Search

Tasks classically performed by human–human teams in today’s workplaces are increasingly given to human–technology teams instead. The role of technology is not only played by classic decision support systems (DSSs) but more and more by artificial intelligence (AI). Reliability is a key factor influencing trust in technology. Therefore, we investigated the reliability participants require in order to perceive the support agents (human, AI, and DSS) as “highly reliable.” We then examined how trust differed between these highly reliable agents. Whilst there is a range of research identifying trust as an important determinant in human–DSS interaction, the question is whether these findings are also applicable to the interaction between humans and AI. To study these issues, we conducted an experiment (N= 300) with two different tasks, usually performed by dyadic teams (loan assignment and x-ray screening), from two different perspectives (i.e., working together or being evaluated by the agent). In contrast to our hypotheses, the required reliability if working together was equal regardless of the agent. Nevertheless, participants trusted the human more than an AI or DSS. They also required that AI be more reliable than a human when used to evaluate themselves, thus illustrating the importance of changing perspective.

In numerous applications, alarm systems play an important role, supporting human decision-making. So far, however, little research dealt with the cognitive mechanisms that are at play in alarm-supported decision-making. In the present study, we aim to disentangle underlying cognitive mechanisms by using drift diffusion modeling. The results showed that going beyond standard approaches of analyzing alarm-system supported binary decision tasks can reveal results unlikely to be captured otherwise. That is, the analyses revealed that the alarm system’s output biased the decision-making process, requiring less evidence to be sampled for agreeing with the system than for disagreeing with the system. Moreover, evidence was accumulated faster on correct than on incorrect alarm system recommendations. Thus, the present results point to promising directions for gaining a more fine-grained picture of automation supported decision making.

In various domains, humans are supported by automated systems. Earlier research has suggested that trust in automated agents differs from trust in other humans. The present studies aimed at taking a multi-dimensional look at effects on trust towards automation and humans. To this end, we conducted two studies to empirically validate a multi-dimensional trust questionnaire to assess performance, utility, purpose, and transparency subdimensions of trust (Study 1, N= 160) and to study experimental effects of support agent (i.e., human vs. decision support system) and failure experience (i.e., none vs. one; Study 2, N= 181). The expected factor structure was confirmed. Moreover, the results showed that being supported by a human mostly impacted the performance subscale. In sum, the findings illustrate the importance to study trust not only uni-dimensionally but to consider different subdimensions, particularly as a single-item trust measurement was mostly correlated to performance and utility subscales.

This article describes a flight simulator study to validate a new command control system for longitudinal load factor nx. The system is called nxControl and actuates thrust, speedbrakes, and wheel brakes that directly influence the total energy state of the aircraft. It aims at manually flying with higher precision and lower workload. After a brief overview of the functionality of nxControl and its cockpit interfaces, the key results of earlier proof-of-concept simulator studies are summarized. The focus lies on the comprehensive flight simulator study with 24 airline pilots to conclusively evaluate the complete system. The task was a highly demanding approach trajectory containing segmented-continuous-descent in gusty wind conditions, touch-down, deceleration, and taxi as well as engine failure at low altitude. With nxControl, the pilots achieved higher precision in airspeed and energy control with lower workload compared to conventional manual thrust control. In addition, nxControl achieved safety benefits in case of an engine failure.

The present study addresses effects of human redundancy on automation monitoring and cross-checking. Thirty-six participants performed a multi-task, consisting of three subtasks that mimic basic work demands of operators in a control room of a chemical plant. One of the tasks was to monitor and cross-check a highly reliable and safety-critical automated process. Participants were randomly assigned to two groups: (1) “Non-redundant”: participants worked on all tasks alone as the only responsible operator. (2) “Redundant”: participants were informed that a second crewmate would work in parallel on the automation monitoring task and that they both were responsible for ensuring safe operation of the automation. Results provide evidence for social loafing effects in automation cross-checking. Participants working redundantly with another crewmate were found to cross-check the automation significantly less than participants, who were working alone. Even if the combined team performance of the participants working in the redundant condition was considered, the number of cross-checks did not significantly differ from the performance in the non-redundant condition. This result suggests that human redundancy can induce social loafing effects which fully compensate a possible reliability gain intended to be achieved by this measure. It challenges the often stated assumption that “four eyes see more than two” and shows that human redundancy does not necessarily lead to enhanced safety in automation monitoring.

