ObjectiveThis study examined how learning, workload and search behaviors were impacted by a chatbot during a self-regulated Web search task, as opposed to more classic search engines.BackgroundArtificial intelligence technologies, including chatbots, are becoming increasingly accessible. These tools have been demonstrated useful to support self-regulated learning, mostly in structured learning contexts. The reduction in workload they offer may, however, prevent key learning strategies from being deployed, especially for Web search.MethodSixty participants were asked to answer a set of essay questions and to rate their workload, effort deployed and literacy while either gathering information from the Internet (Web condition) or by chatting with a chatbot (LLM condition) with the possibility of verifying information on the Web. Several key search strategies and chatbot interaction measures were extracted. A surprise memory test was also presented to evaluate how they learned the content addressed in the essay questions.ResultsMeasures of effort, mental workload, search behaviors and literacy differed significantly across conditions. Performance on the memory test did not vary. Multiple relationships with memory performance, including key Web search and verification behaviors, were found.ConclusionChatbots may help reduce workload and short-term learning with a chatbot may be more influenced by the nature of the interaction with the tool, rather than the tool itself.ApplicationEffective uses of chatbots may require learners to verify the content generated by the chatbot and to show superior engagement. Engagement-promoting learning activities should be considered when using LLM-driven agents to support self-regulated, Web-based search.

ObjectiveTo examine how cognitive workload influences physiological responses, driving performance, and subjective workload in underground coal mine transport, with comparisons between novice and experienced drivers.BackgroundCognitive workload is a key determinant of safety and performance in high-risk driving environments. However, little is known about how driver experience shapes responses to workload demands in underground coal mine transport operations.MethodsSixty drivers (30 novices, 30 experienced) participated in four experimental conditions (baseline, warning stimulus, distraction, dual-task). Data from 54 participants were included in the final analysis after excluding unusable recordings. Physiological indicators (heart rate, heart rate variability, respiratory rate, and variability), driving performance metrics, and subjective workload (NASA-TLX) were analyzed. Linear mixed models were applied to examine relationships among workload, physiology, and performance.ResultsIncreased workload elevated physiological strain and subjective demand while impairing driving performance. Novices exhibited a linear decline in performance, whereas experienced drivers showed possible nonlinear workload-performance patterns.ConclusionDriver experience moderates the workload-performance relationship. Integrating physiological, performance, and subjective indicators provides a more comprehensive understanding of workload effects in underground driving contexts.ApplicationFindings can guide the development of adaptive driver support systems that dynamically respond to workload fluctuations, enhancing safety, reducing errors, and improving operational efficiency in high-risk mining transport environments.

ObjectiveTo characterize the Phase 1 manual anthropometry component of YouthShape.US, document its 57-measure protocol and quality-control procedures, evaluate selected design-relevant differences from legacy and contemporary national reference data, and introduce the publicly accessible YouthShape.US resource.BackgroundAccurate anthropometric data are essential for designing products that safely accommodate children. The most comprehensive U.S. child anthropometry survey was conducted in the 1970s and no longer reflects current populations. Contemporary national sources include few dimensions, while other studies emphasize health outcomes or proprietary sizing systems.MethodPhase 1 of YouthShape.US collected laboratory data from 650 children aged 3-17 years. Participants completed a standardized protocol comprising 57 manual anthropometric measurements obtained with calibrated instruments in standing and seated postures, along with body-composition measures.ResultsComparisons of 90th percentile values across age showed important differences among datasets. Stature and body weight generally tracked contemporary NHANES estimates more closely than the legacy 1970s U.S. child anthropometry values, although some age- and sex-specific deviations were evident. Body weight and waist circumference were substantially larger than values reported in the 1970s, and buttock-knee length was also larger, indicating changes in seated body dimensions relevant to product and environment design.ConclusionYouthShape.US establishes the first large-scale, detailed, publicly accessible anthropometric resource for U.S. children and adolescents in more than five decades and demonstrates substantial secular changes, particularly in body-mass-related dimensions.ApplicationYouthShape.US provides interactive tools that enable practitioners to explore contemporary body dimensions relevant to product fit, safety, and ergonomic design.

