Exoskeletons can reduce physical strain at work, but they can also change how wearers are perceived by others. This may be especially consequential in professional caregiving, where warmth and human connection are central to good care and where gendered expectations frame the domain as feminine. Yet little is known about how exoskeleton use is socially received in care settings. This online experiment examined how wearing an exoskeleton shapes perceptions of caregivers, drawing on the Stereotype Content Model and theories of mechanistic dehumanization. We recruited 223 participants from the general population, who viewed one of four videos showing a female or male caregiver, with or without an exoskeleton, in an eldercare interaction and rated the caregiver’s warmth, competence, and machine-likeness. Exoskeleton use alone showed no overall effect. However, responses varied by caregiver gender: for female caregivers, wearing an exoskeleton came at a social cost, as they were perceived as less warm and more machine-like. For male caregivers, exoskeleton use did not alter these perceptions. Our findings suggest that the social reception of exoskeletons in care may be influenced by gender role stereotypes, refining existing accounts of dehumanization in human–technology interaction.
User eXperience (UX) is a multidimensional concept encompassing emotions, cognition, and product characteristics. From a design standpoint, addressing this multidimensionality necessitates the collaboration and communication of multiple design specialists, such as industrial designers, ergonomists, and engineers. To effectively support this collaborative process, there is a need for a tool that can integrate all UX dimensions throughout the design process. This paper identifies and maps 75 existing tools—both UX-specific and non-UX-specific—across different UX dimensions and phases of the design process (specification, imagination and evaluation). The resulting mapping highlights the absence of a single tool capable of representing UX multidimensionality in a comprehensive, consistent, and unambiguous manner throughout the design process. In response, the paper introduces the “UX grammar” inspired by three main existing approaches: UX Journey Maps, DRED, and SysML. This tool employs natural language statements interconnected through a graphical structure, enabling the visual representation, specification, and measurement of User Experience. An experiment is conducted as a preliminary evaluation focusing on the tool’s usability. The discussion outlines the tool’s strengths and limitations, as well as avenues for future work including a utility-focused evaluation to further assess the benefits of the proposed tool. Finally, several perspectives are envisioned for this grammar in design practice, namely supporting designers in their work and facilitating the design of workers’ experiences with the resulting artifacts. A mapping of 75 UX and non-UX specialized tools for UX Design is drawn up. The absence of a tool to represent all the UX dimensions across design phases is highlighted. A new tool is built around a graph and statements written in natural language. This “grammar” enables UX to be specified, imagined and evaluated in a complete, unambiguous, and uniform way. An experiment is carried out as a preliminary evaluation of the grammar usability. Avenues for future work include a utility-focused evaluation to assess the tool’s benefits in practice.
We explored the predictability of individual factors related to job performance (e.g., cognitive ability and Big Five personality traits) based on applicant’s résumé formatting that was inferred by a computer. Humans typically emphasize the content of the résumé, such as education, work experience, and skills listed, when subjectively evaluating the applicant. Because job applicants may falsify or exaggerate their résumé information to match the prospective employer’s desires, using the content to draw inferences about the applicant has limitations. To address this limitation, we trained an artificial neural network to learn the common visual features within a résumé. We used these visual features as predictors in a machine learning model to predict an applicant’s work-related traits on a sample of 387 résumés. We found that the visual features of a résumé could predict both people’s personality traits and their cognitive ability beyond chance levels. We compared those predictions with the ratings made by two human raters using a linear regression model that predicted the true trait scores with the human rater scores and the neural network model’s predictions. Results show that neural networks significantly improved the explanatory ability of the model compared to only using human ratings. The relationship between résumé style and people’s traits could not be explained by simple measures of visual appearance such as the word count or darkness of content. Using this scalable method, practitioners can extract additional information from their existing selection process to better identify qualified job candidates within large pools of applicants.
