Understanding the role of expectations on human responses to an automated system
Barg-Walkow, Laura Hillary
MetadataShow full item record
As automation becomes increasing ubiquitous, it is important to know how differences in introducing automated systems will affect human-automation interactions. There are two main ways of introducing expected reliability of an automated system to users: explicitly telling operators what to expect or giving operators experience using the system. This study systematically investigated the effect of expectation format initially and over time on: 1) perceptions of reliability and system usage, and 2) human responses to automation (e.g., compliance, reliance, and overall dependence). Initially, there was an effect of expected level for explicit statement groups, whereas there was no effect of expected level for initial exposure groups. Over time, explicit statement groups had more stable perceptions of system reliability than the initial exposure groups. In general, perceived reliability did not converge to actual system reliability (75%) by the end of the study. Additionally, perceived reliability had a weak, but positive relationship with actual system use, whereas perceptions of system use (e.g., perceived dependence) had a strong, but negative relationship with actual system use. Outside of initial effects seen with perceived reliability, there were few initial differences between expectation formats. Almost all groups tended to initially comply more than rely, with the exception of the initial exposure – lower-than group. Over time, level of expectation for initial exposure groups influenced reliance. There were no differences between expectation groups on compliance and dependence over time. In general, dependence and compliance increased or stayed the same as time using the system increased. This pattern was also seen with reliance, with the exception of the initial exposure - higher-than group decreasing reliance over time. Results from this study have implications for both theory and practice. The research findings both support and augment the existing conceptual model of automation. A better understanding of the differential effects of expectation format and introduced level of expectations can lead to introductions of automated systems that are best suited to the system’s goals, ultimately improving system performance.