Women experience distinct safety challenges when walking, cycling, and using public transport, particularly after dark and in relation to unwanted behaviours. Contemporary evidence integrating multiple transport modes and whole-journey stages across Australia and New Zealand remains limited. A cross-sectional online survey was administered to women aged ≥18 years in Australia and New Zealand who walked, cycled, or used public transport to and/or from work or education at least weekly. Descriptive statistical analyses examined travel patterns, safety perceptions, unwanted behaviours, reporting, precautionary strategies, barriers, and preferred improvements. Of 878 consenting respondents, 642 eligible women were included (M age = 42.6 years, SD = 13.3). Perceived safety declined sharply after dark across all modes. Only 25.2% felt safe walking after dark, and 34.4% cycling after dark. Infrastructure type strongly influenced perceptions: 72% felt safe in dedicated bicycle lanes compared with 16.3% on shared roads. For public transport, most women felt safe on board during the day (train: 76.8%; tram/light rail: 73.2%; bus: 78.6%; ferry: 69.4%), but far fewer felt safe after dark (30.7%; 34.1%; 34.2%; 38.7%). Perceived safety was particularly low when walking to/from stops (21.0–28.0%) and when waiting after dark (21.0–29.0%). Unwanted behaviours were common across journey stages, including unwanted staring or gestures, being followed, or made to feel unsafe, and abusive language, particularly on-board public transport. Only 21% agreed there were promising avenues for reporting concerns, and reporting rates were low (walking 13%; cycling 15%; train 18%; tram/light rail 20%; bus 17%). Women's perceptions of safety show clear whole-journey patterning by time of day, infrastructure, and journey stage. Improvements should prioritise lighting and surveillance, protected cycling infrastructure, staffed or monitored waiting areas, and clear, trusted reporting mechanisms.
Low-visibility environments (such as sandstorms and fog) pose a critical risk to road traffic safety. This study replicates typical sandstorm scenarios on the Uma Expressway using a driving simulator, systematically investigating the impact of six types of warning facilities (variable message signs, fixed warning signs, variable speed limit signs, guiding signs, voice navigation, and smart road studs) on driver behavior. First, considering the spatial characteristics of different warning methods along sandstorm-prone sections, nine warning combinations were designed to address different phases of the driving process. Next, driving behavior data under each warning scenario were analyzed from three dimensions: speed adjustment, driving stability, and compliance behavior. Finally, the non-integer rank RSR method was employed to rank the scenarios, determining the optimal warning scheme for low-visibility conditions. Experimental results indicate that different warning schemes significantly influence driver behavior in multiple dimensions. The scheme GK notably optimized driving patterns: average speed decreased by 63.7%, secondary acceleration frequency dropped to 33.33%, steady driving rate increased to 60%, and steering wheel angle variance reached the lowest fluctuation, achieving 100% lane-change anticipation. With an overall score of 0.946, it emerged as the optimal scheme. This suggests that providing effective warning information in advance on low-visibility highway sections is an effective measure to enhance service levels and ensure driving safety. The proposed framework for warning schemes can be extended to optimize traffic safety facilities under compound weather conditions such as fog and heavy rain, offering theoretical support and technical pathways for advancing highway proactive safety measures.
Professional bus and truck drivers are disproportionately involved in severe road crashes in Bangladesh, yet limited evidence explains how work conditions contribute to fatigue-driven exhaustion and unsafe driving intentions. This study develops and tests an extended Job Demands–Resources (JD-R) model to examine how occupational demands and resources shape fatigue-driven exhaustion and risky driving intention among professional drivers. Four demands were examined: shift duration and trip quotas, owner and dispatcher pressure, economic vulnerability, and owner-driver power imbalance. Four resources were assessed: rest stop availability, fatigue risk awareness and training, union and peer social support, and perceived regulatory enforcement. Data were collected through a face-to-face, enumerator-assisted survey of 439 professional bus and truck drivers across five major transport corridors in Bangladesh and analyzed using PLS-SEM and IPMA. The model explained 67.8% of fatigue-driven exhaustion and 42.5% of risky driving intention. Economic vulnerability and shift duration with trip quotas were the strongest demand predictors, while rest stop availability and fatigue risk awareness were the most important protective resources. Fatigue-driven exhaustion was the strongest predictor of risky driving intention, indicating that unsafe intentions are closely linked to cumulative occupational strain. The findings suggest that driver fatigue is shaped by income insecurity, scheduling pressure, limited control, inadequate recovery opportunities, and weak institutional support. The study interprets the trapped driver condition as a context-specific expression of the JD-R health impairment pathway and highlights the need for coordinated policies on wage protection, scheduling accountability, rest infrastructure, fatigue training, enforcement, and driver support.