Abstract– The objective of this panel was to discuss issues related to the development and use of autonomous systems, with specific focus on the overriding themes of ethical considerations and potential liability for Human Factors and Ergonomics (HF/E) professionals who are involved in their development. Chris Brillprovided opening remarks to frame the discussion and introduce the panelists. James Blissdiscussed legal implications related to our collective penchant for developing conservative, false-alarm prone automation. Peter Hancockadvocated for human-centered constraints on autonomous systems, as they may, one day, pose an existential threat to humanity. Dietrich Manzeydiscussed ethical considerations for autonomous systems, including how design can encourage ethical user behavior. Joachim Meyerargued that HF/E professionals have an obligation to help designers understand the ethical implications of poor design, particularly in the context of autonomous systems. Lastly, Alison Vredenburghprovided thoughts on potential liability for HF/E professionals, particularly in light of the relative newness of autonomous systems. The panel then turned to facilitated discussion with panelists and audience members. Specific themes included the boundaries of our responsibilities as HF/E professionals for ill-conceived or morally-objectionable systems, potential implications of manipulating user trust through design, cross-cultural perspectives on public acceptance and legal peril, and how concerns might differ by domain (e.g., medical vs. combat vs. manufacturing). The session concluded with panelists summarizing how ethics influence design and recommendations for how HF/E professionals can potentially protect themselves from legal liability for mishaps involving autonomous systems they helped develop.

The goal of current work is to provide a literature review of the performance consequences of image-guided surgery (IGS) use. Regarding IGS system functionalities, the pattern of results suggests to distinguish between IGS systems which only provide information support (e.g. pointer-based systems) and those which directly intervene in surgeons' decision-making and actions (e.g. instrument disablement). The former offer benefits for patient safety and surgical outcome, as well as improved intraoperative orientation for surgeons, helping them to identify anatomical structures. Furthermore, IGS systems providing information support seem to shorten the time needed for surgery and reduce the subjective workload. IGS systems which intervene in a surgeon's decision-making and actions also have positive impact on patient safety and surgical outcome. In addition, these systems seem to reduce the physiological effort of surgeons. However, they can also prolong surgery and increase the subjective workload.

It is safety critical for pilots to be aware of the aircraft’s energy state in terms of proper altitude and airspeed. A loss of energy awareness is an important human factors issue in modern civil aircraft. In order to maintain the energy awareness and support the manual flying skills, several cockpit display concepts suggest to augment the current energy status of the aircraft on primary flight displays in terms of the total energy angle. An experiment was carried out to determine which effect this additional energy information has on pilots’ flight path control, instrument scanning, and situation awareness. Outcomes of the study show a significant shift of the scanning pattern from airspeed, altitude scale, and engine parameter towards the center of the primary flight display with unchanged situation awareness. In addition, pilots are better able to maintain given speed targets.

Providing operators additional information helping them to validate alarms has been found to be a countermeasure for problems related to the cry wolf effect (i.e., operators ignoring alarms). Adding information can be realized with likelihood alarm systems (LAS) or with access to alarm validity information (AVI). The two studies presented here examined behavior and performance consequences of the combination of LAS and AVI in multi-task settings. It was investigated to what extent concurrent task performance and alert task performance depend on characteristics of the LAS (i.e. proportion of different alert types) and cost of cross-checking AVI. Results suggest that those LAS characteristics varied here do not influence participants’ performance. Secondly, no benefit of LAS over binary alarm systems (BAS) emerged when increasing the cost of accessing AVI. Results are further discussed with regard to participants’ response patterns.

