Human-automation Interaction Research Articles

Mathematical Modeling of the Dynamics of Trust in Automation

Trust in automation (TiA) is crucial for human-automation interactions, but traditionally measured statically via discrete survey data points. We developed a mathematical model capturing the dynamic nature of TiA, incorporating measurable components across different timescales, and validated it through experimental data. Our findings show that: (1) trust timescales are important for predicting automation usage, emphasizing trust’s key role in decision-making; (2) short-timescale trust significantly impacts overall trust dynamics, highlighting the importance of recent interactions; (3) trust levels are generally higher in highly reliable automation scenarios, though this pattern emerges after an initial evaluation period; and (4) automation type, implying varying workloads, does not significantly impact trust dynamics in highly reliable systems but fosters increased trust in lower-reliability systems. This study advances our understanding of trust dynamics in automation, contributing to more intuitive and trustworthy technology development.

Transparency improves the accuracy of automation use, but automation confidence information does not.

Increased automation transparency can improve the accuracy of automation use but can lead to increased bias towards agreeing with advice. Information about the automation's confidence in its advice may also increase the predictability of automation errors. We examined the effects of providing automation transparency, automation confidence information, and their potential interacting effect on the accuracy of automation use and other outcomes. An uninhabited vehicle (UV) management task was completed where participants selected the optimal UV to complete missions. Low or high automation transparency was provided, and participants agreed/disagreed with automated advice on each mission. We manipulated between participants whether automated advice was accompanied by confidence information. This information indicated on each trial whether automation was "somewhat" or "highly" confident in its advice. Higher transparency improved the accuracy of automation use, led to faster decisions, lower perceived workload, and increased trust and perceived usability. Providing participant automation confidence information, as compared with not, did not have an overall impact on any outcome variable and did not interact with transparency. Despite no benefit, participants who were provided confidence information did use it. For trials where lower compared to higher confidence information was presented, hit rates decreased, correct rejection rates increased, decision times slowed, and perceived workload increased, all suggestive of decreased reliance on automated advice. Such trial-by-trial shifts in automation use bias and other outcomes were not moderated by transparency. These findings can potentially inform the design of automated decision-support systems that are more understandable by humans in order to optimise human-automation interaction.

Ironies of automation and their implications for public service automation

Automation of public service provision has gained renewed attention as emerging technologies are said to enable automation of tasks that were previously seen as requiring human involvement. However, the merits of these automation technologies are often exaggerated. More knowledge is needed on public service automation, and much can be learned from adjacent research fields studying human-automation interaction. To lead by example, this work applies Bainbridge's (1983) concept of ironies of automation. The purpose is to (1) present ironies of automation, (2) explicate how these ironies can come into play when implementing automated systems in the public service context, and (3) outline implications that follow for public service automation. This is achieved by relating ironies of automation to contemporary studies on Robotic Process Automation (RPA) developments in Swedish local government. The analysis results in five ironies and a set of implications for public service automation. The ironies and implications for public service automation direct attention to key challenges that must be acknowledged in future automation implementations and show that further investigations and theoretical developments are needed on e.g., problems introduced by automation; tasks, roles, and responsibilities that follow on automation; how to design the interface between humans and automated systems in a way that facilitates monitoring, take-over, and maintenance; and, tools and methods for assessing the impact and quality of automated systems. This paper thus provides a foundation for future empirical investigations and further theoretical development on public service automation.

Work domain modeling of human-automation interaction for in-vehicle automation

Automated driving systems are deployed on public roads with little empirical support for the dominant justifications of enhanced safety and enhanced productivity. Furthermore, development of automated driving systems has been piecemeal rather than systematic while research on driver-automation interaction has relied on individual analysis of accidents and on observational studies of driving behavior in a simulator or on the road. In this paper, we apply Work Domain Analysis to develop a more systematic and comprehensive model of automated driving. We use a strategy of layering the driving automation onto the resulting ion-Decomposition Space for manual driving to mimic the existing design strategy of introducing automation to take over driving functions previously the responsibility of the human driver. Our analysis shows that automation does not unequivocally supports dominant driving values. Furthermore, our analysis revealed subtle interdependencies between human and technological functions. We conclude that an ion Decomposition Space offers a systematic view of driver-automation interaction that can suggest new insights for automation design.