This study examines farmers’ views of data privacy, data ownership, cybersecurity, and governance in Canadian agriculture. Evidence on data protection and security in this context remains limited. To address this gap, we interviewed 45 stakeholders between 2020 and 2025, primarily in Ontario and Nova Scotia. Participants included 31 farmers (three of whom combine farming with another occupation), 13 other agricultural stakeholders, and one government representative. Following Lochmiller’s (2021) three phases of thematic analysis, we coded participants’ reasons inductively into nine first-order categories. These categories were consolidated into six themes and overarching three classifiers: Socio-Technical Risk and Assurance, Technological and Systemic Vulnerabilities, and Economic and Strategic Impact. Across all three classifiers, trust emerged as the central mechanism connecting cybersecurity, governance, privacy, AI, and technology adoption. Farmers generally trusted that the technologies worked while distrusting the organisations behind them. They treated AI as a tool to be verified rather than an authority. Farmers did not reject innovation outright. Resistance arose mainly where technologies seemed opaque, poorly governed, externally controlled, insufficiently regulated, or economically burdensome. The findings reframe cybersecurity adoption in agriculture as a socio-technical process rather than a purely technical issue. This process is shaped by governance, legitimacy, transparency, awareness, and economic realities. They also highlight farmers’ potential to act as agents of change. Informal mechanisms, such as peer learning and community-driven adaptations, should complement formal policies. Transparent, participatory governance is needed to align society with technological progress.
In the Industry 5.0 landscape, placing humans at the center of production requires robust models to manage Human Error Probability (HEP) amid increasing cognitive complexity. Despite the availability of separate evaluation methods, there is a lack of predictive models that explicitly link HEP to the simultaneous interplay of Task Complexity, Operator Characterization, and Work Environment. This research bridges the gap by proposing a novel multimodal analytical framework that uses Multi-Attribute Utility Theory (MAUT) to synthesize structural task complexity (TACOM), operators' physical/cognitive traits, and environmental information. The framework was evaluated through a controlled laboratory study involving 22 operators. Multiple regression analysis explored the distinct statistical contributions of the macro-domains (p < 0.05), providing a preliminary structural evaluation of the model's architecture. Furthermore, a comparative ROC analysis against traditional baselines yielded low discriminative metrics (AUC 0.48 vs. 0.39), indicating that while the findings support conceptual plausibility and preliminary internal structural assessment, static weight configurations are insufficient for robust field predictions. Notably, the framework modeled a pattern consistent with non-linear cognitive-load effects, offering a plausible interpretation, requiring direct future validation, of increased error probability among expert operators under low-stimulation (underload) conditions, alongside vulnerabilities among novice operators due to skill gaps. Quantitative analysis yielded Mean Absolute Error (MAE) values of 0.1897 and 0.1646. At the same time, this overestimation provides a functional safety margin by signaling latent risks; it also highlights the future need for calibrated alert thresholds to prevent alarm fatigue in real-world applications. This study offers a reproducible, context-aware assessment framework to support proactive human risk management in cognitively demanding industrial tasks.
