This study examines how healthcare workers in Mohalla Clinics (MCs) perceived environmental design needs during the COVID-19 pandemic. Specifically, it addressed three questions from the perspective of healthcare workers: What were key environmental design features of MCs that impacted delivery of primary healthcare services? How did these features influence human-environment interactions in MCs during the COVID-19 pandemic? How can insights from these interactions inform design the of MCs resilient to future pandemics? Using a participatory design approach, 35 healthcare workers from 9 MCs were interviewed and they sketched out their daily routines on a plan view of their MC. Analysis of data revealed need for—reconfigurable layouts; unidirectional patient workflow; use of building fenestrations as portals for healthcare services. Therefore, the study proposes—flexible and reconfigurable interior layouts, unidirectional patient workflow, and modular fenestrations—to enhance resilience of MCs for future public health emergencies.
Biometric signals are poised to enhance the research landscape of Human Factors (HF) and related fields. However, many existing research-grade biometric devices run on expensive, proprietary systems, while consumer-grade devices are closed source, offering limited or no access to the data. To address this gap we developed EmotiBit, a scientifically-validated, wearable, open-source, biometric sensor module capable of collecting a range of emotional, physiological, and movement data. Affordable, open-source, non-invasive hardware can lower the barrier to entry for academic labs and industry practitioners to integrate research-grade physiological sensing into their Research & Development (R&D) workflows. After 5 years on the market, we have begun to see trends in the way that EmotiBit is being utilized in both research and industry settings. In this paper, we survey the landscape of scientific applications using EmotiBit and report on the impact of open-source biometric sensing on HF and related fields.
Construction workplaces are hazard-rich environments where workers must maintain sustained attention to prevent human errors. In parallel, construction jobsites are inherently social environments that require collaborative work and peer interactions. Such interactions and peer presence may influence workers’ attention, potentially causing distraction. While previous studies have shown that distraction can impair attentional skills, there remains a paucity of research empirically examining whether peer presence contributes to distraction in construction settings. To address this gap, this study investigated the effects of peer presence on workers’ perceived distraction and attentional allocation within a virtual reality environment. Participants completed the simulated task under two conditions: with peer presence and without peer presence. The results showed that participants allocated significantly less attention to the task at hand and hazardous areas in the presence of the peer. These findings provide important insights into how peer presence may divert workers’ attentional resources and contribute to error.
Efficient walking depends on the interaction between both gait mechanisms and energy demand; however, this relationship may differ across age and functional groups. This study examines this association in younger and older adults. Sixty-four participants were divided into two groups: younger adults aged ≤40 years and older adults with reduced physical function, defined as Short Physical Performance Battery scores ≤9, and. Metabolic outcomes included peak, gross reserve metabolic costs, while gait metrics included 10 spatial, temporal, and variability measures. Pearson correlations were calculated separately to examine the differences within each group. Younger adults demonstrated stronger associations between metabolic outcomes and temporal gait measures, whereas older adults showed stronger associations involving gait variability, asymmetry, and support-time measures. These findings suggest that age plays a critical role when there is a shift in gait and metabolic cost coupling, with inefficient locomotion in older adults more closely related to variability and stability control.
Generative AI (GenAI) applications are increasingly adopted in co-creative workflows, taking an active role in design and development tasks. GenAI offers unique opportunities through content generation and synthesis across resources, yet it can lead to losses in co-creativity, due to limited collaborative interactions between humans and GenAI. This loss results in user overreliance on GenAI outputs, skill degradation, and reduced creativity. This study compares two GenAI interaction styles: Standard, providing direct outputs to user prompts; and Collaborative, supporting problem definition, diverging, converging, and organizing. A user training intervention was also assessed. Findings show that collaborative GenAI led to higher perceptions of collaboration, including exploration and expressiveness. In contrast, standard GenAI led to more direct use of the AI ideas, with fewer discussions and evaluations. Training supported more problem-definition and less direct use of AI ideas without discussion. Both AI teammate design and user training contribute to fostering collaboration in HATs.
