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

A New Experimental Paradigm to Manipulate Risk in Human-Automation Research

OBJECTIVE

Two studies serve as a manipulation check of a new experimental multi-task paradigm that can be applied to human-automation research (Virtual Reality Testbed for Risk and Automation Studies; ViRTRAS), in which a subjectively experienceable risk can be manipulated as part of a virtual reality environment.

BACKGROUND

Risk has been postulated as an important contextual factor affecting human-automation interaction. However, experimental evidence is scarce due to the difficulty operationalizing risk in an ethical way. In the new paradigm, risk is varied by the altitude at which participants carry out the task, including the possibility of virtually falling in case of a mistake.

METHOD

Key components of the paradigm were used to investigate participants' risk perception in a low (0.5 m) and high altitude (70 m) using subjective self-reports and objective behavioral measures.

RESULTS

In the high-altitude condition risk perception was significantly higher with medium to large effect sizes. In addition, results of the behavioral measures reveal that participants habituated with length of exposure. However, this habituation seems to occur similarly in both altitude conditions.

CONCLUSION

The manipulation checks were successful. The new paradigm is a promising tool for automation research. It incorporates the contextual factor of risk and creates a situation which is more comparable to what real-life operators experience. Additionally, it meets the same requirements of other multi-task environments in human-automation research.

APPLICATION

The new paradigm provides the basis to vary the contextual factor of risk in human-automation research, which has previously been either neglected or operationalized in an arguably inferior way.

Measuring trust: a text analysis approach to compare, contrast, and select trust questionnaires

Introduction

Trust has emerged as a prevalent construct to describe relationships between people and between people and technology in myriad domains. Across disciplines, researchers have relied on many different questionnaires to measure trust. The degree to which these questionnaires differ has not been systematically explored. In this paper, we use a word-embedding text analysis technique to identify the differences and common themes across the most used trust questionnaires and provide guidelines for questionnaire selection.

Methods

A review was conducted to identify the existing trust questionnaires. In total, we included 46 trust questionnaires from three main domains (i.e., Automation, Humans, and E-commerce) with a total of 626 items measuring different trust layers (i.e., Dispositional, Learned, and Situational). Next, we encoded the words within each questionnaire using GloVe word embeddings and computed the embedding for each questionnaire item, and for each questionnaire. We reduced the dimensionality of the resulting dataset using UMAP to visualize these embeddings in scatterplots and implemented the visualization in a web app for interactive exploration of the questionnaires (https://areen.shinyapps.io/Trust_explorer/).

Results

At the word level, the semantic space serves to produce a lexicon of trust-related words. At the item and questionnaire level, the analysis provided recommendation on questionnaire selection based on the dispersion of questionnaires' items and at the domain and layer composition of each questionnaire. Along with the web app, the results help explore the semantic space of trust questionnaires and guide the questionnaire selection process.

Discussion

The results provide a novel means to compare and select trust questionnaires and to glean insights about trust from spoken dialog or written comments.

Not All Information Is Equal: Effects of Disclosing Different Types of Likelihood Information on Trust, Compliance and Reliance, and Task Performance in Human-Automation Teaming

OBJECTIVE

The study examines the effects of disclosing different types of likelihood information on human operators' trust in automation, their compliance and reliance behaviors, and the human-automation team performance.

BACKGROUND

To facilitate appropriate trust in and dependence on automation, explicitly conveying the likelihood of automation success has been proposed as one solution. Empirical studies have been conducted to investigate the potential benefits of disclosing likelihood information in the form of automation reliability, (un)certainty, and confidence. Yet, results from these studies are rather mixed.

METHOD

We conducted a human-in-the-loop experiment with 60 participants using a simulated surveillance task. Each participant performed a compensatory tracking task and a threat detection task with the help of an imperfect automated threat detector. Three types of likelihood information were presented: overall likelihood information, predictive values, and hit and correct rejection rates. Participants' trust in automation, compliance and reliance behaviors, and task performance were measured.

RESULTS

Human operators informed of the predictive values or the overall likelihood value, rather than the hit and correct rejection rates, relied on the decision aid more appropriately and obtained higher task scores.

CONCLUSION

Not all likelihood information is equal in aiding human-automation team performance. Directly presenting the hit and correct rejection rates of an automated decision aid should be avoided.

APPLICATION

The findings can be applied to the design of automated decision aids.

Do We Really Need More Stages? Comparing the Effects of Likelihood Alarm Systems and Binary Alarm Systems

OBJECTIVE

This research investigates the potential behavioral and performance benefits of a four-stage likelihood alarm system (4-LAS) contrasting a 3-LAS, a binary alarm system with a liberal threshold (lib-BAS), and a BAS with a conservative threshold (con-BAS).