Spatial disorientation (SD) is a common factor in aviation accidents, especially in novice pilots. An experiment was carried out to determine which of four different attitude indicator concepts in combination with two different display backgrounds (abstract vs. synthetic landscape) proves to be the most beneficial for novice pilot performance. Inexperienced pilots had to recover from unusual attitudes by using the standard moving-horizon display, a moving-aircraft display, a frequency-separated display, and a “mixed” display, with the latter two representing hybrid concepts with movements of both aircraft symbol and horizon bar. Participants performed the task of recovering from unusual attitudes most efficiently with hybrid display concepts, suggesting that these display concepts prevent figure-ground reversals and associated pilot errors. Outcomes of the study suggest that the implementation of hybrid display concepts as a backup option when unwillingly entering Instrument Flight Conditions could be a solution for preventing SD in novice pilots.

Recent research has shown that the use of 3-stage likelihood alarm systems (LAS) has the potential to mitigate performance deficits associated with the use of binary alarm systems (BAS). The additional likelihood information can guide operators’ behavior and improve their decision-making accuracy. Comparisons of LAS with different numbers of stages are missing so far. Therefore, the current study compared a BAS with a 3-stage LAS and a 4-stage LAS. Participants were found to make significantly fewer wrong decisions with the 4-stage LAS than with the other two systems, and still significantly fewer errors with the 3-stage LAS compared to the BAS. We found that this performance benefit resulted from a reduced number of false alarms, whereas no difference was found with regard to misses. Results are further discussed with regard to their theoretical implications for LAS and threshold setting in BAS.

Providing access towards raw data is often considered to be a good solution for improving human decision making in interaction with imperfect automated decision support such as alarm systems. However, there is some evidence that such cross-checking measures are used in an asymmetric manner with respect to the amount of uncertainty involved in the decision. Namely, people seem to accept low amounts of uncertainty when complying with an alarm cue, but not when contradicting it. The current study investigates the question whether this phenomenon is limited to alarm systems and a high risk environment. Within a multi-task PC simulation participants performed a low risk monitoring task which was supported by a system neutrally framed as “assistant system”. In one group the cues emitted by the system were 90% correct, in the other 10% were correct, thus causing a 10% uncertainty about the real state in both conditions. Results show a strong asymmetry as participants in the latter condition spent a high amount of effort in reducing their uncertainty, while participants in the former condition did not. Furthermore participants’ behavior almost exactly replicates the asymmetric cross-checking pattern found in a former study which employed a comparatively high risk monitoring task supported by an “alarm system”. This supports the hypothesis that the observed commission bias represents a general phenomenon in the context of automated decision support, irrespective of the risk attributed to the environment and irrespective of whether the system represents an alarm system or not.

The present study addresses effects of human redundancy on automation monitoring performance. Fourty-six participants performed a multi-task, consisting of three sub-tasks which simulate basic demands of operators in a chemical plant. One of the tasks involved the monitoring of an automated process. Participants were randomly assigned to three groups: (1) “Non-Redundant”: participants worked on all tasks alone. (2) “Redundant”: participants were informed that a second crewmate would work in parallel on the monitoring task. (3) “Redundant-Feedback”: like the “redundant” condition with the additional information that crewmembers’ individual monitoring performance would be tracked and fed back. Results provide evidence of social loafing effects in monitoring performance. Participants in the “redundant” condition cross-checked the automation significantly less than participants in the other groups. Moreover they were more prone to miss automation failures which occurred surprisingly. The anticipation that individual performance will be tracked and fed back after the task in the “redundant-feedback” group reduced this effect. The results suggest that human redundancy does not necessarily constitute an effective measure for enhancing reliability of automation monitoring and that expected positive effects can at least partially be off-set by a sort of social-loafing effect.