ObjectiveWe develop a dynamic model to understand how trust in automation evolves within teams through team member interactions, assessing whether trust converges or diverges over time.BackgroundExisting research assumes that individual and team trust in automation converges over time, but these studies typically last only hours or days. This assumption may not hold for long-duration missions, like deep space exploration, where team dynamics and trust might evolve in a contingent rather than convergent manner. Contingent behavior evolves towards different endpoints based on small perturbations, whereas convergent behavior evolves to similar endpoints.MethodsWe developed a stochastic, discrete-event agent-based trust dynamics model that goes beyond existing models that only consider past interactions with automation. Our model incorporates team conversations, individual automation experience, and turn-taking interactions.ResultsConsistent with human subjects data, the model showed divergent trust behavior where team members' trust levels did not converge to similar values over time.ConclusionsDynamical models of trust in teams can show contingent behavior. Trust in automation within and across teams can diverge, indicating a new mechanism for trust dynamics. Trust calibration strategies should address potential divergence.ApplicationsDesigners should consider the divergence of trust within and between teams, especially for long-duration missions. Methods to calibrate trust in this situation may include structured debriefs or shared automation feedback displays.

ObjectiveThis study examined whether cognitive load produces selective effects on different trust updating pathways in AI-assisted decision making.BackgroundAlthough cognitive load affects trust in automation, its influence on the mechanisms of trial-by-trial trust updating remains unclear.MethodsA dual-task paradigm embedded in a mining exploration task manipulated cognitive load while capturing dynamic trust calibration. Guided by a dual-pathway framework, we operationalized process-based (analytical evaluation of AI recommendation correctness) and outcome-based (heuristic reliance on task outcomes) trust updating pathways. Trust dynamics and behavioral reliance were examined using linear mixed-effects models.ResultsCognitive load shifted the relative influence of the two trust updating pathways. Process-based updating was attenuated under high cognitive load, indicating reduced sensitivity to AI recommendation correctness during trust updating. Outcome-based information gained greater influence under high load, amplifying outcome-driven bias regardless of recommendation correctness. Asymmetric trust updating was evident overall, although the influence of cognitive load on this asymmetry depended on task outcomes. Overall, high cognitive load elevated both subjective trust and behavioral reliance on AI.ConclusionCognitive load shapes trust calibration through mechanism-level reconfiguration rather than global impairment. By revealing how cognitive constraints rebalance dual trust pathways-weakening analytic evaluation while amplifying heuristic outcome reliance-this study advances theoretical understanding of dynamic trust in human-AI collaboration.ApplicationThe results provide practical guidance for the design of AI systems in high-stakes settings, highlighting the need to support analytic trust updating and mitigate over-reliance under cognitive strain.

ObjectiveThis narrative review examines the cognitive, metacognitive, and team competency requirements that may contribute to productive and reliable collaboration between human and AI to address two questions: What capabilities make AI a competent collaborator? What makes humans ready for AI collaboration?BackgroundAs AI systems are increasingly integrated into workplaces and framed as teammates rather than tools, humans face challenges that include maintaining situation awareness, calibrating trust, and working with systems that may surpass them cognitively. We analyzed Human-Agent Teaming (HAT) readiness around two complementary levels: operational team competencies (communication, coordination, and adaptability) and regulatory capacities (trust calibration and metacognitive awareness).MethodWe conducted a structured narrative review of literature from 2010 through January 2026, searching Google Scholar, Scopus, PsycINFO, IEEE Xplore, ACM Digital Library, and Semantic Scholar, complemented by forward citation tracking. After screening 572 records, 192 articles were included for synthesis.ResultsCommunication inflexibility, limited shared understanding, and trust miscalibration emerge as recurring barriers to HAT, while regulatory capacities (trust calibration and metacognitive awareness) represent particularly critical dimensions of HAT readiness that remain to be fully operationalized.ConclusionHAT requires mutual readiness, with both humans and AI developing metacognitive and adaptive capabilities. Despite methodological heterogeneity limiting clear conclusions, cross-training and co-learning methods offer a promising avenue for building shared understanding and calibrated collaboration.ApplicationThis review provides practical principles for designing AI systems that support calibrated collaboration and for preparing humans to work adaptively with AI, thereby enhancing team effectiveness, reliability, and resilience in collaborative work environments.