BACKGROUND

Prior research has shown performance benefits of 3-LASs over conventional lib-BASs due to more distinct response strategies and better discriminating true from false alerts. This effect might be further enhanced using 4-LASs. However, the increase in stages could cause users to reduce cognitive complexity by responding in the same way to the two lower and the two higher stages, thus treating the 4-LAS like a con-BAS.

METHOD

All systems were compared using a dual-task paradigm. Response strategies, number of joint human machine (JHM) false alarms (FAs), misses, and sensitivity were regarded.

RESULTS

Compared with the lib-BAS, JHM sensitivity only improved with the 4-LAS and the con-BAS. However, the number of JHM misses was lowest for the con-BAS compared with all other systems.

CONCLUSION

JHM sensitivity improvements can be achieved by using a 4-LAS, as well as a con-BAS. However, only the latter one may also reduce the number of JHM misses, which is remarkable considering that BASs with conservative thresholds a priori commit more inbuilt misses than other systems.

APPLICATION

Results suggest implementing conservative BASs in multi-task working environments to improve JHM sensitivity and reduce the number of JHM misses. When refraining from designing systems which are miss prone, 4-LASs represent a suitable compromise.

Sepsis surveillance: an examination of parameter sensitivity and alert reliability

Objective

To examine performance of a sepsis surveillance system in a simulated environment where modifications to parameters and settings for identification of at-risk patients can be explored in-depth.

Materials and Methods

This was a multiple center observational cohort study. The study population comprised 14 917 adults hospitalized in 2016. An expert-driven rules algorithm was applied against 15.1 million data points to simulate a system with binary notification of sepsis events. Three system scenarios were examined: a scenario as derived from the second version of the Consensus Definitions for Sepsis and Septic Shock (SEP-2), the same scenario but without systolic blood pressure (SBP) decrease criteria (near SEP-2), and a conservative scenario with limited parameters. Patients identified by scenarios as being at-risk for sepsis were assessed for suspected infection. Multivariate binary logistic regression models estimated mortality risk among patients with suspected infection.

Results

First, the SEP-2-based scenario had a hyperactive, unreliable parameter SBP decrease >40 mm Hg from baseline. Second, the near SEP-2 scenario demonstrated adequate reliability and sensitivity. Third, the conservative scenario had modestly higher reliability, but sensitivity degraded quickly. Parameters differed in predicting mortality risk and represented a substitution effect between scenarios.

Discussion

Configuration of parameters and alert criteria have implications for patient identification and predicted outcomes.

Conclusion

Performance of scenarios was associated with scenario design. A single hyperactive, unreliable parameter may negatively influence adoption of the system. A trade-off between modest improvements in alert reliability corresponded to a steep decline in condition sensitivity in scenarios explored.

Effects of Trust, Self-Confidence, and Feedback on the Use of Decision Automation

Operators often fail to rely sufficiently on alarm systems. This results in a joint human-machine (JHM) sensitivity below the one of the alarm system. The 'confidence vs. trust hypothesis' assumes the use of the system depends on the weighting of both values. In case of higher confidence, the task is performed manually, if trust is higher, the user relies on the system. Thus, insufficient reliance may be due to operators' overconfidence in their own abilities and/or insufficient trust in the decision automation, but could be mitigated by providing feedback. That was investigated within a signal detection task, supported by a system with either higher sensitivity (HSS) or lower sensitivity (LSS) than the human, while being provided with feedback or not. We expected disuse of the LSS and insufficiently reliance on the HSS, in the condition without feedback. The feedback was expected to increase reliance on the HSS through an increase in trust and/or decreases in confidence, and thus, improve performance. Hypotheses were partly supported. Confidence in manual performance was similar to trust in the HSS even though humans' sensitivity was significantly lower than systems' sensitivity. While confidence had not effect on reliance or JHM sensitivity, trust was found to be positively related with both. We found disuse of the HSS, that could be improved through feedback, increasing also trust and JHM sensitivity. However, contrary to 'confidence vs. trust' expectations, participants were also found to make use of the LSS. This misuse could not be reduced by feedback. Results indicate the use of feedback being beneficial for the overall performance (with HSS only). Findings do not support the idea that misuse or disuse of the system may result from comparison of confidence and trust. We suppose it may rather be the product of users' wrong strategy of function allocation, based on the underlying idea of team work in combination with missing assignment of responsibility. We discuss this alternative explanation.

