Automated driving systems currently communicate operational status through binary indicators that tell drivers whether the system is or is not engaged. Therefore, drivers don’t receive any information about declining system confidence before a takeover request occurs, leaving them without intermediate states to prepare for a transition of control (Wang et al., 2024). Transparency cues that communicate system confidence to drivers have been proposed as a solution, yet designs have typically been evaluated post-implementation rather than developed from driver input.
This study tests two transparency cue designs, informed by driver co-design priorities, that differ in their temporal structure. The continuous condition presents a trendline that updates in real-time, showing system confidence as it changes in response to dynamic system conditions (George et al., 2026). The sequential condition presents a categorical display that changes state at fixed thresholds: autopilot engaged, confidence reduced and take over now. Both designs present their cue in the same instrument cluster location. The continuous display provides a visible change history, enabling pattern recognition and anticipatory monitoring. The sequential display provides only the current state without any indication of trends in confidence.
Participants completed a simulated highway drive in a STISIM3 driving simulator while wearing Tobii Pro eye tracking glasses. Prefrontal cortex oxygenation was recorded bilaterally using functional near-infrared spectroscopy (fNIRS) as a physiological measure of cognitive workload. Ten degradation windows occurred during the drive, six of which escalated to a takeover request. Eye tracking measures captured the proportion of each degradation window spent looking at the display versus the road, and the rate at which gaze transitioned between the two regions (George et al., 2026). Lane keeping performance was measured as the standard deviation of lateral position across escalation events (Huang et al., 2025). Post-drive measures included perceptions of situational awareness, cognitive workload, usability, and trust. Semi-structured interviews examined how drivers interpreted and used each display to inform their decisions.
This mixed-methods design addresses whether continuous transparency changes how drivers monitor and respond to declining system confidence compared to a sequential categorical display. The eye tracking and fNIRS data test whether a richer display improves monitoring or increases cognitive demand, while the subjective measures and interviews capture how each display shaped drivers’ understanding of the system. The presentation will report results from both the behavioural and subjective measures alongside the qualitative findings.