Analyzing Riders’ Behavioral Adaptation to Driving Patterns of Advanced Autonomous Vehicles: A Virtual Reality Simulation Study

The necessity of human supervision and intervention during autonomous driving has long been a topic of controversial discussion. From a developer’s perspective, it is expected that users will readily adapt to well-calibrated autonomous driving systems (ADS) due to their superior performance in dynamic driving tasks (DDT) compared to conventional human-driven vehicles. However, when passengers experience an autonomous vehicle (AV), there may be an adjustment period during which they modify their behavior to accommodate the driving patterns of the ADS. Additionally, some passengers might not adapt to autonomous driving at all, highlighting potential limitations in the current ADS development strategy. This work studies the dynamics of human-automation interaction and introduces an “objective method”, which employs a Virtual Reality (VR)-enabled simulation approach for in-depth behavioral analysis concerning riders’ behavioral adaptation to autonomous driving. Specifically, we examined how participants interacted with and intervened in Level 4 ADS operating under conservative, moderate, and aggressive driving patterns in a fully autonomous environment. A realistic urban road network was recreated in VR, integrated with traffic microsimulation to generate various driving scenarios. Twenty-seven participants completed driving tasks across different AV modes, with their intervention behaviors analyzed in relation to traffic conditions and AV aggressiveness. Key findings include: (1) Participants showed higher intention to intervene but lower actual intervention rates under aggressive AV modes compared to moderate and conservative modes, suggesting quicker adaptation to more challenging driving scenarios. (2) Interventions generally proved unnecessary and sometimes detrimental to overall traffic performance in a full-AV environment. (3) Aggressive AV modes significantly improved traffic efficiency, with a 40% increase in average travel speed and a 53% reduction in waiting time. However, human interventions posed the greatest challenge to achieving optimal traffic conditions. This research provides insights into the complex dynamics of human-AV interaction and adaptation, offering valuable implications for AV interface design, implementation strategies, and public acceptance of autonomous driving technologies.

Judgments of Difficulty (JODs) While Observing an Automated System Support the Media Equation and Unique Agent Hypotheses.

We investigated how people used cues to make Judgments of Difficulty (JODs) while observing automation perform a task and when performing this task themselves. Task difficulty is a factor affecting trust in automation; however, no research has explored how individuals make JODs when watching automation or whether these judgments are similar to or different from those made while watching humans. Lastly, it is unclear how cue use when observing automation differs as a function of experience. The study involved a visual search task. Some participants performed the task first, then watched automation complete it. Others watched and then performed, and a third group alternated between performing and watching. After each trial, participants made a JOD by indicating if the task was easier or harder than before. Task difficulty randomly changed every five trials. A Bayesian regression suggested that cue use is similar to and different from cue use while observing humans. For central cues, support for the UAH was bounded by experience: those who performed the task first underweighted central cues when making JODs, relative to their counterparts in a previous study involving humans. For peripheral cues, support for the MEH was unequivocal and participants weighted cues similarly across observation sources. People weighted cues similar to and different from when they watched automation perform a task relative to when they watched humans, supporting the Media Equation and Unique Agent Hypotheses. This study adds to a growing understanding of judgments in human-human and human-automation interactions.

A robust operators’ cognitive workload recognition method based on denoising masked autoencoder

Identifying the cognitive workload of operators is crucial in complex human-automation collaboration systems. An excessive workload can lead to fatigue or accidents, while an insufficient workload may diminish situational awareness and efficiency. However, existing supervised learning-based methods for workload recognition are ineffective when dealing with imperfect input data, such as missing or noisy data, which is not practical in real applications. This study introduces a robust Electroencephalogram (EEG)-enabled cognitive workload recognition model using self-supervised learning. The proposed method, DMAEEG, combines the training strategies of denoising autoencoders and masked autoencoders, demonstrating strong robustness against noisy and incomplete data. More specifically, we adopt the temporal convolutional network and multi-head self-attention mechanisms as the backbone, effectively capturing both the temporal and spatial features from EEG. Extensive experiments are conducted to verify the effectiveness and robustness of the proposed method on an open dataset and a self-collected dataset. The results indicate that DMAEEG performs superior to other state-of-the-art across various evaluation metrics. Moreover, DMAEEG maintains high accuracy in workload inference even when EEG signals are corrupted with a high masking ratio or strong noises. This signifies its superiority in capturing robust intrinsic patterns from imperfect EEG data. The proposed method significantly contributes to decoding EEG signals for workload recognition in real-world applications, thereby enhancing the safety and reliability of human-automation interactions.