ObjectiveThis study aimed to investigate the effects of different monitoring systems on human performance and user experience in safety-critical diagnostic and operational tasks, and examine how these effects are moderated by task-related and human-related factors.BackgroundMonitoring systems are increasingly adopted in safety-critical domains to enhance performance and safety. However, these systems may present both benefits and potential drawbacks, and comprehensive understanding of their multifaceted effects remains limited.Method32 participants performed diagnostic and operational tasks using a nuclear power plant simulator. A mixed experimental design examined three independent variables: monitoring system (human monitoring, camera monitoring, no monitoring as a baseline), task type (accident, non-accident), and participant group (nuclear-based, engineering/science-based). Dependent variables included human performance metrics (accuracy, task completion time) and user experience (stress, workload, privacy concerns, trust, and technology acceptance).ResultsHuman monitoring was associated with significantly higher accuracy compared to no monitoring, but also increased stress and workload. Camera monitoring generally exhibited intermediate values between no monitoring and human monitoring, while eliciting the highest privacy concerns among all conditions. Significant interaction effects emerged between monitoring systems, participant groups, and task types on stress, trust, and technology acceptance.ConclusionMonitoring conditions exhibited double-edged effects: enhancing human performance while simultaneously increasing the perceived burden. These effects were moderated by task characteristics and users' background.ApplicationThese findings can inform the human-centered design of monitoring systems and development of regulatory guidelines in safety-critical domains, emphasizing the need to balance the performance benefits with stress-related human factors issues.

ObjectiveTo explore how novice programmers' trust in Artificial Intelligence-driven Development Environments (AIDEs) relates to their coding performance and AI compliance while programming under time pressure.BackgroundComputer programming has undergone rapid upheaval due to state-of-the-art AIDEs, which provide clever automation for many aspects of software development. A longstanding interest of researchers of automation more generally has been the attitude of trust. Decades of research seek to explain how influencing trust can help to achieve desirable outcomes in different domains, but very limited work has provided similar focus on trust in AIDEs.MethodWe collected subjective measures of trust along with objective measures of performance and AIDE compliance from a diverse group of 27 novice programmers between two study locations.ResultsOur results corroborated traditional understandings of how trust changes through experiences. However, we did not find a relationship between trust and subsequent compliance during programming tasks. Greater compliance was associated with strong performance, and strong performance led to greater subsequent trust.ConclusionOur findings raise new questions about the utility of trust in the context of interacting with AIDEs and generative AI. We call for further research into the effect of trust on compliance to recommendations from imperfect AI.ApplicationThis work can inform the design of training and educational content for generative AI use within and beyond software development. Instructional designers should consider risks of AI misuse and disuse and focus on promoting desirable interaction outcomes, regardless of trust's connection to them.

ObjectiveWe introduce two novel scales to measure two important aspects of motion sickness, dizziness and eye strain.BackgroundThe Fast Motion Sickness Scale (FMS) has become a valuable tool to assess motion sickness whenever quick unobtrusive measures need to be taken. It is well-anchored and provides scalable, quantitative data. However, it does not differentiate among symptoms of motion sickness.MethodWe discuss the benefits and shortcomings of the FMS as well as other, symptom-based rating scales. We present a taxonomy to assess symptom clusters and illustrate the trade-off between fast global and more detailed measures of motion sickness correlates.ResultsWe introduced two separate FMS variants, focusing separately on oculomotor discomfort (FMS-O) and dizziness (FMS-D). Observers have successfully used these scales.ConclusionThe novel scales, FMS-D and FMS-O, can summarily capture, in a non-redundant fashion, dizziness and eye strain. For the third cluster, nausea, the original FMS scale remains successful. Together, they capture the main symptom clusters of motion sickness.ApplicationTogether with the original FMS, the new scales provide a powerful instrument that can be used to sample feelings of dizziness, eye strain, and nausea at frequent intervals. For instance, taken once per minute, they can efficiently monitor well-being in real time in challenging situations, such as autonomous driving or Virtual Reality applications.