Building on Hoff and Bashir’s (2015) model of trust in automation, this study investigates how representational framing influences reliance and trust in a non-normative dual-advisor paradigm. In a Close Air Support–inspired scenario, 80 military trainee pilots repeatedly arbitrated between advice provided by a human advisor, presented as a Joint Terminal Attack Controller (JTAC), and an automated decision aid. Automation type varied between subjects: the automated advisor was framed either as a traditional mapping system or as an AI-based adaptive aid. Declared reliability varied within subjects from 20 to 90
This study explores a dissonance-based risk analysis approach grounded in the Benefit/Cost/Deficit (BCD) cognitive model and its application to adolescent cycling behaviors amid traffic conflicts. The application of the BCD model for cognitive restructuring and risk assessment constitutes a novel educational intervention. This intervention aims to facilitate adolescents’ learning via a human-error-based framework, thereby improving their cognitive abilities and cycling performance. A controlled closed-course field experiment was conducted, incorporating six customized traffic conflict scenarios that involved rule violations by interacting traffic participants. Twenty adolescent cyclists were recruited and randomly allocated to either a control group or an experimental post-experimental group. To evaluate the intervention effects of BCD-based dissonance risk analysis on perceived risk and cycling decision-making, assessments were administered to participants immediately after the intervention and in a one-month follow-up test. During the experiments, objective metrics (including cycling speed, head-turning frequency, and eye-tracking gaze behaviors) and subjective hazard perception indicators were measured. Data analysis was performed using Linear Mixed-Effects Models (LMMs). The results showed statistically significant differences between the experimental and control groups, as well as between the follow-up and baseline control conditions. Compared with the control group, adolescents in the experimental group exhibited a 2
High-speed train driving is a safety-critical human–machine task in which executive functions support continuous monitoring, information integration, and safe decision-making. This study examined how motivational contexts influence performance and physiological responses in executive-function-related tasks in the context of high-speed train driving. The N-back, task-switching, and Stroop tasks were used as representative paradigms with primary demands on updating, shifting, and inhibition, respectively. A 3 (executive-function-related task: N-back, task-switching, Stroop) × 3 (motivational context: baseline, reward, hazard-warning) × 3 (task-internal load level) within-subject design was adopted. Thirty-two trainees who had received high-speed train driving training completed the experiment. Accuracy, reaction time, mean pupil diameter, standardised skin conductance response, and NASA-TLX scores were collected. Task type and task-internal load level showed relatively stable effects on behavioural performance, physiological responses, and subjective workload, with the N-back task showing higher cognitive costs. Motivational context affected some behavioural and physiological measures, but its overall effect was limited. The reward context did not consistently improve response performance, whereas the hazard-warning context may have induced greater risk-focused attention and a more cautious response strategy. Mean pupil diameter showed greater sensitivity than standardised SCR and NASA-TLX scores to task-related and motivational changes, although it should be interpreted as a non-specific indicator of cognitive-effort-related and arousal-related resource mobilisation. These findings suggest that motivational contexts may primarily influence resource investment or compensatory processing rather than directly producing stable behavioural benefits. The study provides preliminary implications for optimising information presentation, risk prompts, and state-monitoring support in high-speed train driving interfaces.
Advances in artificial intelligence and driving automation are fostering a paradigm shift in which technical systems exhibit cognitive and operational capabilities that surpass human performance. These developments have prompted proposals to grant automated systems greater control and authority, including allowing the automation to act authoritatively, fully blocking or taking away control from human users. While such delegation may promise improvements in safety and efficiency, it also carries the risk of disempowering humans if not carefully managed. This tension underscores a central challenge for human–machine interaction (HMI) design: maintaining operator agency and trust when interacting with automated systems that demonstrate ability beyond that of humans. The concept of Control Management, originating from aerospace, offers a pathway for balancing authority between human and automation by defining system policies that guide its architecture, mechanisms, and HMI design. In this conceptual work, we examine the theoretical foundations of human-automation collaboration and describe emerging challenges for driver interaction with authoritative automation. This work contributes: (1) a conceptual classification of authoritative control interventions across automation levels and control authority transition types, (2) Control Management-informed HMI guidelines spanning Avoid, Alert, Inform, Recover, and Return intervention states, (3) ethical implications for preserving driver agency, and (4) research directions for system development, policymakers, and HMI designers.
Startle and surprise are known to potentially incapacitate professionals who respond to emergency situations. In the aviation domain, self-management methods to prevent such incapacitation have been introduced for pilots. This study aims to explore the relevance of similar self-management methods tailored for cabin crew, using subjective evaluation by cabin crew and cabin crew instructors. First, a method was designed and refined using interviews and a focus group with nine subject-matter experts, consisting of four steps: protect, reset, check and act. Second, relevance of the method was quasi-experimentally tested by letting 15 cabin crew members apply it in a simulated firefighting scenario. After performing the scenario, they rated whether the method had positive or negative effects, was difficult or easy to use, and they retrospectively rated their perceived stress before and after applying the method. Participants positively rated the method’s effectiveness and usability, and reported the method to significantly reduce perceived stress. Participants provided several suggestions for improving the method, and for approaches to integrate the method more effectively into existing protocols. The outcomes of this study provide direction on the design and use of self-management methods that can help manage startle and surprise in teams.
