Advances in intelligent driving technology have improved the mobility and independence of older drivers. However, under sudden low-visibility conditions requiring manual takeover, environmental perception is further limited among older drivers due to slower reaction times and reduced information-processing capacity, thereby increasing the risk of a takeover. This study focused on older drivers and established human–machine co-driving takeover scenarios across three representative visibility ranges (50–100 m, 100–200 m, and 200–500 m), including non-driving task conditions. A Dynamic Bayesian Network (DBN)-based takeover risk assessment model was developed to enable probabilistic prediction and inference-based evaluation under different conditions. Key factors influencing takeover risk were identified through model analysis. The results showed that takeover risk exhibited a dynamic pattern across all scenarios, characterized by an initial increase followed by stabilization. Under 50–100 m visibility, the takeover risk in older drivers increased by 1.32% and 2.66% compared with the 100–200 m and 200–500 m conditions, respectively. Engagement in non-driving tasks increased average takeover risk by approximately 3.77% relative to baseline driving. Across all conditions, takeover risk was higher in older drivers than in younger drivers, with a maximum intergroup difference of 1.61%. Among all variables, the velocity node showed the largest relative change value (ROV = 0.37), indicating that it was the most influential risk factor. Under combined low-visibility and non-driving-task conditions, takeover risk in older drivers tended to accumulate and remain elevated during the pre-takeover phase. In contrast, takeover risk in younger drivers increased more markedly when key indicators shifted from low-risk to high-risk states.
Patient safety culture (PSC) refers to the shared values, beliefs, and norms within a healthcare organization that shape the attitudes and behaviors of staff toward patient safety. It reflects how an organization values, perceives, and incorporates safety into its daily practices. PSC in a hospital setting has been directly associated with patient outcomes, the prevention of medical errors, and overall healthcare quality. Enhanced PSC has been related to reducing errors while improving the quality of care for patients. This systematic review addresses the assessment of PSC in a hospital setting using the structural equation modeling (SEM) techniques. A total of 26 studies were identified using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) protocol. The study identified important predictors, mediators, outcomes, and moderators of PSC. The results show that PSC interacts with different variables in complex healthcare networks. The results also suggest that better organizational and job-related conditions can have a positive impact on PSC, leading to improved safety-related behaviors, patient safety variables, psychological variables, and other outcomes.
In complex and dynamically changing human–machine collaborative environments, enhancing operators’ information processing efficiency and cognitive regulation remains a critical challenge. This study investigates the effects and underlying mechanisms of attentional guidance–based coordination interventions on cognitive efficiency, and evaluates whether multimodal physiological and behavioral indicators can reliably distinguish different intervention levels. Three experimental conditions were designed, including no intervention, low-level coordination intervention, and high-level coordination intervention. A multimodal assessment framework was established, integrating subjective workload ratings (NASA-TLX), behavioral performance, eye-tracking features, heart rate variability (HRV), and electroencephalography (EEG). Multivariate pattern analysis (MVPA) with permutation testing was employed, and a leave-one-subject-out (LOSO) cross-validation strategy was adopted. Results showed that attentional guidance interventions significantly reduced subjective workload and reaction time, while response accuracy remained stable. Under the LOSO framework, classification results revealed clear modality-dependent patterns: eye-tracking demonstrated the most consistent discriminative ability, particularly for distinguishing higher task demand conditions, while HRV showed moderate sensitivity to global cognitive workload. In contrast, EEG features exhibited limited discriminative power in subject-independent classification. Multimodal fusion improved performance, especially for conditions with larger cognitive differences, indicating complementary information across modalities. Notably, low- and high-level intervention conditions remained weakly distinguishable across all models. These findings provide new insights into the cognitive mechanisms of attentional guidance in collaborative tasks. This study offers practical guidance for ergonomics-oriented intervention design and evaluation in intelligent human–machine interaction systems.
