Human Factors最新文献

ObjectiveTo examine how domain-switching and user characteristics may predict broad susceptibility to digital deception.BackgroundDespite successful automated filtering techniques, humans remain vulnerable to fraud, losing billions of dollars annually. Many scams are delivered by digitally mediated methods, such as phishing emails or fake social media accounts. However, research typically explores susceptibility to these deceptions independently, making it difficult to draw broad conclusions regarding susceptibility to digital deception.MethodWe recruited a representative sample to investigate how susceptibility to deception may vary across digital domains, particularly when switching between domains (i.e., domain-switching). Participants classified stimuli from five different digital domains (i.e., emails, text messages, news headlines, social media accounts, and voicemails), either randomly (i.e., domain-switching) or in separate blocks, and completed measures of cognitive reflection and digital literacy.ResultsThe results suggest that when users struggle to discriminate between deceptive and legitimate stimuli in one digital deception domain, they are likely to struggle in others. Additionally, the results suggest that while cognitive reflection and digital literacy may help insulate users from deception, domain-switching may generally hinder user performance (i.e., slower responses).ConclusionOverall, individuals appear to be consistently vulnerable to deception across digital domains and this vulnerability can be exacerbated by certain task factors (e.g., domain-switching) and user characteristics (e.g., cognitive reflection and digital literacy).ApplicationTo develop more efficacious interventions that enhance user resiliency, research should consider broad training that incorporates correlates of susceptibility (e.g., cognitive reflection and digital literacy), and more realistic task settings (e.g., domain-switching).

ObjectiveTo examine the effect of dual tasking on hand dominance during a bimanual visuomotor task.BackgroundMany operators need to perform separate tasks with each hand. Yet, there is no comprehensive study examining whether the right-hand visuomotor advantage found in right handers remains stable, increases or attenuates when another task is performed concurrently with the other hand.MethodsTwenty-eight right-handed participants (mean age = 22) performed 2D visuomotor tracking under either unimanual (one target, one hand) or bimanual conditions (two targets, one for each hand). Various gaze contingencies and visual displays were tested. Tracking performance of each hand was evaluated through the mean cursor-target distance.ResultsA clear right-hand advantage was found under all unimanual conditions. Under bimanual conditions, tracking accuracy decreased for both hands albeit more extensively for the left hand than the right when gaze was free, thus amplifying the above right-hand advantage. Prioritization of the right hand was associated with a gaze preference toward this hand. However, this increase in manual asymmetry was greatly alleviated when participants were instructed to fixate straight ahead, a benefit obtained at no cost in terms of overall tracking performance.ConclusionsDuring bimanual/dual tracking, there is a natural tendency for right handers to prioritize their right hand. However, this effect is strongly reduced by fixating straight ahead.ApplicationPerforming separate tasks with the right and left hands is common when piloting an aircraft. Fixating straight ahead may be useful for pilots that seek to divide more equally the negative impact of dual/bimanual tasking.

ObjectiveWe explored transfer of learning from brief practice with different input devices in the Navy's Performance Based Measures Battery (PBM), a psychomotor subset of the Aviation Selection Test Battery (ASTB).BackgroundThe PBM is a set of computerized tests used as a part of the ASTB to select aviators in the U.S. military. Official practice is not available, leading candidates to practice with unofficial re-creations and with or without access to the stick and throttle used on the PBM.MethodOur between-subjects study with 152 cadets from the U.S. Military Academy evaluated the impact of mouse/keyboard or stick/throttle practice on the psychomotor portions of the PBM compared to a control group that was only presented with an informational video.ResultsThe results showed that practice with either input device resulted in improved performance relative to control on the PBM's two-dimensional airplane tracking task (ATT). For the simpler vertical tracking task (VTT), the mouse/keyboard group showed significantly worse performance than either stick/throttle practice or control groups, indicating a transfer cost from practicing with an alternative input device.ConclusionThe results suggest that becoming familiar with the unique dynamics of the ATT may be more important than practicing with the appropriate input device. Conversely, device-specific motor learning appears to be a more impactful determinant of performance for the simpler VTT. This indicates that transfer effects from alternative input devices depend in part on properties of the task.ApplicationThis research can inform practice policies for psychomotor test selection.

