Human Factors最新文献

ObjectiveWe conducted two experiments to understand the effects of computationally diminishing reality on performance, awareness of the environment, and subjective workload.BackgroundAdvances in extended reality (XR) technologies make it possible to alter or remove auditory and visual distractions from an environment. Though distractions are known to harm performance, there is no work examining the effects of removal via XR.MethodAcross two samples, STEM graduate students and Johnson Space Center employees, the effects of reducing distraction during a novel, demanding assembly task via a form of XR (diminished reality) were compared to a full distraction control condition, studied in a virtual reality (VR) environment. In one condition, participants experienced universal attenuation of distractions. In a second condition, attenuation was context-aware: only nontask objects were made less visible and only unimportant off-task audio was eliminated.ResultsBoth experiments found subjective workload could be lowered via a Diminished reality (DR) aid. The STEM graduate student sample showed a benefit of a DR aid for performance and environment awareness; however, the sample of professionals from Johnson Space Center showed no performance differences with the DR aids. There were mixed results regarding awareness of the location of objects and events outside of the assembly task.ConclusionDR aids can have effects similar to those seen in studies that removed distractions entirely. More work is needed to understand the match between distraction removal design and task.ApplicationThese findings contribute to the development of a class of XR aids: Diminished Reality.

ObjectiveThe goal of this meta-analysis is to investigate the effect of partial automation on mental workload, visual behavior, and engagement in nondriving-related tasks.BackgroundThe literature on the human factors of operating partially automated driving offers mixed findings. While some studies show partial driving automation to result in suboptimal mental workload, others found it to impose similar levels of workload to the ones observed during manual driving. Likewise, while some studies evidence a marked increase in off-road glances when the automated system was engaged, other work has failed to replicate this pattern.Method41 studies involving 1482 participants were analyzed using the PRISMA approach.ResultsNo significant differences in mental workload were found between manual and partially automated driving, indicating no changes in mental workload between the two driving modes. A higher likelihood of glancing away from the forward roadway and engaging in nondriving-related tasks was found when the partially automated system was engaged.ConclusionAlthough the adoption of partial driving automation comes with some intended safety benefits, its use is also associated with an increased engagement in nondriving-related activities.ApplicationThese findings add to our understanding of the safety of partial automation and provide valuable information to Human Factors practitioners and regulators about the use and potential safety risks of using these systems in the real-world.

ObjectiveTo study the performance consequences of binary versus likelihood decision support systems in low-prevalence visual search.BackgroundHit rates in visual search are often low if target prevalence is low, an issue that is relevant for numerous real-world visual search tasks (e.g., luggage screening and medical imaging). Given that binary decision support systems produce many false alarms at low prevalence, they have often been discounted as a solution to this low-prevalence problem. By offering additional information about the certainty of target-present indications through splitting these into warnings and alarms, likelihood-based systems could potentially boost hit rates without raising the number of false alarms.MethodWe used a simulated medical search task with low target prevalence in a paradigm where participants sequentially uncovered parts of the stimulus with their mouse. In two sessions, participants completed the task either while being supported by a binary or a likelihood system.ResultsHit rates were higher when interacting with the likelihood systems than with the binary system, at no cost of higher false alarms.ConclusionLikelihood systems are a promising way to tackle the low-prevalence problem, and might further be an effective means to make systems more transparent.ApplicationSimple-to-process information about system certainty for each case might be a solution to low hit rates in domains with low target prevalence, such as radiology.

ObjectiveWe investigated effects of a Back-support industrial exoskeleton (BSIE) across intermittently performed unloaded trunk bending task cycles.BackgroundIndustrial tasks are often performed in the form of task cycles with varying activities and rest breaks after each task cycle. Investigating the effects of BSIEs during such intermittent tasks is crucial to understand translation of their benefits in real-world environments.MethodTwelve participants performed ∼709 task cycles (sustained bending, bending and retraction, standing still, and relaxation activities) with/without BSIE (E/NE) and with/without 45° asymmetry (S/A) towards left until fatigue. Evaluated measures included muscle activity in left (LES)/right (RES) erector spinae and left (LBF)/right (RBF) biceps femoris muscles, endurance, and user perspectives. Temporal effects of fatigue were examined by categorizing muscle activity based on perceived exertion level with Borg scale.ResultsBSIE reduced low-back (LES, RES), and leg (LBF, RBF) mean amplitude by ∼ 18-24% and ∼10-17% respectively. Benefits of BSIE in low-back reduced by ∼11-15% at medium versus low fatigue. Overall, BSIE led to 50% more completed task cycles and was favorably rated by participants in reducing physical demands, most especially during sustained bending portion of the task cycles.ConclusionUsing BSIE in intermittent bending tasks can not only provide benefits in reducing muscle demands but can also delay effects of fatigue in low-back region and increase endurance by enabling wearers to perform more task cycles.ApplicationFindings from this study may be beneficial to practitioners for setting guidelines on implementation of BSIEs in industrial bending tasks.

