The Effects of In-vehicle Automation and Reliability on Driver Situation Awareness and Trust

Abstract

The cognitive construct of situation awareness (SA) has not been well developed in the domain of driving. The objective of this study was to define a new transactional model of SA in various driving behaviors and activities, as influenced by automation and in-vehicle device use. Specifically, this study investigated the implications of adaptive cruise control (ACC) and cellular phone use in driving on a direct and objective measure of SA; investigate the effect of varying reliability of in-vehicle automation (navigation aids) on driver SA and trust; and assess differences in human trust in a human aid versus an automation aid in a simulated driving task.

Twenty participants drove a virtual car and performed a freeway driving task (Experiment A) as well as a suburb navigation task (Experiment B). In the freeway driving, participants were required to drive using ACC or manual control modes, and received navigation information from one of two sources: a human or in-vehicle automation aid via cell phone or separate display screen, respectively. During the navigation driving, participants were required to drive through the suburban area following all traffic signs and directions from the navigation aid under different levels of information reliability (100%, 80% and 60%). A control condition was also used in which aids only presented a telemarketing survey and participants navigated using a map. Driver SA was assessed at the end of each experiment using a SA global assessment technique. Driver workload was collected at the same time using the NASA- TLX. Driver trust in the navigation aid information was measured using a subjective survey of initial subject trust expectations as well as a subjective rating at the close of each trial (end of Experiment B). Across both experiments, multiple dimensions of task performance were measured.

MANOVA results for Experiment A revealed significant main effects for both ACC control mode and navigation aid type on driver performance, but no interaction effect. Findings were similar for driver SA except there was no effect of aid type. ANOVA results indicated use of the ACC system to improve driver SA and operational driver behaviors by reducing the task load in Experiment A. MANOVA results for Experiment B revealed only a significant effect of navigation aid reliability on driver performance and SA. ANOVA results revealed that perfect navigation information generally improved driving performance and driver SA for strategic driving behavior compared to unreliable navigation aid information and the control condition (task-irrelevant information). The results also revealed that drivers had higher initial trust expectations and expectation of fewer errors by the automation compared to the human. However, when participants experienced automation aid errors or inefficiency, their trust in the automation declined more sharply than trust in the human advisor. The results of this empirical work provide insight into the importance of driver SA in operational and strategic type driving tasks and associated actions. It identifies in-vehicle automation and devices as underlying factors in linkages of levels of SA to specific driving behaviors in the transactional model and serves to quantify the impact of the factors on driving performance. Validation of the proposed model and identification of other underlying factors may lead to its future use for predictive purposes.