In tunnels, roadside auditory warnings provide an important safety communication channel for drivers. However, warnings transmitted at the intended volume may not remain clear when reaching drivers. This study examined the effects of broadcast voice, volume, and ambient noise on warning detection, reaction time, speech intelligibility, comprehension, listening effort, and behavioural intention using an audiovisual simulation based on in-situ tunnel recordings. Results showed that the female voice used in this study was associated with better warning detection and speech intelligibility than the male voice. Higher broadcast volume mainly improved detection, whereas high ambient noise reduced speech intelligibility. Speech intelligibility remained low overall and was strongly associated with information comprehension. Higher comprehension was also relevant to higher behavioural intention. Listening effort showed no consistent association with comprehension or behavioural intention. These findings highlight speech recognition as a major limitation and support driver-level evaluation of tunnel warnings.
The construction industry continues to experience high rates of occupational injuries, with slips, trips and falls (STFs) being the main causes. This study reviewed the current research on construction STFs and identified key influencing factors. A bibliometric analysis of 346 papers (2016-2025) revealed remarkable growth in this research area. Keyword analysis highlighted safety management, fall prevention and risk assessment as the main research focuses. A scoping review of 82 relevant papers was conducted to develop a theoretical framework that organises the factors contributing to STFs among construction workers into human, task, machine and management factors. Management influences the human, task and machine factors, which directly affect the occurrence of STFs.
Artificial intelligence has the potential to enhance risk assessment of work-related musculoskeletal disorders in office environments. This study externally validated previously developed machine learning models and evaluated the performance of simplified binary data inputs. Six models trained on data from 810 office workers were validated using an independent dataset of 74 globally recruited participants. Model performance across nine anatomical regions was assessed using F1-scores and AUC-ROC metrics. SHapley Additive exPlanations were used to interpret risk factor contributions, and model performance using ordinal Likert-scale inputs was compared with binary-transformed representations. Generalisability varied by anatomical region, with moderate performance for high-prevalence regions (neck and lower back) and reduced accuracy for low-prevalence regions. XGB showed moderate but comparatively stable performance across several anatomical regions. These findings highlight the need for region-specific validation and support the feasibility of simplified, interpretable, and scalable AI-based tools for practical ergonomic risk assessment in office environments.
Despite advancements in automating observation-based ergonomics assessments, most automations still rely on a single underlying construct (e.g. posture). This fragmented approach, however, derogates from the integrated nature of traditional assessment methods. The few studies that consider multiple aspects do so through indirect or manual procedures. This paper studies a novel approach of automating ergonomics assessment, making use of the well-established European Assembly Worksheet (EAWS), incorporating both posture and load manipulation. To this end, it comparatively examines observation-based and automated calculations, using paper-based documents for the former, and inertial measurement units and smart glasses for the latter. The comparative results show a strong, significant correlation between automated and observation-based EAWS calculations (r = 0.74; p < 0.05 for an assembly task, r = 0.82; p < 0.01 for a palletising task). Additional subjective survey results deviate from the EAWS calculations, but are complementary in identifying effects not detected by it.
Firefighters experience high cardiovascular strain, and inadequate cardiorespiratory fitness (CRF) elevates health and performance risks. This scoping review evaluates field-based tests for estimating CRF via maximal oxygen uptake (VO2max) in fire services. PubMed and SPORTDiscus (1990-2025) were searched; of 140 records identified, 102 were screened and 14 included. Tests required documented criterion validity, a standardised protocol, and evidence of operational adoption. Protocols were appraised against five criteria: specificity, validity, reliability, practical implementation, and safety. The 20 m shuttle run offered the most favourable profile for department-wide monitoring, combining high validity, high test-retest reliability, minimal equipment, indoor feasibility, and scalable group testing. Treadmill and stairmill tests provide robust alternatives for individual assessment but are constrained by equipment and staffing. Submaximal tests show larger error. Standardising a field test with regular reassessment may support training control and early detection of decline.
Given the increasing use of touch-screens, the impact of images of objects upon cursor positioning were examined. As objects may present compatible and conflicting cues for movement, we examined the influence of relevant and irrelevant cues on responses when pointing to images on touch-screens. An experiment (N = 24) examined leftwards or rightwards pointing movements to images of large or small cups with handles that could be compatible or incompatible for grasping and were depicted as full or empty. Handle compatibility only benefited response latencies for empty cups, and this effect was reversed for small cups. Kinematic analyses suggested movements were less efficient to images of cups with compatible handles. Aiming points varied with cup size and contents suggesting that handles influenced performance when they presented lower Indices of Difficulty (as per Fitts' law). Object characteristics can influence screen hot-spots, and this primarily occurs whilst the cursor is in motion.