ObjectiveThis study investigates the effects of workload and task priority on multitasking performance and reliance on Level 1 Explainable Artificial Intelligence (XAI) systems in high-stakes decision environments.BackgroundOperators in critical settings manage multiple tasks under varying levels of workload and priority, potentially leading to performance degradation. XAI offers opportunities to support decision making by providing insights into AI's reasoning, yet its adoption and effectiveness in multitasking scenarios remain underexplored.MethodThirty participants engaged in a simulated multitasking environment, involving UAV command and control tasks, with the assistance of a Level 1 (i.e., basic perceptual information) XAI system on one of the tasks. The study utilized a within-subjects experimental design, manipulating workload (low, medium, and high) and AI-supported-task priority (low and high) across six conditions. Participants' accuracy, use of automatic rerouting, AI miss detection, false alert identification, and use of AI explanations were measured and analyzed across the different experimental conditions.ResultsWorkload significantly hindered performance on the AI-assisted task and increased reliance on the AI system especially when the AI-assisted task was given low priority. The use of AI explanations was significantly affected by task priority only.ConclusionAn increase in workload led to proper offloading by relying on the AI's alerts, but it also led to a lower rate of alert verification despite the alert feature's high false alert rate.ApplicationThe findings from the present work help inform AI system designers on how to design their systems for high-stakes environments such that reliance on AI is properly calibrated.

ObjectiveDynamic measures of team adaptation based in team cognition theory and the measurement of real-time team cognition are developed. The present study examines the validity and context-specificity of this measurement framework for simulation-based team training.BackgroundTeams adapt by reorganizing their coordination behavior to overcome challenges in dynamic environments. Theoretically grounded objective metrics for measuring adaptive skill in teams are needed. We developed dynamic measures of team adaptation to help fill this gap.MethodCommunication data from critical care air transport team training were analyzed using moving window entropy and recurrence-based determinism metrics of communicative adaptation in response to training event perturbations involving stabilizing deteriorating patient status. The measures were validated across four simulation-based training scenarios using objective and subjective metrics of team performance.ResultsWe validated performance prediction in all scenarios, demonstrating generalizability. Critically, teams reorganized significantly more during perturbation segments than routine segments, validating the measures as indices of team adaptation. We also observed context-specificity, wherein the relationships between reorganization and successful performance depended on the training scenario.ConclusionThe communicative reorganization measures advanced in this paper present a valid method for assessing adaptive competencies in teams. These analytics generalize in terms of performance prediction across training scenarios, but they are also context-specific, wherein patterns of effective reorganization depend on the type of scenario.ApplicationWe discuss the practical deployment of the measurement framework in a Team Dynamics Measurement System for assessing team adaptation competencies in critical care air transport team training.

ObjectiveThis study evaluated whether pedestrians can use augmented reality (AR) overlays to guide their road-crossing decisions when crossing two lanes of opposing traffic.BackgroundEmerging technologies for enhancing traffic safety often focus on alerting drivers to hazards. Less attention has been given to understanding how pedestrians respond to technology designed to aid their road-crossing decisions, particularly in more complex traffic.MethodParticipants repeatedly crossed two lanes of opposing traffic displayed in a virtual reality system. Participants in the AR condition viewed matching-colored bars (AR overlays) suspended just above the gaps between cars where there was sufficient time to safely cross a pair of near and far lane gaps. Participants in the control condition performed the same road-crossing task but saw no AR overlays.ResultsParticipants who viewed AR cues were more likely than participants who did not view AR cues to accept gap pairs classified as crossable and less likely to accept gap pairs classified as uncrossable. However, there was no difference between the AR and control conditions in time to spare when exiting the roadway. NASA Task Load Index (2020) responses indicated that perceived performance was higher and perceived frustration was lower in the AR than control condition, but perceived workload was higher in the AR condition.ConclusionThe AR cues helped participants identify crossable gap pairs but did not lead to greater time to spare when exiting the roadway.ApplicationThese results show both the promise and risks of assistive technologies designed to increase pedestrian safety in more complex traffic situations.

