Understanding the effect of workload on automation use for younger and older adults

Abstract

Objective: This study examined how individuals, younger and older, interacted with an imperfect automated system. The impact of workload on performance and automation use was also investigated.

Background: Automation is used in situations characterized by varying levels of workload. As automated systems spread to domains such as transportation and the home, a diverse population of users will interact with automation. Research is needed to understand how different segments of the population use automation.

Method: Workload was systematically manipulated to create three levels (low, moderate, high) in a dual-task scenario in which participants interacted with a 70% reliable automated aid. Two experiments were conducted to assess automation use for younger and older adults.

Results: Both younger and older adults relied on the automation more than they complied with it. Among younger adults, high workload led to poorer performance and higher compliance, even when that compliance was detrimental. Older adults' performance was negatively affected by workload, but their compliance and reliance were unaffected.

Conclusion: Younger and older adults were both able to use and double-check an imperfect automated system. Workload affected how younger adults complied with automation, particularly with regard to detecting automation false alarms. Older adults tended to comply and rely at fairly high rates overall, and this did not change with increased workload.

Application: Training programs for imperfect automated systems should vary workload and provide feedback about error types, and strategies for identifying errors. The ability to identify automation errors varies across individuals, thereby necessitating training.

Figure 1

Screenshots of the components of the Automated Warehouse Management System: (a) the Receiving Packages task was presented on the right side of the computer screen and (b) the Dispatching Trucks task was presented on the left side of the computer screen. If participants held down the space bar, they would see the interior of the truck, as depicted in Panel B. However, when the space bar was not depressed, participants saw only what is depicted in Panel A.

Figure 2

The Receiving Packages search load for the low, moderate, and high workload groups in Experiment 1.

Figure 3

Younger adults’ (a) percentage correct, incorrect, and time-out on the Receiving Packages task and (b) percentage correct, not full, and overloaded in the Dispatching Trucks task for each workload condition.

Figure 4

Younger adults’ (a) compliance with and (b) reliance on the automation for trials in which the automation was correct as well as trials in which the automation was incorrect (false alarm or miss, respectively). Optimal compliance or reliance is approximately 100% for correct trials and 0% for incorrect trials. Error bars depict the standard error of the mean.

Figure 5

The Receiving Packages search load for the low, moderate, and high workload groups in Experiment 2.

Figure 6

Older adults’ (a) percentage correct, incorrect, and time-out on the Receiving Packages task and (b) percentage correct, not full, and overloaded in the Dispatching Trucks task for each workload condition.

Figure 7

Older adults’ (a) compliance with and (b) reliance on the automation for trials in which the automation was correct as well as trials in which the automation was incorrect (false alarm or miss, respectively). Optimal compliance or reliance is approximately 100% for correct trials and 0% for incorrect trials. Error bars depict the standard error of the mean.