Autonomous systems increasingly employ default settings to guide user decisions in repeated interactions, yet the proportion of rounds in which a system preselects a particular option, termed the default recommendation rate (DRR), remains largely unexplored. This study investigated whether DRR asymmetrically influences drivers' compliance and adjustment, and whether this effect is moderated by baseline decision preference and automation trust. Two studies were conducted using a multi-round binary decision task framed within a scenario involving trade-offs between comfort and energy conservation. Study 1 (N = 50) employed five DRR levels (0.1–0.9), and Study 2 (N = 64) refined the gradient to eleven levels (0–1.0) while incorporating measures of baseline preference and automation trust. Results showed that DRR exerted a significant positive effect on final compliance, yet this effect was asymmetric: behavioural adjustments converged toward a moderate level under extreme DRR values, with a group-level transition zone between DRR 0.3 and 0.4. Baseline preference shifted the intercept of the adjustment curve, energy-oriented users maintained higher compliance, but did not alter sensitivity to DRR. Automation trust did not moderate the DRR–compliance relationship, but positively moderated the association between baseline preference and compliance, amplifying intrinsic tendencies. These findings demonstrate that cumulative defaults serve as an implicit signal users integrate over repeated rounds to calibrate between system guidance and personal judgment, providing empirically grounded parameters, particularly the DRR 0.3–0.4 zone, for designing defaults that respect user autonomy in human-in-the-loop systems.
Procedures and checklists are essential elements of operational safety in various work domains. Multiple studies have identified the need for a standardized design methodology for domain-specific procedures and checklists due to inconsistent approaches. Consequently, there is an increasing need for widely accepted general methodology and guidelines. The purpose of this review is to summarize and synthesize the literature that has examined design considerations for procedures and checklists across multiple domains. As part of this review, two groups of findings were identified: definitions and key dimensions. First, definitions of procedure and checklist were summarized and critically evaluated in terms of five factors (purpose, format, utility, work environment, and expected outcome). Second, based on analyses of the literature, three key dimensions related to the procedure and checklist design were identified: the information, medium, and ecosystems dimensions. For each dimension, directions for future research efforts to improve the procedure and checklist design are proposed.
Monitoring workers' emotional states is a key objective of occupational ergonomics in safety-critical industries; however, systematic research on neurophysiological monitoring approaches in industrial worker populations remains limited. This study used functional near-infrared spectroscopy (fNIRS) to examine the feasibility of objectively monitoring high-arousal emotional states in high-risk workers via prefrontal hemodynamic signatures. A multi-level feasibility assessment was implemented, integrating behavioral validation, statistical significance testing, and effect size analysis. Fifty-five male coal miners underwent prefrontal fNIRS recording while high-arousal positive and negative emotions were induced using validated video stimuli, with resting state as the baseline. Hemodynamic responses across six prefrontal regions of interest (ROIs) were quantified using a general linear model. Both emotional conditions elicited significant prefrontal hemodynamic changes relative to resting state, with large effect sizes (Cohen's d = 0.807–1.243), supporting the sensitivity of fNIRS to emotional arousal. Five of six ROIs significantly differentiated between the two emotional states ( d = 0.32–0.55), with weaker prefrontal responses under negative emotion, suggesting reduced cognitive control resources that may link negative emotions to unsafe behaviors. The ventrolateral prefrontal cortex (VLPFC) exhibited the strongest hemodynamic response (d > 1.2) and the highest discriminability (d = 0.50–0.55); proof-of-concept classification (78.2% detection, 60.0% discrimination) confirmed its suitability as the primary monitoring target. A left-lateralized discriminability pattern was also observed, providing guidance for optimizing sensor placement. Based on these findings, a Detect-then-Discriminate (D2) monitoring framework is proposed, establishing portable fNIRS as a viable, non-invasive, and motion-tolerant approach for operator state monitoring in safety-critical work environments.