ObjectiveThe impact of the context in which automation is introduced to a decision-making system was analyzed theoretically and empirically.BackgroundPrevious work dealt with causality and responsibility in human-automation systems without considering the effects of how the automation's role is presented to users.MethodsAn existing analytical model for predicting the human contribution to outcomes was adapted to accommodate the context of automation. An aided signal detection experiment with 400 participants was conducted to assess the correspondence of observed behavior to model predictions.ResultsThe context in which the automation's role is presented affected users' tendency to follow its advice. When automation made decisions, and users only supervised it, they tended to contribute less to the outcome than in systems where the automation had an advisory capacity. The adapted theoretical model for human contribution was generally aligned with participants' behavior.ConclusionThe specific way automation is integrated into a system affects its use and the perceptions of user involvement, possibly altering overall system performance.ApplicationThe research can help design systems with automation-assisted decision-making and provide information on regulatory requirements and operational processes for such systems.

ObjectiveTo better understand automation transparency, we experimentally isolated the effects of additional information and decision recommendations on decision accuracy, decision time, perceived workload, trust, and system usability.BackgroundThe benefits of automation transparency are well documented. Previously, however, transparency (in the form of additional information) has been coupled with the provision of decision recommendations, potentially decreasing decision-maker agency and promoting automation bias. It may instead be more beneficial to provide additional information without decision recommendations to inform operators' unaided decision making.MethodsParticipants selected the optimal uninhabited vehicle (UV) to complete missions. Additional display information and decision recommendations were provided but were not always accurate. The level of additional information (no, medium, high) was manipulated between-subjects, and the provision of recommendations (absent, present) within-subjects.ResultsWhen decision recommendations were provided, participants made more accurate and faster decisions, and rated the UV system as more usable. However, recommendation provision reduced participants' ability to discriminate UV system information accuracy. Increased additional information led to faster decisions, lower perceived workload, and higher trust and usability ratings but only significantly improved decision (UV selection) accuracy when recommendations were provided.ConclusionIndividuals scrutinized additional information more when not provided decision recommendations, potentially indicating a higher expected value of processing that information. However, additional information only improved performance when accompanied by recommendations to support decisions.ApplicationIt is critical to understand the potential differential impact of, and interaction between, additional display information and decision recommendations to design effective transparent automated systems in the modern workplace.

ObjectiveThis study investigated the effect of reliability on the function allocation (FA) boundary by examining the interaction effect of degree of automation (DOA) and reliability on routine performance, failure performance, and attention allocation.BackgroundAccording to the lumberjack effect, an increase in DOA will typically improve routine performance, while failure performance may remain undeteriorated until a specific, high DOA threshold is reached. This threshold can be regarded as the FA boundary. Considering that both DOA and reliability can influence failure performance through attention allocation, it is crucial to investigate how reliability affects the FA boundary.MethodParticipants performed three MATB tasks, one of which, the system monitoring task, was supported by four types of automation: information acquisition (IAc), information analysis (IAn), action selection (AS), and action implementation (AI). From IAc to AI, the DOA incrementally increased. Additionally, automation reliability was set to three levels, namely, 87.50%, 68.75%, and 56.25%.ResultsFor routine performance, participants assisted by AS reacted more rapidly to gauge malfunctions than those supported by IAc or IAn. For failure performance, participants aided by AI corrected gauge malfunctions less frequently than other participants. Correspondingly, participants supported by AI exhibited fewer fixation counts on the system monitoring task than did others.ConclusionIt appears that the FA boundary lies between AS and AI. However, there is insufficient evidence to support the effect of reliability on the FA boundary.ApplicationThese findings can provide useful insights for improving the design of automated systems in complex working environments.