Modern train cabs have multiple technological systems aiming to minimise the risk of an incident. This qualitative study of train driver instructors investigated: what in-cab systems and warnings are perceived to be the most helpful for train drivers? And, what themes best describe the relationship between train driving and in-cab technology?. Five online focus groups with train driver instructors (n = 19) involved participants undertaking a two-part prioritisation exercise considering the current technologies and safety features in the train cab and ranking the three they considered to be most helpful and most unhelpful. Open discussion about the importance of the technologies in supporting safety, the driver instructors’ opinions about using the technologies, and any safety gaps followed. The safety systems considered most helpful were the AWS (Automatic Warning System), Vigilance, DRA (Driver Reminder Appliance), and TPWS (Train Protection Warning System). Three key themes were identified. “The patchwork cab”, highlights the participants’ experience of using the numerous and sometimes overlapping in function in-cab technologies; “eyes, ears, and staying aware in the cab”, highlights challenges around distraction; and “balancing alerts and silence” highlights workload dynamics and task engagement. Additions of new safety features while retaining legacy systems has resulted in a patchwork like cab design that may not effectively support the driver in their work and decision making. The management of acknowledged distractions has become normalised and is accepted rather than designed out of the cab. In-cab technology has an important role to play in safe vehicle operations as humans and machines become essential team-mates. However, saturation point has been reached for this patchwork approach to in-cab technology, and it will be a challenge to the industry to change this method and redesign systems rather than adding to them.
Supporting ecodriving in battery-electric vehicles (BEVs) requires feedback aligned with drivers’ mental representations, as effective regulation depends on know-how and know-why. When mental representations are inadequate or confidence exceeds actual knowledge, this can undermine performance and feedback processing. This study primarily examined drivers’ mental representations of ecodriving via thematic analysis, focusing on knowledge gaps (missing beliefs, situational references, reasoning depth) and references to input–comparator–output information. As a complementary and exploratory component, knowledge accuracy, uncertainty due to a lack of knowledge, driving behaviour, and performance across different feedback approaches are additionally analysed to contextualise and further specify the qualitative insights. In a driving simulator study, participants (N = 63) drove under one of three conditions: no feedback (G1), real-time consumption trace (G2), or optimal speed recommendation (G3). Afterwards, they provided ecodriving tips and technical explanations, offering insights into their understanding. Qualitative analysis showed broad familiarity with general ecodriving principles (e.g., smooth driving) but little precise or technically grounded guidance. Misconceptions were common across all groups, especially on regenerative braking, acceleration, and pedal use. Exploratory quantitative comparisons suggested that G3 reported lower uncertainty and drove slower in constant-speed phases than G1. G2 used mechanical braking more and regenerative braking less than G3. We did not observe statistically reliable group differences in knowledge accuracy or mean energy consumption. Findings indicate that cognitively aligned feedback must go beyond prescribing speed selection or energy raw data. To foster robust ecodriving and reduce uncertainty, systems should support causal understanding and accurate, transferable mental representations.
Adaptive automation is increasingly explored in air traffic control as a means to manage variability in operational demand while preserving controller engagement, situation awareness, and authority. This paper reports the COntroller adaptive Digital Assistant (CODA) contribution on making adaptive task delegation governable in supervisory work. We specify an adaptive automation strategy for non-critical air traffic control support tasks that combines explicit mode logic, authority contracts, and interface mechanisms that render delegation inspectable at the point of action. Adaptation is formalised as a closed-loop policy driven by three trigger classes—traffic demand, task/load demand, and an operator-readiness dimension—computed over both current and short-horizon predicted states to anticipate demand peaks and to support timely, bounded transitions. The strategy operationalises three graduated proactivity modes with explicit authority management: Manual (monitor-and-inform), Proactive-Light (recommend-and-confirm), and Proactive-Strong (initiate-and-notify with veto), designed to support appropriate reliance and maintain continuous legibility of task ownership. We detail delegation rules that constrain automation to supportive functions, and we present a human–machine interface that externalises automation state, proposed actions, and ownership while incorporating concise, actionable rationale cues intended to support controllability with minimal overhead. Finally, structured workshops with expert controllers are synthesised to derive operational acceptability constraints on delegation boundaries, transition timing, preview horizons, and explanation selectivity, grounding the design in the realities of air traffic control work.