Despite the prevalence of course registration systems in post-secondary education, there is little research on the usability of those systems. The present study investigated the influence of error message design and task type on error recovery in course registration. Forty-four college students ( M age = 20.48 years; 33 female, 11 male) went through a course registration simulation, which included two error tasks: waitlisted (i.e., full course) and linked (i.e., additional course needed) and two error message design conditions (altered and original message design). Usability was assessed as effectiveness and efficiency. Results showed that although usability did not differ between error message design, the linked task induced higher levels of errors made and time spent compared to the waitlisted task. Our findings contribute to a better understanding of course registration systems and inform future design of these systems.
Upper-limb prostheses increasingly combine user-intent decoding with contextual autonomy, improving robustness but risking reduced sense of agency when assistance overrides user control. We present agency-constrained shared autonomy (ACSA), which treats control authority as an explicit, auditable interaction budget. ACSA computes an Agency Loss Index (ALI) from the user-intent distribution and restricts autonomy to an agency-feasible envelope. SetACSA then selects the autonomy-preferred grasp within that envelope. Using the open MeganePro/NinaPro DB10 dataset, we created an offline replay benchmark comprising 3,814 grasp episodes from five able-bodied participants and four transradial amputees. SetACSA improved accuracy from 0.532 ± 0.021 for user-only control to 0.580 ± 0.014 while maintaining low agency loss (ALI = 0.012 ± 0.001). Confidence-weighted blending achieved comparable accuracy (0.560 ± 0.017) but substantially greater agency loss (ALI = 0.087 ± 0.004). These results support evaluating prosthetic shared autonomy through a performance–agency frontier rather than accuracy alone.
Exoskeletons enhance user abilities but often require users to adapt their task strategies. Training may be critical to this process, yet few studies have evaluated whether training methods improve performance. This exploratory study examined whether additional training on how exoskeletons alter physical capabilities improves performance more than extended practice with exoskeletons alone. Nine participants (5 with structured training and 4 without) completed an exoskeleton-assisted lifting task alongside a computerized flanker task before and after a 1-hour practice session. Task performance, self-efficacy, and self-reported strategies were measured. Analyses showed no significant improvements from practice or training, though notable trends emerged: performance on both tasks worsened with practice, but trained participants showed smaller declines in lifting performance and higher self-efficacy than untrained participants. Both groups reported similar strategies and negative experiences with exoskeleton use. Future research should explore whether different instructional content or feedback methods better support exoskeleton-assisted task performance.
Chronic stress contributes to adverse physical and mental health outcomes. Regular mindfulness practice has been shown to reduce stress and improve well-being. Although digital mindfulness tools are widespread, most applications rely on pre-recorded guided content and offer limited support for monitoring users’ internal processes or evaluating practice effectiveness. To address this gap, this project presents Recalibrate , a user-centered web application that integrates generative AI, a large language model (LLM), and neural text-to-speech (TTS) to support brief, personalized mindfulness practices within everyday digital contexts. Using a research-through-design approach, the system emphasizes reduced cognitive workload through short guided meditations, breath awareness, and optional reflective journaling. A typical session begins with a mood and somatic check-in, which informs the generation of a personalized one-minute meditation script and journal prompt. The meditation is delivered through natural voice output, followed by optional journaling, intention setting, and feedback. By combining adaptive AI-generated guidance with lightweight reflective practices, Recalibrate contributes a transferable set of design considerations for creating supportive, low-demand reflective systems.
Autonomous operations in nuclear power involve complex human factors challenges. This is because the industry has been held to a high standard for safety due to a combination of consequence and public perception of risk and consequence. In industries like manufacturing and transportation, the risks associated with autonomous system failures are typically managed incrementally, allowing for faster adoption and iterative learning based on real-world data. Advanced autonomous concepts in nuclear power often require a redefinition of the operator’s role, which may bring about skill degradation, trust miscalibration, and compromised situation awareness. While these human factors challenges are not new to process control environments with high automation, the industry’s strong regulatory environment, rarity of high-consequence events, and defense-in-depth philosophy mean that addressing these challenges calls for tailored, evidence-based solutions. In this paper, we review characteristics within nuclear power that present human factors challenges that take on added complexity with autonomous operations.