No Effect of Cue Format on Automation Dependence in an Aided Signal Detection Task

OBJECTIVE

To investigate whether manipulating the format of an automated decision aid's cues can improve participants' information integration strategies in a signal detection task.

BACKGROUND

Automation-aided decision making is often suboptimal, falling well short of statistically ideal levels. The choice of format in which the cues from the aid are displayed may help users to better understand and integrate the aid's judgments with their own.

METHOD

Participants performed a signal detection task that asked them to classify random dot images as either blue or orange dominant. They made their judgments either unaided or with assistance from a 93% reliable automated decision aid. The aid provided a binary judgment, along with an estimate of signal strength in the form of either a raw value, a likelihood ratio, or a confidence rating (Experiments 1 and 2) or a binary judgment along with either a verbal or verbal-visuospatial expression of confidence (Experiment 3). Aided sensitivity was benchmarked to the predictions of various statistical models of collaborative decision making.

RESULTS AND CONCLUSION

Aided performance was suboptimal, matching the predictions of some of the least efficient models. Most importantly, performance was similar across cue formats.

APPLICATION

Results indicate that changes to the format in which cues from a signal detection aid are rendered are unlikely to dramatically improve the efficiency of automation-aided decision making.

The Effect of Culture on Trust in Automation

Trust in automation has become a topic of intensive study since the late 1990s and is of increasing importance with the advent of intelligent interacting systems. While the earliest trust experiments involved human interventions to correct failures/errors in automated control systems, a majority of subsequent studies have investigated information acquisition and analysis decision aiding tasks such as target detection for which automation reliability is more easily manipulated. Despite the high level of international dependence on automation in industry, almost all current studies have employed Western samples primarily from the U.S. The present study addresses these gaps by running a large sample experiment in three (U.S., Taiwan, and Turkey) diverse cultures using a “trust sensitive task” consisting of both automated control and target detection subtasks. This article presents results for the target detection subtask for which reliability and task load were manipulated. The current experiments allow us to determine whether reported effects are universal or specific to Western culture, vary in baseline or magnitude, or differ across cultures. Results generally confirm consistent effects of manipulations across the three cultures as well as cultural differences in initial trust and variation in effects of manipulations consistent with 10 cultural hypotheses based on Hofstede's Cultural Dimensions and Leung and Cohen's theory of Cultural Syndromes. These results provide critical implications and insights for correct trust calibration and to enhance human trust in intelligent automation systems across cultures. Additionally, our results would be useful in designing intelligent systems for users of different cultures. Our article presents the following contributions: First, to the best of our knowledge, this is the first set of studies that deal with cultural factors across all the cultural syndromes identified in the literature by comparing trust in the Honor, Face, Dignity cultures. Second, this is the first set of studies that uses a validated cross-cultural trust measure for measuring trust in automation. Third, our experiments are the first to study the dynamics of trust across cultures.

Acoustic features of auditory medical alarms-An experimental study of alarm volume

Audible alarms are a ubiquitous feature of all high-paced, high-risk domains such as aviation and nuclear power where operators control complex systems. In such settings, a missed alarm can have disastrous consequences. It is conventional wisdom that for alarms to be heard, "louder is better," so that alarm levels in operational environments routinely exceed ambient noise levels. Through a robust experimental paradigm in an anechoic environment to study human response to audible alerting stimuli in a cognitively demanding setting, akin to high-tempo and high-risk domains, clinician participants responded to patient crises while concurrently completing an auditory speech intelligibility and visual vigilance distracting task as the level of alarms were varied as a signal-to-noise ratio above and below hospital background noise. There was little difference in performance on the primary task when the alarm sound was -11 dB below background noise as compared with +4 dB above background noise-a typical real-world situation. Concurrent presentation of the secondary auditory speech intelligibility task significantly degraded performance. Operator performance can be maintained with alarms that are softer than background noise. These findings have widespread implications for the design and implementation of alarms across all high-consequence settings.

An investigation of sepsis surveillance and emergency treatment on patient mortality outcomes: An observational cohort study

Objective

To determine the prevalence of initiating the sepsis 3-h bundle of care and estimate effects of bundle completion on risk-adjusted mortality among emergency department (ED) patients screened-in by electronic surveillance.

Materials and Methods

This was a multiple center observational cohort study conducted in 2016. The study population was comprised of patients screened-in by St. John Sepsis Surveillance Agent within 4 h of ED arrival, had a sepsis bundle initiated, and admitted to hospital. We built multivariable logistic regression models to estimate impact of a 3-h bundle completed within 3 h of arrival on mortality outcomes.