ObjectiveTo develop and preliminarily validate a behavior- and height-based strategy-group-specific dynamic seat-adjustment model that generates coordinated low-frequency seat trajectories to reduce subjective discomfort during prolonged public-transport sitting.BackgroundDynamic seating may relieve sitting-related discomfort, but existing strategies rarely model coupled seat adjustments while considering behavioral and anthropometric variability.MethodsSixty-six adults completed counterbalanced 2-h resting and phone-use sessions. Every 15 min, they adjusted seven seat dimensions to their preferred configuration and rated overall and regional discomfort before and after adjustment. An improvement-driven weighted multi-output random forest was trained within Behavior × Height-Based Strategy Group strata, using elapsed sitting time as the only input and seven-dimensional seat configuration as the output; discomfort improvement was used only for sample weighting. Predicted configurations were smoothed into low-frequency trajectories. An independent within-subject crossover experiment compared Static, Mean-trajectory, and Dynamic conditions.ResultsPreferred configurations differed by behavior and height-based strategy group and changed systematically over time. Multi-output forests outperformed independent single-output models, supporting coordinated trajectory learning, whereas improvement-driven weighting mainly provided a relief-oriented training bias. In validation, the Dynamic condition reduced overall discomfort relative to Static seating across both behaviors and time points. Compared with the non-individualized Mean-trajectory baseline, Dynamic seating produced additional region- and behavior-dependent perceived-comfort benefits, especially in neck and trunk-related regions.ConclusionA stratified multi-output model can generate coordinated time-varying seat trajectories associated with lower subjective discomfort.ApplicationThis framework offers a preliminary basis for low-frequency passive adjustment in transportation seats with comparable geometry while requiring limited passenger input.

ObjectiveThis systematic review and meta-analysis (SRMA) aimed to synthesize and critically appraise existing experimental evidence regarding the impact of heat exposure on driving performance.BackgroundClimate change, characterized by rising temperatures, has been linked to more road accidents and related injuries. Nonetheless, no comprehensive systematic review or meta-analysis has examined the direct effects of increased ambient temperature on driving performance.MethodRegistered in PROSPERO (#CRD42024584269), this SRMA searched major databases and grey literature from their inception to October 2024 and appraised the quality and confidence of the evidence via an established research methodology. A final pool of 7 studies, totaling approximately 207 participants, met the inclusion criteria.ResultsOf the five coded dependent variables, only reaction time (RT) had a sufficient number of studies to support a meta-analysis. There was moderate confidence that heat exposure impairs drivers' RT (SMD = 0.72, 95% CI: 0.32-1.12, p = 0.0004, I[2] = 65%, p = 0.02). On the other hand, insufficient evidence precluded a definitive assessment of heat's impact on other driving performance metrics, including lateral control, longitudinal control, signal detection, and driving violations, owing to inconsistent findings, limited data and heterogeneity of performance criteria.ConclusionThis synthesis confirms that heat is likely to impair RT, yet evidence concerning other performance aspects remains insufficient. The current study serves as a foundational step in systematically documenting a critical evidence gap in a topic of mounting public health relevance.ApplicationThis paper provides insights for technological and infrastructural adaptations to enhance road safety in the context of rising global temperatures.

ObjectiveThis study aimed to isolate the source of memorability benefits in recognition-based graphical authentication. The effects of retrieval (recall vs. recognition) and font representation (plain vs. distinctive) on usability measures were examined.BackgroundStrong system-assigned passwords are difficult to remember. This motivates next-generation authentication to develop schemes that are memorable. Some researchers claim that their memorability improvement is due to dual-coding in long-term memory, while others highlight recognition retrieval. A distinctiveness theory was employed to transform text into images. This study examined whether text distinctiveness and recognition could yield memorability improvements while still using strong system-assigned alphanumeric passwords.MethodParticipants learned strong alphanumeric passwords using both traditional (recall) and Use Your Illusion (UYI) (recognition) authentication schemes. The UYI scheme presented the passwords in either plain or distinctive fonts. After practice with each authentication type, participants completed a test login after 5 minutes and 3 weeks later. Measurements included initial practice errors, login errors, login time, and ease ratings.ResultsManipulating font distinctiveness did not significantly improve memorability. Recognition-based UYI interactions significantly outperformed recall-based password entry on all metrics after a 3-week delay. Text-based recognition achieved high memorability, comparable to performance reported for a graphical authentication benchmark.ConclusionAuthentication memorability was primarily driven by the retrieval type. Transforming passwords into images did not enhance memorability, possibly due to a ceiling effect. However, employing a recognition-based scheme dramatically increased text password memorability.ApplicationFindings suggest designers can significantly enhance authentication memorability by leveraging recognition memory, even with standard text.

