Proceedings of the Human Factors and Ergonomics Society ... Annual Meeting. Human Factors and Ergonomics Society. Annual meeting最新文献

This study examines how performance feedback influences trust and self-confidence during interactions with dynamically reliable automation. Trust and self-confidence are crucial components of human-automation collaboration, governing reliance decisions and decision-making processes. In this experiment, 80 participants engaged with an automated assistant whose reliability fluctuated across tasks, receiving performance feedback throughout. Contrary to expectations, trust and self-confidence remained stable, showing little sensitivity to changes in reliability or feedback. This suggests that performance feedback may moderate variability in trust, stabilizing perceptions of automation over time. However, this stabilization could lead to complacency and overconfidence. To develop systems that promote calibrated trust and optimize team performance, future research should investigate individual differences in trust calibration, situational awareness, and prior experience with automation. Understanding the complex interplay between feedback, trust, and self-confidence is essential for effective human-automation collaboration in dynamic environments.

Artificial intelligence (AI) systems need to adapt to changing circumstances to maintain relevance in dynamic environments. Inspired by the adaptive advantages of human forgetting, this study investigates the integration of a forgetting function into an AI system. We implemented this mechanism as a training window within the Cognitive Shadow (CS) system, an AI designed to learn and emulate human decision models. This training window hyperparameter-applicable to supervised machine learning algorithms-aims to address the issue of concept drift by prioritizing recent information. The effectiveness of this addition was tested with a simple strategy game similar in dynamics to rock-paper-scissors. Participants played individually against an AI opponent for three 60-round sessions. CS was trained during Session 1 to learn the decision patterns of the player and actively predicted and countered human decisions in Sessions 2 and 3. Analyses showed that including the training window significantly improved prediction accuracy in both Sessions 2 and 3 by emphasizing recent, relevant data. These findings highlight the potential of incorporating human-inspired forgetting mechanisms to enhance AI performance in interactive and dynamic environments, with implications for future decision support systems.

Developments in artificial intelligence (AI) are transforming everyday tasks, including accessing information, learning, and decision making. Generative AI is representative of these changes as it can generate content traditionally reserved for humans with increased efficiency and reduced effort. This includes technologies like ChatGPT and other tools that exploit large language models, typically taking the form of conversational agents (chatbots). These technologies can be useful for self-regulated learning as is the case for Web browsing. It is, however, unclear whether learning with chatbots may be efficient as opposed to other Web-based approaches given the reduced effort related to chatbot interactions. This study assessed how interacting with a chatbot may affect short-term learning and the role of mental effort. Memory performance was equivalent across participants who either interacted with a chatbot or browsed the Internet to find information for answering essay questions. Differences in self-reported workload were, however, found across conditions.

We measured time-to-collision (TTC) judgments from participants with age-related macular degeneration (AMD), and normal vision (NV) controls, with an audiovisual virtual reality system that simulated vehicles approaching in a 3D traffic environment. The vehicle was presented visually only, aurally only, or both simultaneously, allowing us to determine the relative importance of visual and auditory cues with psychophysical reverse correlation. Results indicated that TTC judgments were based on both auditory and visual cues in the AMD and NV groups; the AMD group relied, at least in part, on their residual vision. A multimodal advantage was not observed in either group. TTC estimation in the AMD group was surprisingly similar to that in the NV group. However, the AMD group showed a higher relative importance of "heuristic" cues compared to more reliably accurate cues favored by the NV group, suggesting that similar performance may be achieved through different cue-weighting strategies.

Patient safety event (PSE) reports, which document incidents that compromise patient safety, are fundamental for improving healthcare quality. Accurate classification of these reports is crucial for analyzing trends, guiding interventions, and supporting organizational learning. However, this process is labor-intensive due to the high volume and complex taxonomy of reports. Previous work has shown that machine learning (ML) can automate PSE report classification; however, its success depends on large manually-labeled datasets. This study leverages Active Learning (AL) strategies with human expertise to streamline PSE-report labeling. We utilize pool-based AL sampling to selectively query reports for human annotation, developing a robust dataset for training ML classifiers. Our experiments demonstrate that AL significantly outperforms random sampling in accuracy across various text representations, reducing the need for labeled samples by 24% to 69%. Based on these findings, we suggest that incorporating AL strategies into PSE-report labeling can effectively reduce manual workload while maintaining high classification accuracy.