Motion sickness (MS) is commonly induced by inconsistency between vestibular and visual information and, in complex motion environments, may cause dizziness, nausea, and even vomiting. Effective methods for alleviating seasickness remain limited. This study proposed a non-invasive visual compensation system to reduce motion sickness by presenting visual cues consistent with the direction of the ship's vertical motion through an LED flowing-light pattern in the peripheral visual field. A six-degree-of-freedom (6-DOF) motion platform was used to simulate realistic ship motion, and the vertical motion component highly associated with motion sickness was transformed into peripheral visual flowing-light cues, thereby providing visual input consistent with vestibular motion information. Twenty participants were recruited. Subjective questionnaires and objective physiological measures were used to evaluate the system, and two central visual tasks were introduced to simulate real travel scenarios. The results showed that, when the visual compensation system was activated, the increases in skin temperature and heart rate relative to baseline, as well as the subjective questionnaire scores, were generally reduced. However, when visual compensation was removed in the subsequent phase, the lower levels of subjective and objective responses were not maintained, suggesting that the continuity and stability of peripheral visual cues may be important conditions for sustaining the intervention effect. This study supports the feasibility of visual compensation for alleviating motion sickness and provides preliminary empirical evidence for human factors design in maritime transportation.
This study investigates the auxiliary role of haptic vibration feedback in stabilizing control intentions during motor imagery (MI)-based operation of robotic arms in brain-computer interface (BCI) systems. In industrial human-robot collaboration (HRC) tasks, operators of conventional BCI systems overly depend on visual feedback. In hazardous industrial settings or for operators with limited mobility unable to adjust their viewing angles, this heavy reliance on visual cues frequently leads to unstable control intentions, ultimately undermining the precision of electroencephalogram (EEG) signal identification. To tackle this issue, a vibration coding scheme for three-finger was introduced, detailing spatial mapping encoding through the activation of single or multiple vibrators. This scheme is integrated into a wearable glove that facilitates real-time communication with the robotic arm. The effectiveness of the scheme was evaluated through two sequential experiments: Experiment 1 identified comfortable and distinguishable vibration intensities while exploring the coding parameter space under multi-vibrator stimulation. Experiment 2 assessed the scheme's efficacy in stabilizing MI control intentions. The results indicate that the scheme, optimized through Experiment 1, effectively stabilizes control intentions and enhances the reliability of EEG recognition, thus offering a promising haptic feedback solution for industrial BCI robotic arm systems.
Wearable technologies are increasingly used in health monitoring, but the specific usability and human–device interaction factors that shape long-term adoption remain insufficiently examined. This study investigates how cardiovascular sensor placement influences comfort, donning and doffing effort, skin interaction, and movement interference elements central to prolonged wearability. Four commercial heart rate wearables were evaluated across four popular placements (wrist, upper arm, chest, and ankle) during both routine and physical tasks.A comprehensive evaluation framework was developed, combining subjective physical and psychological comfort ratings, posture analysis, donning/doffing time, and skin assessments. From these data, a Wearability Score was developed and integrated with heart rate accuracy (Performance Score) to estimate overall user experience in a unified score using data-driven modeling as an analytical tool. Results showed that chest-worn ECG devices offered the highest measurement accuracy but scored lowest on wearability. Wrist and upper arm placements, on the other hand, provided a more balanced trade-off between wearability and accuracy.Our findings highlight that wearable device design should optimize measurement performance alongside user-centered physical ergonomics in order to support long-term use. The proposed framework offers a replicable approach for identifying sensor placements that maximize measurement accuracy without compromising physical and psychological comfort dimensions, usability aspects, and skin health-related concerns.
