The Effect of Automation Reliability on User Automation Trust and Reliance in a Search-and-Rescue Scenario

Abstract

Advances in modern day technology are rapidly increasing the ability of engineers to automate ever more complicated tasks. Often these automated aids are paired with human operators who can supervise their work to ensure that it is free of errors and to even take control of the system if it malfunctions (e.g., pilots supervising an autopilot feature). The goal of this collaboration, between humans and machines, is that it can enhance performance beyond what would be possible by either alone. Arguably the success of this partnership depends in part upon attributions an operator develops that help guide their interaction with the automation. One particular factor that has been shown to guide operator reliance on an automated ‘teammate’ is trust. The following study examined 140 participants performing a simulated search-and-rescue task. The goal of this experiment was to examine the relationship between automated agent's reliability, operator trust, operator reliance, and performance scores. Results indicated that greater automation reliability is positively correlated with greater user reliance ( r = .66), perceived trust ( r = .21), and performance scores ( r = .34). These results indicate that more reliable aids are rated as significantly higher in terms of perceived trust and relied upon more than less reliable aids. Additionally, the size of the effect is much larger for operator behaviors (i.e., reliance) compared to more subjective measures (i.e., self-reported trust).