Trust and system reliability can influence a user's dependence on automated systems. This study aimed to investigate how increases and decreases in automation reliability affect users' trust in these systems and how these changes in trust are associated with users' dependence on the system. Participants completed a color identification task with the help of an automated aid, where the reliability of this aid either increased from 50% to 100% or decreased from 100% to 50% as the task progressed, depending on which group the participants were assigned to. Participants' trust, self-confidence, and dependence on the system were measured throughout the experiment. There were no differences in trust between the two groups throughout the experiment; however, participants' dependence behavior did follow system reliability. These findings highlight that trust is not always correlated with system reliability, and that although trust can often influence dependence, it does not always determine it.

Autistic individuals face challenges in successful employment, emphasizing the need for targeted workplace support. This study explored collaborative dynamics within neurodiverse teams during a simulated remote work task by applying Hidden Markov Models (HMMs) to heart rate data. Eighteen participants formed nine dyads: six nonautistic (NA-NA) pairs and three autistic-non-autistic (ASD-NA) pairs. Dyads completed two trials of a collaborative programming task over Zoom, alternating roles between trials. Heart rate data were collected, segmented, and transformed to extract features reflecting participants' interactions. The final HMM was fitted with seven hidden states, and transition probabilities were derived for each dyad type. Results showed that NA-NA dyads exhibited more frequent transitions among states compared to ASD-NA dyads, potentially suggesting more varied interaction patterns. These findings demonstrate the utility of HMMs in capturing collaborative behaviors through physiological signals and highlight their potential in helping develop effective support strategies for neurodiverse teams.

To support the ongoing adaptation and implementation of an Emergency Department (ED)-based clinical decision support (CDS) tool to prevent future falls, we interviewed older adults (n=15) during their ED stay. We elicited their feedback on the written and verbal content of the existing CDS, feelings about the automated risk-screening aspect of the CDS and asked them to identify barriers that would prevent them from following up with the Falls Clinic to which the CDS supports referral placements. Our findings suggest that the older adults interviewed saw the CDS simply as another tool that they trusted their ED physician/APP to interact with. The identified barriers to follow-up reflect common access barriers such as transportation availability and clinic distance. For CDS tools to impact real-life patient outcomes, we must consider patient's needs and limitations and appropriately match interventions.

The transition period from automation to manual, known as the takeover process, presents challenges for drivers due to the deficiency in collecting requisite contextual information. The current study collected drivers' eye movement in a simulated takeover experiment, and their Situation Awareness (SA) was assessed using the Situation Awareness Global Assessment Technique (SAGAT) method. The drivers' Stationary Gaze Entropy (SGE) was calculated based on the percentages of time they spent on six pre-defined Areas of Interests (AOIs). Three critical time windows were extracted by using the takeover alert time spot and the hazard perceived time spot. The result indicated that drivers with higher SAGAT scores would spread their attention among multiple AOIs. Also, drivers' SGE and SA have a linear relationship only at the last time window (hazard perceived to the end) wherein SGE potentially functions as an evaluative metric for assessing SA in the future.

Employment is an important aspect of independent adulthood, yet autistic adults typically face substantial barriers in the labor market, including high rates of un- and under-employment. To promote an inclusive workplace, the present study explored collaboration dynamics between autistic and non-autistic adults as they worked toward shared team goals in an online setting. We recruited nine dyads, including three dyads of non-autistic adults with an autistic adult (NA-AA), and six dyads of non-autistic adults (NA-NA). Our findings demonstrated that neurodiverse collaboration (autistic and non-autistic adults together) could lead to improved task efficiency at the group level and higher perceived team performance in individuals. However, in these collaborative settings, autistic adults reported higher levels of depression, anxiety, and stress compared to their non-autistic partners. Our findings demonstrate the unique contributions that autistic adults may bring into the workplace and highlight the need to develop workplace technologies supporting their collaborative experiences.