Benchmarking automation-aided performance in a forensic face matching task

Carragher and Hancock (2023) investigated how individuals performed in a one-to-one face matching task when assisted by an Automated Facial Recognition System (AFRS). Across five pre-registered experiments they found evidence of suboptimal aided performance, with AFRS-assisted individuals consistently failing to reach the level of performance the AFRS achieved alone. The current study reanalyses these data (Carragher and Hancock, 2023), to benchmark automation-aided performance against a series of statistical models of collaborative decision making, spanning a range of efficiency levels. Analyses using a Bayesian hierarchical signal detection model revealed that collaborative performance was highly inefficient, falling closest to the most suboptimal models of automation dependence tested. This pattern of results generalises previous reports of suboptimal human-automation interaction across a range of visual search, target detection, sensory discrimination, and numeric estimation decision-making tasks. The current study is the first to provide benchmarks of automation-aided performance in the one-to-one face matching task.

Will we as Passengers Use Highly Automated Vehicles? Examining the Importance of Role Adaptation on People’s Acceptance of the Automation

With increasing driving automation, driving tasks are shifting from drivers to automation systems and their roles are gradually changing from drivers to passengers. While efforts have been made to explore influencing factors of highly automated vehicles (HAVs) usage intention, few studies have linked role adaptation (RA) to variables from existing acceptance models. To fulfill this research gap, a HAV acceptance model was established based on RA, and other factors (i.e., situational trust, anxiety, and perceived usefulness [PU]). The proposed model validity was verified by subjective ratings collected from a driving simulator experiment involving 105 participants each of whom rode vehicles with three different automated driving styles. Results revealed that RA was an important factor in forming HAVs acceptance during initial human-automation interaction stages and could be increased by improving situational trust, PU, or reducing anxiety. Practically, these results can provide valuable guidance for enhancing consumers’ HAV acceptances.

Effect of interface design on cognitive workload in unmanned aerial vehicle control

Unmanned Aerial Vehicle (UAV) control interfaces are critical channels for transferring information between the vehicle and an operator. Research on system performance has focused on enhancing vehicle automation and some work has evaluated cognitive workload for existing UAV interfaces. The potential for usable interface design to reduce cognitive workload during the early design phase has been largely overlooked. This study addresses these gaps by: (1) evaluating the effectiveness of a contemporary UAV interface design tool (the Modified GEDIS-UAV) to moderate user workload; (2) examining the effectiveness of various UAV interface designs for minimizing cognitive workload under different control task pacing; and (3) exploring the use of eye tracking measures, traditionally applied in other domains, as indicators of cognitive workload in UAV operations. We prototyped three different interface designs, classified as “baseline”, “enhanced” and “degraded” interfaces. Cognitive workload in UAV operation was manipulated in terms of levels of vehicle speed (“low” and “high”). Physiological and subjective measures of workload were collected for all combinations of interface design and task demand. Results revealed the “enhanced” interface to yield the lowest operator cognitive workload and supported operator resilience to increased control task demand, as compared to the “baseline” and “degraded” interfaces. In addition, task demand was found to elevate operator cognitive workload, particularly in terms of "mental" and "temporal" demands and operator perceptions of "performance". The study also demonstrated utility of eye-tracking technology for detecting cognitive workload in UAV operations. This research provides practical guidance for UAV control interface design to manage operator workload. The methods employed in the study are applicable to interface evaluation for various types of UAVs and other unmanned systems to enhance human-automation interaction.

A Regulatory Perspective: Have We Done Enough on Grasping Automation Failure?

This paper responds to Skraaning and Jamieson’s target paper “The Failure to Grasp Automation Failure.” We acknowledge that the target paper made important contributions to automation research in the human factors community. It analyzed automation failure events in complex operational systems in contrast to the vast majority of laboratory research on human-automation interaction. The paper presented a taxonomy of automation failure. The analysis and taxonomy demonstrate the integration of approaches to grasping automation failures from system instrumentation and controls, human factors engineering, and human reliability analysis. We reviewed the regulatory framework related to use of automation in nuclear power plants and examined whether the framework elements adequately address “Failure to Grasp Automation Failure” using the taxonomy in the target paper. Overall, we believe that the target paper could enhance the consideration for potential automation failures in the design and regulatory review process of automation technologies.

