Rebecca Wiczorek

Supporting attention allocation in multitask environments: effects of likelihood alarm systems on trust, behavior, and performance

Objective: The aim of the current study was to investigate potential benefits of likelihood alarm systems (LASs) over binary alarm systems (BASs) in a multitask environment. Background: Several problems are associated with the use of BASs, because most of them generate high numbers of false alarms. Operators lose trust in the systems and ignore alarms or cross-check all of them when other information is available. The first behavior harms safety, whereas the latter one reduces productivity. LASs represent an alternative, which is supposed to improve operators’ attention allocation. Method: We investigated LASs and BASs in a dual-task paradigm with and without the possibility to cross-check alerts with raw data information. Participants’ trust in the system, their behavior, and their performance in the alert and the concurrent task were assessed. Results: Reported trust, compliance with alarms, and performance in the alert and the concurrent task were higher for the LAS than for the BAS. The cross-check option led to an increase in alert task performance for both systems and a decrease in concurrent task performance for the BAS, which did not occur in the LAS condition. Conclusion: LASs improve participants’ attention allocation between two different tasks and therefore lead to an increase in alert task and concurrent task performance. The performance maximum is achieved when LAS is combined with a cross-check option for validating alerts with additional information. Application: The use of LASs instead of BASs in safety-related multitask environments has the potential to increase safety and productivity likewise.

Measures of Reliance and Compliance in Aided Visual Scanning

Objective: We study the dependence or independence of reliance and compliance as two responses to alarms to understand the mechanisms behind these responses. Background: Alarms, alerts, and other binary cues affect user behavior in complex ways. The suggestion has been made that there are two different responses to alerts—compliance (the tendency to perform an action cued by the alert) and reliance (the tendency to refrain from actions as long as no alert is issued). The study tests the degree to which these two responses are indeed independent. Method: An experiment tested the effects of the positive and negative predictive values of the alerts (PPV and NPV) on measures of compliance and reliance based on cutoff settings, response times, and subjective confidence. Results: For cutoff settings and response times, compliance was unaffected by the irrelevant NPV, whereas reliance depended on the irrelevant PPV. For subjective estimates, there were no significant effects of the irrelevant variables. Conclusion: Results suggest that compliance is relatively stable and unaffected by irrelevant information (the NPV), whereas reliance is also affected by the PPV. The results support the notion that reliance and compliance are separate, but related, forms of trust. Application: False alarm rates, which affect PPV, determine both the response to alerts (compliance) and the tendency to limit precautions when no alert is issued (reliance).

Benefits of Decision-Support by Likelihood versus Binary Alarm Systems Does the number of stages make a difference?

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.

On the Relation Between Reliance and Compliance in an Aided Visual Scanning Task

Alarms, alerts, and other binary cues affect user behavior in complex ways. One relevant distinction is the suggestion that there are two different responses to alerts – compliance (the tendency to perform an action cued by the alert) and reliance (the tendency to refrain from actions as long as no alert is issued). An experiment tested the dependence of the two behaviors on the Positive and Negative Predictive Values of the alerts (PPV and NPV) to determine whether these are indeed two different behaviors. Results suggest that the compliance is relatively stable and unaffected by irrelevant information (the NPV), while reliance is also affected by the PPV. The results are discussed in terms of multiple-process theories of trust in information sources.

Investigating users’ mental representation of likelihood alarm systems with different thresholds

ABSTRACT This study investigated effects of threshold setting in likelihood alarm systems (LASs) on safety, performance and participants’ mental representation of LASs. Three LASs were compared under two workload conditions. LASs had the same lower (first) threshold separating non-alerts from warnings but differed with regard to their higher (second) threshold dividing warnings from alarms. Behaviour, performance and trust were assessed. Participants of all three conditions responded to almost every alarm but only to parts of the warnings. Threshold setting affected safety, as the most liberal second threshold led to the highest joint human–machine sensitivity. Increasing workload decreased concurrent task performance. A negative correlation of trust in warnings and response rates to warnings reflected reliance and implies a mental representation of graduation of non-alerts. Results suggest that LAS threshold setting allows a precise guidance of users’ behaviour and show the importance of knowledge regarding users’ mental representation for the design of LASs.

On the Influence of Walking on Hazard Detection for Prospective User-Centered Design of an Assistance System for Older Pedestrians

Abstract As older pedestrians are at high risk of being involved in car crashes, an assistance system is currently under development. One of it’s aims is to encourage them to stop walking before looking for traffic. The approach was evaluated in two studies. Age group -young vs. old- and motoric condition -walking vs. standing- served as independent variables in both experiments. Experiment one was conducted in a pedestrian traffic simulation with a traffic related visual hazard detection task with simulated walking. Analysis revealed no age-specific dual-task costs for accuracy and response time. This unexpected result was ascribed to the insufficient operationalization of the walking task, which lacked important aspects of real walking such as requirements of keeping the balance. Therefore, experiment two, comprised real walking but a simple visual task. In the second experiment older participants missed more targets than younger. More important, number of errors increased as a function of motor load only for older participants. Response times were enhanced for older participants and faster for both groups while standing compared to walking. Results are discussed with regard to the development of an assistance systems for older pedestrians and theoretical implications for prospective user-centered experimental design.

Asymmetric effects of false positive and false negative indications on the verification of alerts in different risk conditions

Indications from alerts or alarm systems can be the trigger for decisions, or they can elicit further information search. We report an experiment on the tendency to collect additional information after receiving system indications. We varied the proclivity of the alarm system towards false positive or false negative indications and the perceived risk of the situation. Results showed that false alarm-prone systems led to more frequent re-checking following both alarms and non-alarms in the high risk condition, whereas miss-prone systems led to high re-checking rates only for non-alarms, representing an asymmetry effect. Increasing the risk led to more re-checks with all alarm systems, but it had a stronger impact in the false alarm-prone condition. Results regarding the relation of risk and the asymmetry effect of false negative and false positive indications are discussed.