The concept of Simplified Vehicle Operations (SVO) is proposed to reduce input complexity and training time for future Advanced Air Mobility (AAM) pilots. So far, flight controllers and input concepts for SVO are often evaluated using computational simulations or in small-scale experiments with professional test pilots. This paper presents findings from a quantitative user study on the flight performance of n = 31 novices and n = 7 system designers in the execution of standardized Mission Task Element (MTEs) in a flight simulator. These MTEs include complex transition phases with a switch from rotary-wing to fixed-wing flight and vice versa. Absolute values from multiple subjective, performance, and input-based parameters indicate the general adequacy of the tested SVO concept, even for beginners. At the same time, multiple parameters also show significant differences between the two groups. The paper discusses these findings and SVO concepts in general to improve the future design of SVO concepts for minimally trained pilots.
The existing literature on Mental Workload (MWL) has used subjective and behavioral measures to understand processes involved in MWL. More recently, physiological measurements have also shown promising perspectives in this regard. However, their heterogeneity and interindividual variability leads to question their reliability, dependencies and covariations for daily use. The present work aims to identify objective psychophysiological markers that reliably reflect standardized MWL variations and their relationship with subjective and behavioral responses. We defined pre-standardized taskload levels to assess the stability of physiological markers, across MWL variations. Data was collected under a computer-based controlled task (MATB-II) mimicking aircraft pilots’ activity during flight. Standardization procedures were applied to the MATB-II, allowing the identification of scenarios inducing low, medium and high MWL, keeping constant the number and type of subtasks involved. Results showed that participants’ subjective workload and performances followed the taskload manipulation. Then, we identified at least three types of physiological measures (cardiac, ocular and electroencephalographic) affected by taskload, notably discriminating high workload conditions. Correlation analyses showed that cardiac and EEG measures are consistently associated with subjective MWL estimations. Inconsistencies were also observed when randomizing taskload levels for several physiological measures (heart rate, heart rate variability, engagement ratio). These results extend our knowledge of MWL to varying levels of taskload, while questioning the reliability of psychophysiological markers for the detection of evolving MWL. Ultimately, this knowledge could be applied in industries, supporting product development, to assess new designs with regards to their impact on users’ workload and performances, throughout the design cycle.
Psychological stress among nuclear power plant (NPP) operators represents a critical human factors challenge, as it can impair cognitive performance and decision-making, especially during high-risk scenarios. While prior research has identified various stressors, the interdependencies and systemic influences among these factors remain underexplored. This study adopts an integrated DEMATEL–ANP approach to quantify the causal relationships and relative importance of operator stressors. A previously developed grounded-theory indicator system—comprising seven primary dimensions and 22 secondary indicators—serves as the conceptual input. DEMATEL is used to map the direct and indirect influences among stressors, and ANP is employed to derive global weights within the interactive stress network. A case study involving 16 expert evaluations demonstrates that alarm load, task complexity, and interface management load are the most influential primary stress dimensions, while emergency response and novel event handling emerge as key secondary stressors. The integrated DEMATEL–ANP model not only identifies dominant stress drivers but also reveals how stressors propagate through feedback and dependence relationships. The findings offer a systemic perspective on operator stress formation and provide a decision-support basis for prioritizing stress-mitigation strategies in NPP operations.