The industry is shifting from rigid systems towards dynamically adjustable architectures with flexible degrees of automation (DOA). A decision-role structure (DRS) defines how decision authority and responsibility for DOA adjustments are allocated between humans and machines. This study examined four DRSs, including human-only (HO), machine-only (MO), serial decision (SD) and parallel decision (PD) in simulated submarine track management tasks. The HO condition exhibited fewer DOA adjustments and prolonged high DOA, yielding better routine performance but poorer situation awareness (SA) and takeover performance. Participants tended to increase DOA under both SD and PD; downward requests were often rejected under SD and delegated to the machine under PD. The PD condition showed higher proportion of effective adjustments, higher SA and better takeover performance without compromising routine performance. The MO condition involved more manual operations and yielded suboptimal outcomes. Overall, the PD condition best balanced routine and takeover performance through proactive adjustments.
PRACTITIONER SUMMARY:Noting differences in active and passive lumbar contributions to trunk extension moment across harvesting postures and ground slopes, this study examined whether a back-support exosuit reduces tissue loads and suggests the exosuit can effectively reduce lumbar muscle activity during kneeling/squatting and passive moment during stooping, with consistent effects across slopes.
Security surveillance operators face a volume of visual information incompatible with the limitations of cognitive processing. Traditional multiplex interfaces present multiple camera feeds simultaneously, assuming that broader visual access enhances threat detection. Yet cognitive research shows that visual processing is serial: only a limited subset of information can be processed in depth at any given moment. Using a surveillance microworld, we compared a conventional multiplex interface against a monoplex interface that supports serial processing by displaying a single camera feed at a time. A total of 128 participants completed four 8-min scenarios comprising incidents that varied in saliency, duration, severity and specificity. Although incident detection did not differ between interfaces, the monoplex condition significantly improved situation awareness with no increase in perceived workload. These findings challenge the simultaneous visual access assumption underlying prevailing surveillance design and support a cognitively grounded approach that supports cognitive fit rather than information availability.
Collaborative robots are increasingly deployed in Industry 5.0 disassembly cells, yet how operators adapt across repeated cobot interactions is rarely characterised multimodally. The Day-2 axis of the MultiPhysio-HRC corpus was analysed: forty-two participants performed up to five repetitions each of a Fanuc CRX-20 cobot-assisted disassembly and a matched manual control, with 12-channel dry EEG, ECG, EDA, EMG and respiration recorded throughout; primary inference was based on the paired-analysis cohort (n = 39, ≥ 3 repetitions per context). Mixed-effects models and per-subject physiological regressions revealed a within-subject adaptation signature confined to cobot: state anxiety, workload, frustration and arousal declined while dominance grew, whereas manual produced flat or worsening trajectories. Repeated-measures analyses showed stronger aggregate physiological-state coupling during cobot than manual work (subject-level permutation p ≈ .001), although no individual channel showed a reliable context-specific effect. Cross-subject prediction of individual adaptation slopes was largely unsuccessful.
During emergency evacuation, visual perception of the sign is critical for route selection, yet how personality traits modulate this process remains unclear. Using a virtual reality subway fire evacuation task with eye tracking, this study examined the interactive effects of sign position and sign brightness on visual search behaviour across the Big Five personality traits (OCEAN). 62 participants completed two evacuation trials in a 2 × 2 design. Personality traits were treated as continuous variables, and trial-level data were analysed using linear and generalised linear mixed-effects models. Low-position signs consistently attracted higher AOI Visits and longer AOI Fixation Time. Personality traits significantly moderated visual search strategies and attentional allocation to the sign, indicating substantial individual differences in emergency wayfinding.
Mental fatigue is an important human factors issue affecting safety performance in high-risk operations. Previous studies have relied on short laboratory tasks and often examined the effects of time-on-task and cognitive load separately. This study investigated their joint effects on mental fatigue in a real drilling environment using EEG cross-frequency coupling (CFC) and a Transformer model. Cognitive load and time-on-task jointly influenced mental fatigue, as reflected in changes in local efficiency and global network reconfiguration. The Transformer framework achieved accuracies of 90.2% and 89.3%, while SHAP analysis identified efficiency entropy as the most important feature. Overall, CFC features effectively characterised mental fatigue in real operations and may support fatigue monitoring, human factors risk identification, and safety management in high-risk settings.
