS) database has been used to get back the participants' natural commuting driving behaviors in the latest month, which could help participants self-report their stress levels and perceived commuting time (PCT) more accurately.However, based on the retrospective evaluation, participants' previous evaluations might afect the following judgments.Terefore, for each participant, these commuting driving videos in various scenarios were arranged based on a random order to avoid this bias.In addition, the sampling substantially afected the generalizability of the fndings on a population level.Tis study was conducted with commuting drivers only in Harbin, China, not addressing other specifc groups (such as professional drivers).Te current results should also be verifed in other cities or countries.Finally, future research is worth exploring the diferent impacts of other possible explanatory variables specifc to snow weather conditions and road surface conditions on drivers' stress levels and the infuence of diferent driving scenarios (such as congestion and trafc accidents) on the commuting time gap.

Figure 1 :
Figure 1: Te geographical location of the research area and survey sites in Harbin.

Modifed 7-item 1
In the last month, how often have you been upset because of something that happened unexpectedly? 1 When driving, how often have you been upset because of some driving events that happened unexpectedly? 2 In the last month, how often have you felt that you were unable to control the important things in your life? 2 When driving, how often have you felt that you were unable to control the vehicle? 3 In the last month, how often have you felt nervous and stressed?3 When driving, how often have you felt nervous and stressed?4 In the last month, how often have you dealt successfully with irritating life hassles?5 In the last month, how often have you felt that you were efectively coping with important changes that were occurring in your life? 4 When driving, how often have you felt that you were unable to cope with the changing trafc conditions?6 In the last month, how often have you felt confdent about your ability to handle your personal problem?7 In the last month, how often have you felt that things were going your way? 8 In the last month, how often have you found that you could not cope with all the things that you had to do? 5 When driving, how often have you found that you could not cope with the driving events?9 In the last month, how often have you been able to control irritations in your life? 10 In the last month, how often have you felt that you were on top of things?11 In the last month, how often have you been angered because of things that happened that were outside of your control? 12 In the last month, how often have you found yourself thinking about things that you have to accomplish? 13 In the last month, how often have you been able to control the way you spend your time?6 When driving, how often have you felt that you were unable to control the arrival time?14 How often have you felt that difculties were piling up so high that you could not overcome them?

Table 1 :
Original 14-item of the Perceived Stress Scale (PSS) and modifed 7-item.

Table 2 :
Descriptive statistics of participants' self-reported stress levels in various scenarios.

Table 3 :
Descriptive statistics of the commuting time gap in various scenarios.

Table 4 :
Hierarchical linear multiple regression models for predicting the overall stress scores in various scenarios.

Table 5 :
Descriptive statistics of participants' self-reported stress levels in adjusted baseline driving scenarios.

Table 6 :
Descriptive statistics of participants' self-reported stress levels in adjusted comparative driving scenarios.