ObjectiveTo propose a model of how attentional effort varies over time in a vigilance task and how this effort relates to subjectively inferred context. To propose an estimation methodology and test the empirical validity of the proposed model in a naturalistic dataset.BackgroundAttentional effort in a task can vary based on how an individual subjectively perceives the task context. However, both attention exertion and subjective context perception are not directly observable. We present a methodology for estimating a structural model that explicitly incorporates subjective models of context perception and attention allocation policies. To our knowledge, this is the first methodology to estimate a structural model of attentional effort dynamics.MethodA Bayesian model of attentional allocation that integrates subjective perceptions of task-relevant context is developed. An estimation methodology based upon expectation-maximization algorithm is proposed to uncover how the allocation of attentional effort is adapted to subjectively perceived context.ResultsThe methodology is applied to a naturalistic dataset of Major League Baseball umpire decisions, revealing context perception (i.e., how umpires infer game situations) and attention allocation policy (i.e., how umpires adjust attentional effort). Model reveals that umpires adjust attentional effort based on inferred game criticality and status bias.ConclusionThis work advances understanding of vigilance failure by providing a structural account for contextual inference determines attentional effort. The estimated model closely tracks empirically observed decision accuracy patterns in a naturalistic dataset.ApplicationThe proposed model enables counterfactual predictions, allowing exploration of hypothetical interventions to improve decision accuracy in environments that require sustained attention.

ObjectiveIn this study, we aimed to show that post-task self-reported mind-wandering can be influenced by task performance.BackgroundRetrospective self-report scales are widely used to measure thought content such as task-unrelated thoughts or mind-wandering in sustained attention or vigilance research. Self-reported thought content is presumed to be a predictor of performance. However, it is possible performance affects how people report their thought content.MethodIn a remote online experiment, we used a fixed order Sustained Attention to Response Task (SART) to force errors by manipulating an expected stimulus. We then assessed self-reported thought content.ResultsWe were successful in forcing errors in the SART. Participants in the forced error version of the task reported having higher task-unrelated thoughts than those participants in a version of the task which did not force an error, despite the tasks being identical up until the forced error.ConclusionPost-task thought content probes (and similar thought content measures) are apparently affected by task performance despite their conventional use as a predictor of that performance. The current method of using post hoc thought content probes is thus a poor choice for studying the impact of thought content on performance.ApplicationA fixed order SART with forced errors is a novel way to investigate relationships between performance and self-report measures of thought content.

ObjectiveObserve how instruction to avoid rounding the low back while lifting a relatively light mass impacts isometric lifting strength.BackgroundAs opposed to manual materials handling training directives recommending whole-body techniques such as a squat lift, targeting specific body regions such as low back curvature, theoretically affords workers greater flexibility to organize the rest of the body to reduce musculoskeletal loading without reducing physical performance. However, providing these directives during sub-maximal tasks may not prompt prioritization of physical performance as individuals self-organize, eventually making the intervention ineffective.MethodsForty participants (50% female) lifted a crate with and without the instruction to avoid rounding the low back. Postures at the initiation of crate lifting were replicated to test isometric strength.ResultsAt the group-level, instruction decreased low back flexion (p < 0.0001) but did not change strength (p = 0.862). However, high heterogeneity motivated examining individual responses. Thirty-seven participants (92.5% of the sample) exhibited greater than 40% of their flexion range-of-motion during baseline lifting, a threshold below which passive tissue strain is typically minimized. Yet, 22 participants (55%) were unsuccessful in reducing low back flexion below this threshold with instruction. Independent from these postural response groups, 23 maintained (57.5%), 8 increased (20%) and 9 decreased (22.5%) isometric strength.ConclusionOn average, physical performance potential was maintained in response to a low back postural directive. However, personalized movement coaching is needed to ensure the desired response for all.ApplicationManual materials handling training should include personalized movement coaching that considers both musculoskeletal loading and performance.