Navigating crossings safely requires quick and accurate stopping decisions. Accidents at railway level crossings may result from internal factors such as pedestrian distraction or external factors such as the absence of effective safety devices. Although a wide range of safety systems is deployed worldwide at railway level crossings, little empirical work has examined how they support human perceptual, cognitive, and motor processes underlying action inhibition. In this study, stopping behavior was investigated using a stop-signal task with six different signal conditions designed from two theoretical perspectives: perception-action (PA) coupling, which emphasizes rapid motor responses to direct cues, and situation awareness (SA), which supports detecting, interpreting, and anticipating critical information. Participants performed an online task where they controlled a moving avatar and had to stop upon different signals, either in a simple task or while solving arithmetic problems as a distraction. Results showed that PA signals triggered the fastest reaction times but with higher error risk, while SA signals promoted safer but slower responses. More importantly, signals combining perceptual affordances with selected situational cues achieved the most effective balance, reducing collision risk without sacrificing speed. These findings contribute to a better understanding of the complementary roles of perception-action coupling and situation awareness in stopping behavior and provide practical, human-centered guidance for the design of safety systems at railway level crossings and other time-critical pedestrian environments.
The maritime industry is undergoing a transformative shift with the integration of digital technologies altering maintenance management practices. This qualitative study explores the following question: What opportunities and risks do shipping companies associate with maintenance management when adopting digital technology? The empirical data material stems from a workshop involving stakeholders from 14 Norwegian shipping companies. The findings highlight significant opportunities for more effective maintenance strategies, enhanced data integration, and improved real-time collaboration between ship and shore. The transition poses several challenges related to system interoperability, data ownership, competence development, and informal social structures. The increasing reliance on digital tools raise concerns about power structures, cultural resistance, insufficient digital competencies, and the erosion of crew autonomy and expertise. These findings emphasize the need for a systemic approach that integrates technical, organizational, and social dimensions. To mitigate future risks, strategic planning, robust change management, and fostering a culture of adaptability and trust are essential. This study underscores the importance of aligning technological innovations with organizational goals to ensure sustainable and effective maintenance management in the future.
The integration of collaborative robots (cobots) is transforming manufacturing, yet empirical evidence on their impact on employees’ work experiences and well-being remains limited and often yields contradictory findings. This exploratory study used a within-subject quasi-experimental design to compare manual and cobot-assisted modalities of a repetitive battery disassembly task and to examine how cobots alter motivational work characteristics, cognitive workload, and psychological stress. Thirty-eight participants performed the task under both modalities, while self-report questionnaires and physiological signals were collected. Results showed that manual tasks provided higher autonomy and job complexity, whereas cobot-assisted tasks offered greater task variety, information processing, problem-solving, and skill variety. Physiological data revealed reduced stress and muscle activation during cobot-assisted work, partially supported by self-reports, which indicated lower physical but higher mental demands. This study offers contextually grounded insights into how introducing a cobot into a specific manufacturing task can simultaneously reduce physical strain while increasing cognitive demands and psychological stress. These findings highlight the importance of human-centered implementation strategies that support employee adaptation during transitions to collaborative robotics.
The aim of this study was to examine drivers’ hazard perception in an urban scene, under two different lighting conditions (daytime and after dark) and for two different levels of driving automation (manual driving and hands-off SAE level 2 automation). Forty-eight participants took part in four experimental drives in a driving simulator, encountering six different hazardous/potentially hazardous events in each drive. Results showed that drivers detected hazards significantly earlier and were more likely to react to the hazards in daytime, compared to after dark environments, particularly when pedestrians were approaching the road from the left. However, there was no significant difference in their response time towards the hazards, between the daytime and after dark environments. In terms of driver response in the two automation levels, the majority of drivers were proactive and reacted before the potentially hazardous events turned into actual hazards during manual driving, but responses were more reactive during automated driving. These findings highlight the need to account for context of the driving environment such as lighting conditions and levels of driving automation when designing systems or protocols that aim to support hazard perception and timely driver response.