Operators in complex environments are supported by alarm-systems that indicate when to shift attention to certain tasks. As alarms are not perfectly reliable, operators have to select appropriate strategies of attention allocation in order to compensate for unreliability and maintain overall performance. This study investigates how humans adapt to differing alarm-reliabilities. Within a multi-tasking flight simulation, participants were randomly assigned to four alarm-reliability conditions (68.75%, 75%, 87.5%, 93.75%), and a manual control group. In experimental conditions, one out of three subtasks was supported by an alarm-system. Compared to manual control, all experimental groups benefited from alarms in the supported task, with best results for the highest reliability condition. However, analyses of performance and eye-tracking data revealed that the benefit of the lowest reliability group was associated with an increased attentional effort, a more demanding attention allocation strategy, and a declined relative performance in a non-supported task. Results are discussed in the context of recent research.

Two experiments are reported that investigate to what extent performance consequences of automated aids are dependent on the distribution of functions between human and automation and on the experience an operator has with an aid. In the first experiment, performance consequences of three automated aids for the support of a supervisory control task were compared. Aids differed in degree of automation (DOA). Compared with a manual control condition, primary and secondary task performance improved and subjective workload decreased with automation support, with effects dependent on DOA. Performance costs include return-to-manual performance issues that emerged for the most highly automated aid and effects of complacency and automation bias, respectively, which emerged independent of DOA. The second experiment specifically addresses how automation bias develops over time and how this development is affected by prior experience with the system. Results show that automation failures entail stronger effects than positive experience (reliably working aid). Furthermore, results suggest that commission errors in interaction with automated aids can depend on three sorts of automation bias effects: (a) withdrawal of attention in terms of incomplete cross-checking of information, (b) active discounting of contradictory system information, and (c) inattentive processing of contradictory information analog to a “looking-but-not-seeing” effect.

The study investigates how complacency and automation bias effects in interaction with automated aids are moderated by system experience. Participants performed a supervisory control task supported by an aid for fault identification and management. Groups differed with respect to how long they worked with the aid until eventually an automation failure occurred, and whether this failure was the first or second one the participants were exposed to. Results show that negative experiences, i.e., automation failures, entail stronger effects on subjective trust in automation as well as the level of complacency and automation bias than positive experiences (correct recommendations of the aid). Furthermore, results suggest that commission errors may be due to three different sorts of effects: (1) a withdrawal of attention in terms of incomplete cross-checks of information, (2) an active discounting of contradictory system information, and (3) an inattentive processing of contradictory information analogue to a “looking-but-not-seeing” effect.

The present study investigates performance consequences of a new approach of automated support for surgeons. “Navigated-Control” (NC) represents an advancement of image-guided navigation that does not only support the surgeon in navigating through a patient's anatomy, but also can stop the surgical device if it comes too close to risk structures which need to be protected to ensure patient safety. It is explored, how NC affects different aspects of surgical outcome, workload and stress, and situation awareness. Fourteen advanced students of medicine performed a simulated Mastoidectomy with and without NC support. The results reveal that NC support can reduce both, the risk of intra-operative injuries, as well as the physiological stress level of surgeons. However, “cost effects” emerged with respect to subjective workload, and a reduced spare capacity compared to unsupported surgeries. These latter effects do not seem to be related to the principle of NC but technical constraints of current implementations.

Most theories about operators' responses to alarm systems suggest that the operators' behavior is guided by their trust towards the system which in turn results from the subjective perception of system properties, namely the perceived reliability of the alarm system. However, some doubts about that assumption have arisen as recent research has not proven the mediating effect of trust. The purpose of this research was to examine the relationship between alarm system properties, trust, and behavior. The alarm reliability was varied while keeping the other system properties constant. It was found that participants' response-rates to alarms were predicted by their perceived alarm reliabilities. However, no mediation by trust could be established. These results suggest that operators' behavior is not always guided by their trust towards the system. Under specific circumstances their compliance rather depends on rational consideration regarding the most efficient strategy.