ObjectiveWe examine AI trust miscalibration-the discrepancy between an individual's trust in AI and its actual performance-among university students. We assess how the length of explanations and students' expertise shape the likelihood of alignment with AI recommendations.BackgroundThe relationship between explainability and users' trust in AI systems has been scarcely addressed in the current literature, even though AI-assisted processes increasingly affect all professions and hierarchical levels. Given that human-AI relationships are often formed during education, it is crucial to understand how individual and contextual factors influence students' assessment of AI outputs.MethodWe conducted in-class experiments with 248 students from multiple universities. Participants solved GMAT questions, then viewed an AI recommendation-sometimes correct, sometimes incorrect-with varying explanation depth and eventually could revise their initial answer; student's final answer being in line with AI recommendation operationalized our measure of "trust." We estimated logistic models with control variables, including mixed-effects specifications to account for repeated observations.ResultsExplanation complexity is associated with higher trust on average, but its relevance depends on who reads it and whether AI is correct. Students who previously answered correctly exhibited lower willingness to defer, especially when AI was incorrect; conversely, agreement and consistency effects significantly amplified trust. These behavioral patterns highlight conditions under which AI-generated explanations can foster critical engagement or conversely encourage uncritical acceptance.ConclusionOur results point to a "AI knows better" heuristic at work-especially among nonexperts-where polished presentation is easily read as reliability, encouraging uncritical agreement with incorrect recommendations; in parallel, experts benefit more from deeper rationales when AI is accurate, yet still display under-reliance of correct assistance in many cases. Overall, trust calibration is driven less by any single cue than by the alignment of student performance, AI reliability, and explanation design, with prior agreement acting as a powerful amplifier of subsequent alignment.ApplicationOur findings imply that instructional approaches should promote independent reasoning before exposure to AI, deploy concise but diagnostically informative explanations, and include brief verification steps before accepting AI recommendations, especially for nonexperts who are more prone to harmful switches. Simple monitoring tools that track helpful versus harmful changes could support a more discerning and productive use of AI tools.

ObjectiveWe examine how risk and automation failures in conditional driving automation (SAE Level 3) influence drivers' calibration of trust and reliance behavior in the form of system use and monitoring.BackgroundConditionally automated driving brings a challenging new role for drivers, who are permitted to engage in non-driving-related activities but must take back control in certain situations.MethodsParticipants completed three drives in a driving simulation with conditional driving automation. The first drive was with low risk and the second drive was with high risk implemented in the simulation. The third drive included either early or late automation failure.ResultsParticipants reported lower trust, took over manual control more often, and monitored more when driving under high risk than when driving under low risk. After experiencing an automation failure, trust decreased immediately but fully recovered over time. Driver's monitoring increased and decreased immediately as the failure started and ended. The timing of automation failure did not influence its impact on trust.ConclusionThe results indicate that drivers respond appropriately to risk. Trust develops dynamically in case of an automation failure, but failure timing does not influence this process. From an applied perspective, drivers would benefit from assistance in re-calibrating trust after automation failure.ApplicationBased on our findings, we argue that incorporating drivers' mental model formation process into the feedback loop of trust and reliance behavior calibration could enhance the theoretical understanding of trust calibration.

ObjectiveThis study identifies cybersecurity vulnerabilities and risks in robotic-assisted surgery (RAS) and proposes a cybersecurity framework and an assessment tool for RAS systems.BackgroundRAS systems are increasingly integrated into networks which raise cybersecurity concerns. These systems can enhance surgical outcomes but are potential cyberattack targets, which can affect clinician care, patient safety, and organizational operations.MethodSurveys and interviews were conducted with stakeholders (clinicians, researchers, cybersecurity professionals, and hospital administrators) to collect perspectives on RAS cybersecurity. Thematic analysis was used to develop an RAS cybersecurity framework. Then, stakeholders contributed to creating an RAS cybersecurity assessment tool using Failure Modes, Effects and Criticality Analysis (FMECA).ResultsSurvey responses (n = 84) revealed that 48.8% of respondents were familiar with RAS cybersecurity. Only 24.6% of clinical respondents were aware of their organization's cybersecurity policy. Interviews (n = 15) identified vulnerabilities such as inadequate training, limited communication between manufacturers and healthcare systems, and gaps in regulations. Failure modes focused on consequences of cyberattacks on RAS systems, with severity assessments related to patient health and technology reliability/integrity completed and outcome actions identified.ConclusionUnderstanding RAS cybersecurity challenges is still in its infancy. Key vulnerabilities include insufficient training, limited data sharing, and external threats. The framework illustrates the interconnectedness of stakeholders, while the FMECA assessment tool addresses current vulnerabilities in RAS systems.ApplicationRAS cybersecurity vulnerability and risks should be carefully considered when integrating systems into healthcare organizations, and the RAS cybersecurity assessment tool can be used by stakeholders to systematically identify and analyze potential cybersecurity failure modes.