Results

Approximately 3% ED patients were screened-in by electronic surveillance within 4 h of arrival and admitted to hospital. Nearly 7 in 10 (69%) patients had a bundle initiated, with most bundles completed within 3 h of arrival. The fully-adjusted risk model achieved good discrimination on mortality outcomes [area under the receiver operating characteristic 0.82, 95% confidence interval (CI) 0.79-0.85] and estimated 34% reduced mortality risk among patients with a bundle completed within 3 h of arrival compared to non-completers.

Discussion

The sepsis bundle is an effective intervention for many vulnerable patients, and likely to be completed within 3 h after arrival when electronic surveillance with reliable alert notifications are integrated into clinical workflow. Beginning at triage, the platform and sepsis program enables identification and management of patients with greater precision, and increases the odds of good outcomes.

Conclusion

Sepsis surveillance and clinical decision support accelerate accurate recognition and stratification of patients, and facilitate timely delivery of health care.

Benchmarking Aided Decision Making in a Signal Detection Task

OBJECTIVE

A series of experiments examined human operators' strategies for interacting with highly (93%) reliable automated decision aids in a binary signal detection task.

BACKGROUND

Operators often interact with automated decision aids in a suboptimal way, achieving performance levels lower than predicted by a statistically ideal model of information integration. To better understand operators' inefficient use of decision aids, we compared participants' automation-aided performance levels with the predictions of seven statistical models of collaborative decision making.

METHOD

Participants performed a binary signal detection task that asked them to classify random dot images as either blue or orange dominant. They made their judgments either unaided or with assistance from a 93% reliable automated decision aid that provided either graded (Experiments 1 and 3) or binary (Experiment 2) cues. We compared automation-aided performance with the predictions of seven statistical models of collaborative decision making, including a statistically optimal model and Robinson and Sorkin's contingent criterion model.

RESULTS AND CONCLUSION

Automation-aided sensitivity hewed closest to the predictions of the two least efficient collaborative models, well short of statistically ideal levels. Performance was similar whether the aid provided graded or binary judgments. Model comparisons identified potential strategies by which participants integrated their judgments with the aid's.

APPLICATION

Results lend insight into participants' automation-aided decision strategies and provide benchmarks for predicting automation-aided performance levels.

The use of adaptable automation: Effects of extended skill lay-off and changes in system reliability

This experiment aimed to examine how skill lay-off and system reliability would affect operator behaviour in a simulated work environment under wide-range and large-choice adaptable automation comprising six different levels. Twenty-four participants were tested twice during a 2-hr testing session, with the second session taking place 8 months after the first. In the middle of the second testing session, system reliability changed. The results showed that after the retention interval trust increased and self-confidence decreased. Complacency was unaffected by the lay-off period. Diagnostic speed slowed down after the retention interval but diagnostic accuracy was maintained. No difference between experimental conditions was found for automation management behaviour (i.e. level of automation chosen and frequency of switching between levels). There were few effects of system reliability. Overall, the findings showed that subjective measures were more sensitive to the impact of skill lay-off than objective behavioural measures.

Investigating Benefits of Likelihood Alarm Systems in Presence of Alarm Validity Information

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.

A Multidisciplinary Sepsis Program Enabled by a Two-Stage Clinical Decision Support System: Factors That Influence Patient Outcomes

Sepsis is an inflammatory response triggered by infection, with risk of in-hospital mortality fueled by disease progression. Early recognition and intervention by multidisciplinary sepsis programs may reverse the inflammatory response among at-risk patient populations, potentially improving outcomes. This retrospective study of a sepsis program enabled by a 2-stage sepsis Clinical Decision Support (CDS) system sought to evaluate the program's impact, identify early indicators that may influence outcomes, and uncover opportunities for quality improvement. Data encompassed 16 527 adult hospitalizations from 2014 and 2015. Of 2108 non-intensive care unit patients screened-in by sepsis CDS, 97% patients were stratified by 177 providers. Risk of adverse outcome improved 30% from baseline to year end, with gains materializing and stabilizing at month 7 after sepsis program go-live. Early indicators likely to influence outcomes include patient age, recent hospitalization, electrolyte abnormalities, hypovolemic shock, hypoxemia, patient location when sepsis CDS activated, and specific alert patterns.