This study investigated drivers' responses to three types of driving hazards (anticipation, complexity, and surprise) during a simulated drive, considering variations in hazard perception, cognitive load, cue utilisation, and visual attention. Drivers recorded the shortest brake response distances for anticipation (i.e. cued) hazards, intermediate distances for complexity (i.e. multi-cue) hazards, and the longest for surprise (i.e. uncued) hazards. Although cognitive load did not differ across hazard types nor levels of cue utilisation, attention metrics such as fixation duration and time to first fixation varied significantly depending on hazard predictability and complexity. Cue utilisation did not influence cognitive load nor visual attention but was associated with earlier braking responses for complexity hazards, suggesting its relevance in situations requiring the integration of multiple cues. These findings may inform driver training programs and advanced driver assistance systems by supporting interventions that align with drivers' perceptual-cognitive abilities and the demands of different driving environments.
Mental and behavioural health (MBH)-related visits account for a substantial portion of emergency department (ED) care. EDs have increasingly adopted technological interventions to support MBH patients and staff. This review examined ED-based technological interventions and their impact on patient safety and quality of care for MBH populations. An initial search on six databases led to identification of 37,432 articles and final selection of 34 peer-reviewed articles on technology interventions. Key findings indicate that screening interventions effectively identify underlying drug and alcohol use, depression, and suicidal ideation in patients presenting to the ED with non-psychiatric complaints; remote psychiatry consultations improve access to mental health care; crisis management intervention programs are well-accepted by patients while waiting in the ED; and documentation interventions enhance communication with providers, and LLMs show strong potential in identifying psychiatric indicators. This review delineates key implications, implementation challenges, and future research directions for ED MBH care delivery.
Engaging experiences often occur when we are deeply absorbed in a challenging activity matching our skills-a state known as flow. Widely sought after in education, sports, and gaming, flow combines deep concentration with intrinsic enjoyment. Yet, its relationship with affective states remains debated. This study examined the relationship between flow and affect in video game players. Participants played a Tetris-like game under three difficulty conditions: too easy, optimal, and too difficult. Flow and affect were assessed through questionnaires, facial expressions, and physiological data (pupillometry, electrodermal activity, cardiac measures). The optimal condition was associated with the highest reported flow levels, increased positive affect, and moderate arousal. The too difficult condition triggered high arousal and negative affect, while the too easy condition showed low arousal. These results suggest that flow-related experiences may be associated with positive affective states, with implications for adaptive game design and player engagement monitoring.
This study examined the effects of inhibitory load and task engagement on workload using self-report, behavioural, and physiological measures. Thirty-five adults completed Stroop task versions designed to manipulate inhibitory load-by varying Stroop congruency-and task engagement-through adaptive and non-adaptive pacing. A greater inhibitory load (compared to a lower load) resulted in a greater perceived workload, along with impaired performance, increased visual engagement (reduce blinking), and reduced boredom. Greater task engagement (compared to lower engagement) resulted in even greater perceived workload, along with more efficient information processing, increased visual engagement, and a slight reduction in parasympathetic tone from baseline (reduced heart rate variability). It also led to increased and anticipated mental fatigue while mitigating the increases in sleepiness observed under lower engagement. These findings demonstrate that perceived workload is sensitive to both inhibitory load and task engagement. Combining high inhibitory load and adaptive pacing produced the greatest perceived workload, offering a promising paradigm for advancing workload research.
Accurately predicting operators' situation awareness (SA) is important in many work contexts. However, current well-validated SA measurement methods require task interruption, motivating alternative measurement approaches. We developed a multimodal ensemble model, using six different non-invasive physiological sensors, to predict high/low SA based on Endsley's three levels. We used a dataset of 31 participants performing the MATB-II, with periodic freeze-probe assessments about the task state to objectively measure SA. We fit logistic regression models for each sensor, combined them using a weighted average, and evaluated them on unseen data. This approach significantly outperforms various baselines, including shuffled-labels, random guessing, and constant-class predictions for all but level 3 SA. Level 3 was more challenging to predict as models did not significantly outperform the constant-class baseline. Sensor importance analysis identified electroencephalogram as consistently most important, followed by eye-tracking. These findings demonstrate effective high/low SA predictions with linear classifiers using physiological data.
This study aimed to develop and evaluate a tactile manual input driver-vehicle interface to support simple non-driving-related tasks and emergency braking in conditionally automated vehicles. A within-participant simulator experiment with 30 drivers compared the proposed interface with a conventional interface for simple non-driving-related task usability, and with the foot brake pedal for emergency braking. User acceptance was explored by comparing participants who experienced the proposed interface with selected respondents from a previous questionnaire study. The results showed that participants completed the item-switching task significantly faster and exhibited more efficient interaction behaviours using the proposed interface. Questionnaire responses indicated higher scores on several acceptance variables than selected respondents. In the emergency braking task, reaction times were significantly shorter using the proposed interface than the foot brake pedal. These findings provide practical implications for designing tactile manual input interfaces that support simple functions and rapid driver intervention in conditionally automated vehicles.