ObjectiveTo investigate how different types of automated cues (direct versus indirect) of a decision support system (DSS) affect performance and attentional mechanisms in visual search.BackgroundVisual search tasks in safety- and health-critical domains are increasingly supported by DSS. While such systems commonly provide explicit cues, these cues differ in whether they directly indicate target locations or merely signal target presence. Empirical evidence comparing these cue types, particularly at the level of attentional processes, remains limited.MethodParticipants completed a simulated medical visual search task in which they searched for target letters embedded in noise. Using a within-subjects design, participants performed the task under three conditions: no DSS support, indirect cueing, and direct cueing. Human-alone performance exceeded system-alone performance. Performance was assessed using signal detection measures, and response times and eyetracking were used to examine attentional guidance and search termination.ResultsDirect cueing increased sensitivity relative to indirect cueing and no support. Response times and eyetracking measures showed that direct cues guided attention more efficiently and reduced search effort, whereas indirect cues produced attentional patterns descriptively similar to the no-cue condition.ConclusionDirect cues enhanced sensitivity and attentional efficiency in visual search, whereas indirect cues provided little benefit over unaided search when human performance exceeded system performance.ApplicationThe results highlight cue format as a critical design factor for DSS in applied visual search, suggesting that location-specific cues are particularly effective when DSS complement rather than outperform human observers.

ObjectiveTo examine how moral intensity and perceived trustworthiness of an AI certifier influence individuals' willingness to change their decision due to AI-generated suggestions.BackgroundAI-supported decision-making is increasingly used in professional contexts and situations with a high moral intensity. Yet it is unclear how different factors influence the willingness to base decisions on AI-generated suggestions in such situations. Perceived trustworthiness in the AI system is key for effective and ethically sound deployment. Investigating the interplay of these factors is critical for understanding AI-supported ethical decision-making and designing AI systems that support responsible decision-making in morally complex workplace environments.MethodWe conducted a 2 × 2 vignette-based online experiment with a representative US sample (n = 546), manipulating moral intensity and AI certifier trustworthiness. Participants made an initial workplace decision (whom to let go), received a deviating AI suggestion, and then made a second decision. Effects of independent and control variables on decision changes were analyzed using hierarchical logistic regression and χ[2-]tests.ResultsDecision makers were more likely to change decisions in low moral intensity scenarios due to AI suggestions. Certifier trustworthiness had no significant effect, whereas a positive general attitude towards AI increased the likelihood of changing decisions, suggesting that overall, AI attitude may overshadow perceived certifier trustworthiness.ConclusionHuman factors play a more prominent role than AI certifications when it comes to trust-building in AI suggestions for ethical decision-making in the workplace. Further research is needed to clarify how these factors interact with perceived certifier trustworthiness and other contextual factors.ApplicationEncouraging reliance on AI-based recommendations via certifications alone is challenging. Organizations should focus on their employees' general attitudes towards AI to support AI-based ethical decision-making in the workplace. AI systems could play a significant supporting role particularly in decision situations with high moral intensity.

ObjectiveThis study investigates how task priority, task difficulty, and their interaction influence task-switching decisions in a complex multitasking environment, where operators manage multiple supervisory tasks distributed across displays.BackgroundThe Strategic Task Overload Management (STOM) model posits that task priority and difficulty influence task-switching behavior, but empirical research has yielded inconsistent results.MethodParticipants performed four concurrent supervisory tasks, each shown on a separate display, using a simulated unmanned aerial vehicle monitoring platform. Task priority was manipulated through performance-based rewards, and the difficulty of the target detection task varied across four levels.ResultsParticipants were more likely to switch to the target detection task and dwelled longer on it when it was high-priority. Increased task difficulty led to lower overall switching frequency and longer dwell time on the more difficult task. Manipulations of priority and difficulty mutually reinforced perceptions of each other and amplified each other's impact on task-switching behaviors.ConclusionTask priority has a significant impact on task-switching behaviors in complex multitasking scenarios. The impact of task difficulty on switching behaviors is more pronounced at the global level than at the local level, where a threshold effect is observed. The two attributes mutually reinforce each other's impact.ApplicationThis study provides implications for refining the STOM model and offers insights into designing interventions that support effective task switching.