Occupational and environmental medicine (OEM) clinicians operate at the intersection of clinical medicine and the legal system. The impact of the clinical decisions and determinations made by OEM clinicians extends far beyond clinical outcomes. The integration of artificial intelligence clinical decision support systems (AI-CDSS) for OEM-specific clinician tasks requires careful consideration of the domain-specific potential harms that may arise from clinician use of this technology in OEM, as such hazards may not be readily apparent using conventional AI safety and risk assessment frameworks for the general healthcare domain. This conceptual systems-based hazard analysis applies the System-Theoretic Process Analysis (STPA) framework to the use of AI-CDSS in clinical OEM, focusing on domain-specific hazards. This analysis identified legal, regulatory, economic, and public safety hazards that should be addressed to ensure responsible and ethical integration of AI-CDSS in clinical OEM.
Learner engagement is commonly viewed as a key factor in successful learning. In online settings, limited face-to-face interaction can make learners more prone to distraction and reduced attention, highlighting the importance of monitoring and sustaining engagement. Recent advances in generative AI allow systems to infer learners’ cognitive and emotional states from multimodal cues, enabling more personalized and adaptive instructional support. However, little research has examined how such systems can dynamically adapt both the type and timing of feedback based on learners’ moment-to-moment engagement states inferred from multimodal signals. This work presents an adaptive multimodal AI-driven tutoring system that infers learners’ states by interpreting real-time visual, auditory, and behavioral cues. Based on the inferred learner state, the AI tutor determines when and how to intervene to sustain engagement. The system is structured as a closed-loop cognitive architecture: perception (capturing real-time multimodal cues), decision (aggregating the multimodal inputs into four affective metrics), and action (delivering feedback based on the inferred state by mapping each metric to a feedback type and timing strategy). This work presents a high-fidelity, interactive multimodal AI tutoring system that illustrates the feasibility of integrating multimodal cues to enable adaptive instructional feedback and engagement-aware intervention in online learning contexts.
Risk management and risk communication are inherent components of human factors work. Human factors professionals must confront how human capabilities and limitations impact the safe and effective use of devices and medications and their role in attaining an informed balance between the potential for harm and benefit. Notably, effective risk management and communication depend on systemic elements including the audience, the product, and the environment. For instance, communications regarding an in-clinic dialysis system for use by physicians and patients will be different from how risk is communicated for a kitchen appliance designed for home use. This proceedings paper provides a brief introductory overview of risk communications and summarizes tools and strategies for effective risk communication.
An accurate mental model of the temporal characteristics of complex task environments, or temporal awareness (TA), presumably underlies human performance when interacting with dynamic systems. Nevertheless, objective methods for characterizing TA from observable behavior remain elusive. Observable data from human-system interactions may be used to estimate operators’ TA. We applied an interpretable optimal clustering tree model to temporal data collected from participant interactions with a modified NASA Multi-Attribute Task Battery (MATB)-II to determine whether distinct behavioral patterns associated with differing levels of TA could be recovered from objective data. The model, constructed using two temporal measures theorized to reflect covert TA, time-since-last-interaction and the temporal response index, recovered behavioral patterns associated with nontrivial differences in participants’ task performance. These findings highlight the importance of temporal cognition in human performance, provide means for objective performance monitoring in complex, dynamic task environments, and ways to improve operator training and interface development.
We present an information-theoretic framework for identifying which physiological and eye-gaze-based indicators, and at what temporal scales, are most informative of cognition-related reliance behavior during conditionally automated (SAE Level 3) driving. We analyze data collected from an in-person driving simulator study in which participants interact with a conditionally automated vehicle in a single continuous drive. By evaluating physiological signals (heart rate, galvanic skin response) and eye-gaze fixations through mutual information and conditional mutual information, the approach enables assessment of both overall and uniquely contributed informational relevance without assuming linear relationships or predefined model structures, making it suitable for continuous, non-trial-based settings. Results show substantial inter-participant variability in both informative features and optimal time scales, suggesting that no single physiological indicator is universally optimal. The proposed framework highlights the importance of personalized sensing strategies and provides a scalable methodology for future larger-scale studies on cognitive-state inference in automated driving.