Framing Automation and Human Error in the Context of the Skill, Rule and Knowledge Taxonomy

Automation errors may result in human performance issues that are often difficult to grasp. Skraaning and Jamieson (2023) proposed a taxonomy for classifying automation errors into categories based on the visible symptoms of design problems, so as to benefit the design of training scenarios. In this paper, we propose a complementary classification that is based on the mechanisms of human-automation interaction guided by Rasmussen’s Skill, Rule and Knowledge (SRK) taxonomy. We identified four main failure classes and expect that this classification can support automation designers.

The Designers’ Perspective on Autonomous Mining Systems and Sociotechnology

Digitalization and automation technology offer new possibilities to increase productivity and obtain higher levels of autonomy in mining operations. Introducing autonomous systems into mining is not only a technical problem in terms of effectiveness and efficiency, nor a problem of safety in human-automation interactions. The systems also need to be designed and developed so that they foster healthy and attractive working environments. The design and development phase of new mining technology has not been extensively studied previously. To fill this knowledge gap, we investigated technology developers’ basic assumptions about humans and their interactions with the technology they develop. We conducted five semi-structured workshops within an EU funded project concerned with developing digitalization and automation solutions for the mining industry. The data suggests that many critical functions will still be under human control in future mining systems. The results also indicate increased complexity in the interaction between autonomous systems and humans as the technology becomes more advanced. As a result, we suggest that a human perspective, based on sociotechnical principles, should not only be considered in implementing the technology at mines but also in the early conceptual phases of developing and designing the technology. This will ensure healthy and attractive work environments in the future mining industry.

Trust, risk perception, and intention to use autonomous vehicles: an interdisciplinary bibliometric review

AbstractAutonomous vehicles (AV) offer promising benefits to society in terms of safety, environmental impact and increased mobility. However, acute challenges persist with any novel technology, inlcuding the perceived risks and trust underlying public acceptance. While research examining the current state of AV public perceptions and future challenges related to both societal and individual barriers to trust and risk perceptions is emerging, it is highly fragmented across disciplines. To address this research gap, by using the Web of Science database, our study undertakes a bibliometric and performance analysis to identify the conceptual and intellectual structures of trust and risk narratives within the AV research field by investigating engineering, social sciences, marketing, and business and infrastructure domains to offer an interdisciplinary approach. Our analysis provides an overview of the key research area across the search categories of ‘trust’ and ‘risk’. Our results show three main clusters with regard to trust and risk, namely, behavioural aspects of AV interaction; uptake and acceptance; and modelling human–automation interaction. The synthesis of the literature allows a better understanding of the public perception of AV and its historical conception and development. It further offers a robust model of public perception in AV, outlining the key themes found in the literature and, in turn, offers critical directions for future research.

Use of Human-Automation Taxonomies for System Modeling

Skraaning and Jamieson indicate the importance of developing a definition of automation failure in order to guide development of human performance models to predict human-automation interactions. They go on to develop a taxonomy including four categories: Elementary automation failures; Systemic automation failures; Human-automation breakdowns; and Human and organizational slips and misconceptions. They suggest that the latter category be excluded from classes of automation failures. They then proceed to identify important factors within each category. This paper presents a complementary taxonomy designed to identify factors contributing to automation-related failures in order to develop probes to elicit predictions of failure stories for some proposed new system design from a range of experts. While there is overlap between these two taxonomies, there are important differences. It is suggested that they could be integrated or used jointly both for knowledge elicitation and to identify requirements regarding needed capabilities of predictive performance models.

