Characterizing Driver Trust in Vehicle Control Algorithm Parameters

Abstract

Human factors research in vehicle automation has focused on user interfaces such as performance feedback through visual and auditory displays (Blanco et al., 2015). Another approach is to use vehicle dynamics and vibrations as communicative tools for guiding attention (e.g., Morando, Victor, & Dozza, 2016; Walker, Stanton, & Young, 2006; Wiese & Lee, 2007). In our previous study (Price, Venkatraman, Gibson, Lee, & Mutlu, 2016), we showed that the steering wheel deadband, or lateral movement of the vehicle while maintaining lane position, was negatively associated with trust—more lateral movement led to less trust in the algorithm. The present study extends these findings by using Bayesian statistical methods with new control algorithm data. Although the inclusion of additional algorithm characteristics did not improve the trust model, the use of Bayesian statistical methods provides a useful tool to incorporate prior knowledge into an analysis.