Phone use while driving, particularly visual-manual phone engagement (VMPE) like texting, dialling, and browsing, can significantly increase crash likelihood due to prolonged diversion of attention from the driving task. However, significant knowledge gaps remain regarding the circumstances surrounding VMPE, whether different types of VMPE are initiated in different contexts, and whether different driver types have distinct VMPE patterns. Additionally, there is limited understanding of drivers' VMPE across entire trips. This study analyzed naturalistic driving data from 44 drivers collected over a three-week period. The data encompassed 557 full trips, driving kinematics, VMPEs, driving context, driver demographics, and psychosocial factors. The study aimed to predict the onset of VMPE as a function of driving context and driver-related variables. Results showed that the largest contributions to VMPE onset prediction was the interaction between driving context, and a set of driver-related variables (i.e., driver's attitudes, injunctive norms, and perceived behavioral control toward speeding and phone use). Specifically, drivers with positive attitudes, injunctive norms, and perceived behavioral control toward speeding and phone use were more likely to engage in VMPE in certain driving contexts, such as, while waiting in traffic, during morning drives with low traffic, and during trips that had limited idling. VMPE driver profiles revealed three types: Consistent Context-Independent VMPE, Selective Context-Specific VMPE, and Minimal VMPE. Together, these findings highlight the interplay between context and the driver's psychosocial factors, which influence their VMPE. They emphasize the importance of considering driver characteristics when designing interventions specifically aimed at reducing VMPE.
Some vehicle manufacturers have replaced physical driving-related controls (e.g., windshield wiper settings) with touchscreen controls. However, there is limited research investigating how this change may affect driver behaviour. We conducted a driving simulator study to compare touchscreen and physical controls for performing driving-related tasks across five systems: exterior lighting, windshield wipers, turn indicators, hazard lights, and climate. Tasks varied by length and were completed under lower demand and higher demand road conditions. We also conducted a preliminary exploration of whether driver behaviour is affected by touchscreen layout changes, which may occur due to an over-the-air software update. Thirty-two participants completed the study (17 M, 15 W; mean age = 34.8). For tasks requiring more than one button press, the touchscreen was associated with worse task performance (e.g., longer task completion time) and worse driving performance (e.g., higher speed and lane position variance, longer reaction time to lead vehicle braking) than physical controls. For all tasks, the touchscreen was associated with increased visual distraction. A minor layout change (switching the location of two controls on the screen) impaired task performance and increased long glances to the touchscreen for the first task after the layout change. These results can be used to inform touchscreen design guidelines and requirements for retaining task-appropriate physical controls for certain driving-related tasks. For example, our findings suggest that touchscreens should be avoided for any critical driving-related tasks requiring more than one button press. Further, if over-the-air software updates are used, manufacturers may want to restrict initial interactions while the vehicle is in motion.
Ensuring effective human supervision in conditionally automated driving is critical yet challenging, as directive alerts may increase attentional burden or foster passive compliance. This study examined whether nudges could encourage stated monitoring decisions without degrading user experience, and how hazard characteristics and individual differences modulate their effectiveness. Through a scenario-based experiment (N = 1032), we evaluated eight between-subject message conditions: four framing nudges combining perspective (self-focused vs. prosocial) and goal framing (gain vs. loss) in a 2 × 2 design, two social norm nudges (i.e., descriptive and injunctive), a reminder nudge, and a control. For within-subject factors, participants responded to hazards that varied in visibility (i.e., visible vs. invisible) and source (i.e., vehicles vs. vulnerable road users (VRUs)). Results revealed that self-gain framing, self-loss framing, prosocial-gain framing, and descriptive norm increased stated monitoring decisions relative to control, while the injunctive norm nudge performed consistently worse than other nudge conditions. Nudge effects on subjective evaluations varied by individual characteristics and hazard context: perceived safety improvements under multiple nudge conditions were primarily observed among participants with a High Big Five profile, and the acceptance gap between hazard sources observed in the control condition disappeared under nudge conditions. Hazard characteristics and individual differences also systematically shaped stated monitoring decisions and subjective evaluations. These findings offer design implications for adaptive human–machine interfaces that may support drivers' stated monitoring decisions from a human-centered perspective, contributing to the safe and scalable deployment of automated driving.
This research examines the determinants of cycling adoption intention among 983 non-bicycle users in Mexico City. We applied the Theory of Planned Behaviour (TPB) within an SEM-MIMIC framework, and using Bayesian informative hypothesis testing, we discovered key socio-psychological and contextual drivers. The results show that behavioural and normative beliefs are significant positive predictors, while being female and having long travel times are deterrents—with travel time proving to be the stronger negative influence. To translate the results into policy, we mapped the results were onto the Capability, Opportunity, Motivation, and Behaviour (COMB) framework. This analysis shows that increasing adoption requires a multi-pronged approach: urban cycling courses to enhance psychological capability and motivation, particularly for women, and financial incentives for e-bikes to improve physical opportunity by overcoming the spatial barriers of travel time.