Exploring remote operation of heavy vehicles—findings from a simulator study

AbstractFully automated driving has posed more challenges than expected, and remote operation of heavy vehicles is increasingly getting attention. Therefore, human remote operators may have an essential role in compensating for the technological shortcomings in vehicle automation. This poses challenges in designing the work of human remote operators of automated heavy vehicles. This paper present findings from a research project performed in collaboration between the RISE Research Institutes of Sweden and Scania. In the project, human-automation interaction requirements and challenges for remote operator work were explored through a simulator study. Before the study, three main operator tasks were defined: assessment, assistance, and remote driving. The simulation occurred in a transportation scenario where operators handled ten trucks driving on a public road and in confined areas (transportation hub). Fifteen participants completed the study. The results provide examples and insights into classical automation-related challenges in a new context—the remote operation of heavy vehicles. Instances of challenges with situational awareness, out-of-the-loop, trust, and attention management were found and are discussed in relation to HMI design and requirements. In addition, it was found that transitions between relatively passive monitoring and more active assistance and driving were performed more fluently than expected. In general, supervisory control of ten vehicles in parallel was seen as a feasible task given the conditions in the simulated environment.

How Are Automation Failures Characterized in the Driving Domain? Insights From a Scoping Review

We agree with Skraaning and Jamieson’s assertion that the “failure” construct is not always clearly defined in human-automation interaction research. We applied their proposed taxonomy to explore how failures have been characterized in driving automation research based on a recent scoping review we conducted. We discuss the insights gained and the challenges of using the taxonomy to characterize driving automation failures: (1) Utilizing the taxonomy confirmed that driving automation research is limited in failure scenarios tested. (2) Applying the taxonomy to empirical studies on driving automation is challenging due to limited information on underlying failure mechanisms. (3) Failures can be difficult to classify due to the complexity of technology and environmental factors. (4) Researchers and designers should know failure mechanisms, but drivers will not.

Does it do the same as we would? How trust in automated shipboard systems relates to seafarers’ professional identity

In this qualitative study, we explore the concept of trust in safety-critical automated shipboard systems and how it relates to the professional identity of seafarers. For maritime safety, it is critical that human–automation interaction builds on appropriate trust in an automated system. Although there is momentum in researching trust in automation, few studies have addressed this topic using qualitative methods and sociocultural approaches. This study builds on rich data from interviews and observations of maritime officers on battery-electric domestic car ferries with state-of-the-art automation in Norway. The results show how seafarers progressed from initial skepticism to trust and sometimes overreliance on the systems. The development of trust was contingent on systems behaving according to professional standards rooted in professional identity, such as sailing according to good seamanship. This comparison was enabled by individual hands-on experiences. Professional identity clarifies the context-specific antecedents of trust exhibited by seafarers when core tasks are automated. Understanding and, to some degree, aligning systems with professional identity could help achieve trustworthy systems with appropriate trust exhibited by users. The safety opportunities and challenges created by the interplay between trust and professional identity are discussed.

Exploring the effects of automation malfunction on team communication and coordination in ships' engine rooms

AbstractAutomation malfunctions within complex socio‐technical systems reserve the potential to significantly affect human performance. In the context of maritime operations, varying consequences of automation malfunction on human performance can be observed. This study introduced a two‐step research framework to examine the repercussions of such malfunctions, particularly those related to communication and coordination among human teams in ship engine rooms. Initially, a qualitative semi‐structured interview was conducted with seven professional marine engineers to explore the potential impact of hypothetical automation malfunction on team communication. Subsequently, a quantitative survey involving 32 professional marine engineers employed coordination demand analysis (CDA) to scrutinize changes in team coordination resulting from malfunction. The findings indicate that an automation malfunction within an engine room can precipitate an abrupt overload of the socio‐technical system. This can significantly increase communication frequency among engineers, particularly in relation to the physical and organizational aspects of the environment. Furthermore, the study highlights the influence of disparate levels of expertise among team members on coordination demands. A positive correlation was discovered between differences in expertise and increased coordination demands within a team. These insights underscore the necessity for future research on human–automation interaction, specifically focusing on individual differences and nontechnical skills.

Operator Role Classification in Human-Automation Interaction: A Systematic Review

Technical advancements, such as autonomous systems, prompt a reevaluation of operator roles within human-machine systems in manufacturing. This paper presents a systematic review of 370 records concerning applications in relevant application areas. It offers classifications of activities within human-machine systems, providing a nuanced exploration of operator roles and interaction dynamics. Through this analysis, three distinct perspectives emerge, focusing on agent dependency, information processing, and functional roles. The results highlight the importance of clear role descriptions and the current lack of standardized approaches and offering valuable insights for future research and practical applications.