Evaluating the discomfort of seated occupants caused by whole-body vibration has been a challenging task due to the complex dependence on various factors related to the vibration and human characteristics. Existing studies typically added manually selected and extracted features into vibration discomfort evaluation models, which can be affected by vibration magnitude and seating conditions. To address the limitations, a vibration discomfort evaluation method with adaptive feature extraction based on deep learning is proposed. The method fuses the raw time-domain accelerations at the seat pan and backrest using a one dimensional-convolutional neural network, without the need for manual feature selection and extraction. Whole-body vibration experiments were then conducted and subjective responses regarding vibration discomfort were obtained. The performance of two different data fusion approaches on the vibration discomfort evaluation was compared, and later contrasted with the features-manually extracted vibration discomfort evaluation models. The results showed that the proposed method is able to extract adaptively discriminative and representative features from raw data, and accurately evaluate subjective discomfort for different vibration magnitudes and road excitations, achieving an accuracy of nearly 97%. This method demonstrates superior classification performance compared to models based on manually-extracted features, providing a more comprehensive and effective vibration discomfort evaluation of seated occupants.
Intelligent cockpits increasingly incorporate AI-enabled voice assistants to support goal-driven secondary tasks during driving. Yet when these interactions become effortful, they may be associated with negative emotions and less stable vehicle control. The affective process through which interaction burden is associated with driving behavior remains insufficiently understood. This study investigates whether interaction-evoked emotion can be modeled as a mediating construct linking voice-interaction burden and driving behavior in dual-task driving. A high-fidelity driving-simulation experiment with 31 participants was conducted, in which drivers completed navigation tasks through voice commands while driving. Behavioral, acoustic, and self-reported emotional data were analyzed using statistical methods and structural equation modeling (SEM). The results show that emotions arising during voice-assisted task completion differ from those typically examined in conventional traffic contexts. Lower interaction quality, indexed from interaction turns and completion time, was associated with greater frustration and lower perceived control; these affective responses coincided with less stable lane keeping and speed regulation. Speech features primarily reflected these emotional responses rather than exerting a direct effect on driving behavior. Overall, the findings support a theoretically specified user-experience pathway in which interaction-evoked emotion helps explain the association between interaction burden and driving outcomes. These results can inform the design of driver monitoring and in-vehicle voice interfaces by highlighting how speech cues may help detect emotional strain during voice-based secondary tasks.
Partially automated driving systems still rely on human drivers to monitor the environment and resume control when needed. Due to reduced vigilance and situation awareness during automation, drivers may misjudge the level of risk and the system’s capability to handle sudden conflict scenarios, which can undermine overall safety. Effectively communicating risk-predictive information generated by the automation through the Human–Machine Interface (HMI) can help drivers better understand system behavior and adjust their trust accordingly. In this study, we conducted a 3 × 2 × 2 within-subjects experiment involving intersection conflict scenarios (HMI: baseline vs. risk-alert vs. multimodal alert; risk level: low vs. high; turning direction: left vs. right). Thirty participants took part in the simulation-based experiment, with psychological measures, peripheral biosignals, neural activity, and driver-initiated takeover behavior recorded throughout the experiment. Results showed that risk-predictive information amplified the difference in drivers’ subjective risk evaluations between high-risk and low-risk conditions, and promoted more effective cognitive control in high-risk scenarios. The multimodal alert HMI (integrating visual and auditory risk prediction information) increased situational trust and reduced takeover frequency. High-risk and right-turn conditions led to higher perceived risk, lower trust, and more frequent takeovers. These findings provide new insights into how risk-predictive HMI designs influence drivers’ performance in partially automated driving, contributing to safer and more effective human–automation interaction.