Research on bus transit service quality relies on correlational factor models that cannot determine whether extracted dimensions represent autonomous constructs or manifestations of a single evaluative disposition. This study develops and internally validates a bifactor-based instrument for measuring user perceptions of intelligent transportation technologies in Curitiba's bus rapid transit system (N = 382, self-administered online questionnaire, quota-sampled). An integrated pipeline combining exploratory factor analysis, confirmatory bifactor modeling (WLSMV), ontological classification, and dual bootstrap validation was applied. The general factor accounted for 91.9% of the reliable variance (ωh/ω = 0.919). Of the four EFA-derived domains, two were reclassified as thematic after ontological classification; only Service Adequacy (ωS = 0.317 initial, 0.355 post-OC) and Fare Value (ωS = 0.261 initial, 0.311 post-OC) retained autonomous specific variance. Users recognized the city's technological capacity (M = 7.0) but rejected fare compensation (M = 2.9). Bootstrap validation confirmed structural stability (93.5% of converged replications achieving CFI ≥ 0.90). The resulting three-score architecture offers transit agencies a parsimonious monitoring framework grounded in the distinction between thematic groupings and structural dimensions.
Mobility as a Service (MaaS), as a pivotal solution for sustainable urban transportation, critically depends on user adoption and acceptance for its successful implementation. Existing MaaS adoption research leaves two issues insufficiently addressed. First, many studies explain adoption mainly through established technology-adoption constructs, giving limited attention to platform-specific perceptions such as perceived fairness and incentive mechanism effectiveness. Second, most empirical studies estimate average net effects and therefore provide limited insight into how different combinations of conditions can generate high adoption intention. To address these gaps, this study develops an extended unified theory of acceptance and use of technology (UTAUT) framework and combines partial least squares structural equation modeling (PLS-SEM) with fuzzy-set qualitative comparative analysis (fsQCA). Based on survey data from MaaS-aware early adopters in Beijing and Shanghai, PLS-SEM was used to estimate the net effects of eight latent constructs on intention to adopt MaaS (IAM), while fsQCA was used to identify sufficient configurations of the same conditions leading to high IAM. The PLS-SEM results show that performance expectancy, social influence, facilitating conditions, and price value positively affect IAM, whereas perceived risk has a negative effect. Notably, perceived fairness and incentive mechanism effectiveness did not show statistically significant net effects. However, the fsQCA results reveal seven sufficient configurations for high IAM, indicating that these two conditions can become important within specific causal recipes. The seven configurations are further interpreted as three configuration-based user archetypes: rational gain-oriented, community trust-driven, and fair transaction-oriented. The configurational results underscore that MaaS adoption is driven not by a single logic, but by multiple equifinal pathways, each associated with a distinct user decision-making paradigm. Therefore, effective platform strategies and policy interventions should shift from generic approaches to segment-specific designs that align with diverse adoption logics.
Autonomous taxis (ATs) represent a major AI-driven mobility innovation, yet public acceptance continues to lag due to concerns over safety, ethical accountability, and trust. This study develops and empirically validates a dual-process trust framework to explain how Chinese consumers form attitudes and adoption intentions toward ATs. Grounded in Dual-Process Theory and the Technology Trust Model, the framework differentiates between heuristic cues (brand signals, aesthetic appeal, social influence, and perceived anthropomorphism) that shape affective trust and systematic cues (algorithmic transparency, perceived safety, institutional trust, perceived control, perceived usefulness, and perceived risk) that shape cognitive trust. Both trust forms jointly drive consumer attitude, which in turn predicts adoption intention. The moderating roles of AI knowledge and consumer involvement in the attitude–intention relationship are also assessed. Survey data from 524 urban consumers were analysed using partial least squares structural equation modelling (PLS-SEM). The results largely support the dual-pathway mechanism and reveal distinct effects of heuristic and systematic cues on affective and cognitive trust. Importantly, AI knowledge negatively moderates the attitude–intention relationship, suggesting that more knowledgeable consumers may evaluate autonomous taxis more critically even when they hold favourable attitudes. By incorporating a human factors perspective, the study further links consumer trust formation to users' cognitive, affective, and perceived physical interaction with autonomous mobility systems. This study advances autonomous mobility research by integrating emotional and rational trust processes into a unified model and offers practical recommendations for enhancing public trust through improved transparency, human-centred design, and calibrated AI-literacy initiatives.