The purpose of this panel was to provide a general overview and discussion of some of the most current and controversial concepts and trends in human-automation interaction. The panel was composed of eight researchers and practitioners. The panelists are well-known experts in the area and offered differing views on a variety of different human-automation topics. The range of concepts and trends discussed in this panel include: general taxonomies regarding stages and levels of automation and function allocation, individualized adaptive automation, automation-induced complacency, economic rationality and the use of automation, the potential utility of false alarms, the influence of different types of false alarms on trust and reliance, and a system-wide theory of trust in multiple automated aids.

Image-guided navigation (IGN) systems support the surgeon in navigating through the patients' anatomy. Previous research on IGN has focused on technical feasibility and clinical applications. Yet, as the introduction of IGN corresponds to a partial automation of the surgeon's task, well known issues of human-automation interaction might play a crucial role for the success of IGN as well. The present study represents a first attempt to assess the impact of IGN on four key issues of human automation-interaction, i.e., workload, situation awareness, trust, and skill degradation, from the surgeons' perspective. A nation-wide survey among 213 German surgeons from 94 different hospitals was conducted. Results revealed (1) a workload-shift due to IGN rather than a reduction of workload, (2) benefits of IGN with respect to situation awareness, (3) comparatively high levels of perceived reliability, trust and reliance, and (4) skill degradation as a possible risk, albeit only for inexperienced surgeons.

The present study investigates how human performance consequences of automated decision aids are moderated by the operator's performance state and the aid's level of automation. Participants performed a simulated supervisory control task with one of two decision aids which provided different degrees of support for fault identification and management. One session took place during the day, another one during the night, after a prolonged waking phase of more than 20 hours. Results show that both, primary and secondary task performance benefit from automated support compared to manual performance. During the night, participants supported by the higher automated aid were better able to maintain a high level of performance. Clear evidence for automation bias was found, but only during the day session. Automation verification was performed more carefully during the night, indicating a less complacent behavior when operators used the decision aids in a state of sleepiness and fatigue.

The effects of misses of an automated alarm and fault diagnosis system on different manifestations of automation misuse were examined. 24 participants operated a complex multi-task process control simulation. During training, they either experienced automation misses or were only informed that failures might occur. The experience of misses reduced complacency towards the alarm function of the decision aid as well as omission errors but did neither affect complacency towards the aid's diagnostic function nor commission errors. Implications of this specific effect of automation misses for the design of training measures as well as the theoretical understanding of automation misuse are discussed.

The present study investigates effects of automated decision aids in terms of expected performance benefits of automation support and possible negative performance consequences. The negative consequences include the possible effects of automation bias and performance decrements when returning to manual performance in case of automation failure. Three automated aids that support fault identification and management are compared in a simulated supervisory control task. Results show that primary and secondary task performance improved with automated support compared to manual performance, with effects directly dependent on the level of automation (LOA). Effects of automation bias emerged independent of LOA. For return-to-manual performance, weak indications of an automation-induced skill loss were only found in the highest LOA. These arose in manual fault management, but not in fault identification performance.

The present study addresses effects of social loafing and social compensation in automation monitoring. Thirty-six participants performed a multi-task, consisting of three sub-tasks which simulate work demands of operators in a chemical plant. One of the tasks involved the monitoring of an automated process. Participants were randomly assigned to three different groups: (1) “Non-Redundant”: participants worked on all tasks alone. (2) “Redundant”: participants were informed that a second crewmember would work in parallel on the monitoring task. (3) “Informed-Redundant”: like the “redundant” condition with the additional information that the crewmate's monitoring performance might be low. Results provide evidence of social loafing and social compensation effects in automation monitoring. Participants in the “redundant” condition cross-checked the automation significantly less than participants in the other groups. This result suggests that human redundancy might not always be the best solution to enhance safety, but might even lead to riskier operator behavior.