This study presents a systematic approach to assessing and optimizing thermal experience in mobile gaming scenarios through machine learning methods. Through experimental investigation involving 67 participants and four smartphone models representing distinct device categories (Premium Standard, Gaming Flagship, Horizontal Foldable, and Vertical Foldable), the study developed optimized temperature monitoring strategies and established critical thermal thresholds. Analysis of eight temperature monitoring points revealed that front surface temperature served as the primary predictor of thermal sensation, accounting for 65% of model predictive power, with accurate thermal experience assessment achievable using three strategically placed monitoring points while maintaining 96% of full system accuracy. Random Forest and Support Vector Regression outperformed linear approaches in both thermal sensation regression (R2 = 0.773) and satisfaction classification (accuracy = 0.856); participant- and device-independent validation confirmed these models retain predictive value for unseen users and devices, with generalization varying most by device architecture for the foldable model. The research identified critical temperature ranges related to user thermal experience, with comfort zones typically below 38°C and discomfort zones generally above 45°C. These findings provide concrete guidance for thermal experience assessment strategies design and sensor placement optimization in mobile gaming devices. The research contributes to mobile thermal management by identifying optimal monitoring strategies, demonstrating the effectiveness and limitations of machine learning approaches, and establishing empirically grounded, context-dependent temperature thresholds.
Interdisciplinary engineering design requires rapid transitions from individual assessment to collaborative deliberation, yet the supporting neural mechanisms remain unclear. We utilized 22-channel functional near-infrared spectroscopy (fNIRS) hyperscanning to monitor 24 three-member teams (N = 72)—comprising structural, aesthetic, and environmental engineers—during a cruise-cabin design task. As participants shifted from individual scoring to group consensus, wavelet amplitude increased across prefrontal and inferior frontal regions, reflecting higher cognitive team workload, alongside a more integrated functional network. These neural shifts were significantly modulated by key behavioral factors: (1) higher decision quality was associated with selective decoupling between orbitofrontal and inferior frontal circuits, suggesting reduced affective bias; (2) greater prior expertise predicted lower right inferior frontal activation, consistent with neural efficiency; and (3) superior interdisciplinary communication capability (ICC) enhanced left inferior frontal activation while suppressing orbitofrontal–inferior frontal links. Furthermore, discipline-specific barriers shaped these dynamic responses: structural engineers facing dense jargon exhibited widespread prefrontal decoupling, whereas aesthetic engineers strengthened orbitofrontal–dorsolateral connections to reconcile subjective preferences. Ultimately, effective paradigm shifts in team decision-making rely on field-specific neural reconfigurations, linking neurophysiology directly to ergonomic outcomes like workload and communication efficiency.
Minimal work has investigated exoskeletons to support the neck, particularly in high-performing populations. One such group, pilots and aircrews, have high rates of neck pain and injury. When this group experiences high-g forces and uses extreme head postures, the risk increases. This work proposes a single participant pilot study on the development and evaluation of a dual-mode, multi-link exoskeleton prototype in normal and elevated g-level situations. A Human Centrifuge System was used to produce the elevated g-level environment. The performance of the system is measured with EMG sensors. In ’free’ mode, the prototype resulted in only minimal changes in muscle activity compared with the clean condition. At 2.5Gz, the ’damped’ mode of the exoskeleton reduced overall muscle activity by an average of 45.3% for four muscles compared to the clean condition. The outcomes of the pilot study support the feasibility of neck support exoskeletons for neck safety in high-performance settings, such as with pilots and aircrews.
Addressing the challenge of leveraging unstructured data for smart product-service systems (PSS) development, this study constructs a human-AI collaboration analysis framework integrating large language models (LLMs) reasoning with computational grounded theory (CGT) methodological constraints, thereby establishing an automated closed-loop for requirement attribution analysis. By employing LLMs to simulate the three-level coding procedure of grounded theory, this method transforms massive fragmented feedback into structured theoretical models, breaking efficiency bottlenecks while preserving the explanatory depth of qualitative analysis. Empirical analysis on automotive smart cockpit data shows the framework efficiently builds user experience attribution models with consistency and precision and identifies key design elements. Quantitative validation against expert benchmarks verifies the framework's effectiveness in complex semantic attribution tasks. The proposed framework converts fragmented feedback into structured design insights, offering methodological support for smart PSS optimization.