Partially automated vehicles control vehicle speed and roadway position but require drivers to monitor roadway and automation events that may go undetected by the automation. Although drivers must remain vigilant, their ability to detect and respond to roadway hazards declines over time. Higher automation reliability can lead to high trust resulting in overreliance and poorer monitoring. The study assessed drivers’ ability to remain vigilant when detecting hazardous automation events and how different automation reliability affects trust and vigilance. Participants completed a 40-minute simulated driving task under one reliability condition. Participants responded slower over time, suggesting a decline in vigilance toward hazardous automation behaviors. Moderate reliability supported better performance and user state. These findings suggest that while designers may aim to maximize reliability, consideration must be given to how reliability affects user performance and state.
Automated diagnostic aids can improve decision making, yet operators often fail to use them effectively. This study examined whether personality traits and decision-making styles predict aid-use efficiency in a signal detection task. One hundred seventy-five undergraduate participants completed a signal detection task with and without a 93% reliable automated decision aid and completed measures of conscientiousness, neuroticism, satisficing, and decision difficulty. After preregistered exclusions, Bayesian analyses examined whether individual differences predicted aid-use efficiency, quantified as the ratio between observed and optimal aided performance. The aid significantly improved sensitivity, providing evidence of its diagnostic value. However, Bayesian correlations provided evidence against relationships between aid-use efficiency and conscientiousness, neuroticism, satisficing, or decision difficulty. Additional analyses showed that neuroticism was associated with greater decision difficulty, whereas conscientiousness was associated with lower decision difficulty. These findings suggest that individual differences in personality and decision style may not explain inefficiencies in automation use.
Transformer-based large language models are becoming increasingly popular in supporting qualitative analysis, but their applications and validation are often poorly characterized. This scoping review followed PRISMA-ScR rules to understand how such models are used in the analysis of interview and conversational data. Searches of ProQuest, Web of Science and IEEE Xplore (2020–2025) returned 3,102 records; 207 were assessed at full text, and 84 peer-reviewed studies were included. Support for thematic analysis was the most common application (43/84, 51%), typically with full transcripts analyzed using zero-shot prompting. Decoder-only models dominated (69/84, 82%), with GPT-4 (51%) and GPT-3.5 (21%) most frequent and Llama the leading open-source alternative (10%). Validation, when present, relied mainly on human-LLM comparison ( n = 33) or repeated model runs ( n = 12). Trends regarding reporting model versions and prompts were inconsistent. Standardized reporting and ethical guidance are needed to support reproducible use.
As socially capable artificial intelligence (AI) rapidly advances in sophistication, prevalence, and ubiquity, it becomes increasingly important to understand how human users perceive AI-generated messages. The present study examined whether users’ AI self-efficacy and attitudes toward AI predict their evaluations of AI-generated text. One hundred ninety-one participants rated 32 GPT-4-generated passages on likeability and perceived informational value, and completed measures of AI self-efficacy and attitudes toward AI. Multiple regression models were used to analyze the self-report data. Results indicated that both trait-level constructs contributed unique variance to each outcome, with AI self-efficacy more closely associated with perceived value and attitudes more closely associated with likeability. These findings suggest that perceived interaction competence and general evaluative orientation independently contribute to user experience with socially capable AI agents, which have practical implications for the design of AI systems.
Multiagent systems deployed in dynamic environments depend on timely and accurate task allocation between human operators and autonomous agents. Current approaches predominantly assign leadership authority to a single agent, either a human supervisor or an automated arbitrator, leaving out examination of shared leadership. This study compared shared and centralized leadership in a simulated search-and-rescue task in which 46 participants coordinated with one autonomous agent and four unmanned aerial vehicles (UAVs) to detect and classify objects. Under centralized leadership, participants held unilateral authority over task allocation. Under shared leadership, authority was negotiated each trial through a claiming-granting mechanism. Results showed that shared leadership produced significantly higher performance than centralized leadership. Analysis of UAV assignment revealed that perceived competence predicted allocation across both structures, while trust was significant only under shared leadership. These findings demonstrate that shared authority yields measurable performance advantages and engages reliance mechanisms differently than fixed-authority.