From Trust in Automation to Decision Neuroscience: Applying Cognitive Neuroscience Methods to Understand and Improve Interaction Decisions Involved in Human Automation Interaction

Human automation interaction (HAI) systems have thus far failed to live up to expectations mainly because human users do not always interact with the automation appropriately. Trust in automation (TiA) has been considered a central influence on the way a human user interacts with an automation; if TiA is too high there will be overuse, if TiA is too low there will be disuse. However, even though extensive research into TiA has identified specific HAI behaviors, or trust outcomes, a unique mapping between trust states and trust outcomes has yet to be clearly identified. Interaction behaviors have been intensely studied in the domain of HAI and TiA and this has led to a reframing of the issues of problems with HAI in terms of reliance and compliance. We find the behaviorally defined terms reliance and compliance to be useful in their functionality for application in real-world situations. However, we note that once an inappropriate interaction behavior has occurred it is too late to mitigate it. We therefore take a step back and look at the interaction decision that precedes the behavior. We note that the decision neuroscience community has revealed that decisions are fairly stereotyped processes accompanied by measurable psychophysiological correlates. Two literatures were therefore reviewed. TiA literature was extensively reviewed in order to understand the relationship between TiA and trust outcomes, as well as to identify gaps in current knowledge. We note that an interaction decision precedes an interaction behavior and believe that we can leverage knowledge of the psychophysiological correlates of decisions to improve joint system performance. As we believe that understanding the interaction decision will be critical to the eventual mitigation of inappropriate interaction behavior, we reviewed the decision making literature and provide a synopsis of the state of the art understanding of the decision process from a decision neuroscience perspective. We forward hypotheses based on this understanding that could shape a research path toward the ability to mitigate interaction behavior in the real world.

A two-stage clinical decision support system for early recognition and stratification of patients with sepsis: an observational cohort study

OBJECTIVE

To examine the diagnostic accuracy of a two-stage clinical decision support system for early recognition and stratification of patients with sepsis.

DESIGN

Observational cohort study employing a two-stage sepsis clinical decision support to recognise and stratify patients with sepsis. The stage one component was comprised of a cloud-based clinical decision support with 24/7 surveillance to detect patients at risk of sepsis. The cloud-based clinical decision support delivered notifications to the patients' designated nurse, who then electronically contacted a provider. The second stage component comprised a sepsis screening and stratification form integrated into the patient electronic health record, essentially an evidence-based decision aid, used by providers to assess patients at bedside.

SETTING

Urban, 284 acute bed community hospital in the USA; 16,000 hospitalisations annually.

PARTICIPANTS

Data on 2620 adult patients were collected retrospectively in 2014 after the clinical decision support was implemented.

MAIN OUTCOME MEASURE

'Suspected infection' was the established gold standard to assess clinical decision support clinimetric performance.

RESULTS

A sepsis alert activated on 417 (16%) of 2620 adult patients hospitalised. Applying 'suspected infection' as standard, the patient population characteristics showed 72% sensitivity and 73% positive predictive value. A postalert screening conducted by providers at bedside of 417 patients achieved 81% sensitivity and 94% positive predictive value. Providers documented against 89% patients with an alert activated by clinical decision support and completed 75% of bedside screening and stratification of patients with sepsis within one hour from notification.

CONCLUSION

A clinical decision support binary alarm system with cross-checking functionality improves early recognition and facilitates stratification of patients with sepsis.

Complacency and Automation Bias in the Use of Imperfect Automation

OBJECTIVE

We examine the effects of two different kinds of decision-aiding automation errors on human-automation interaction (HAI), occurring at the first failure following repeated exposure to correctly functioning automation. The two errors are incorrect advice, triggering the automation bias, and missing advice, reflecting complacency.

BACKGROUND

Contrasts between analogous automation errors in alerting systems, rather than decision aiding, have revealed that alerting false alarms are more problematic to HAI than alerting misses are. Prior research in decision aiding, although contrasting the two aiding errors (incorrect vs. missing), has confounded error expectancy.

METHOD

Participants performed an environmental process control simulation with and without decision aiding. For those with the aid, automation dependence was created through several trials of perfect aiding performance, and an unexpected automation error was then imposed in which automation was either gone (one group) or wrong (a second group). A control group received no automation support.

RESULTS

The correct aid supported faster and more accurate diagnosis and lower workload. The aid failure degraded all three variables, but "automation wrong" had a much greater effect on accuracy, reflecting the automation bias, than did "automation gone," reflecting the impact of complacency. Some complacency was manifested for automation gone, by a longer latency and more modest reduction in accuracy.

CONCLUSIONS

Automation wrong, creating the automation bias, appears to be a more problematic form of automation error than automation gone, reflecting complacency.

IMPLICATIONS

Decision-aiding automation should indicate its lower degree of confidence in uncertain environments to avoid the automation bias.

Benefits of Decision-Support by Likelihood versus Binary Alarm Systems

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.