Linking Behaviour and Perception to Evaluate Meaningful Human Control over Partially Automated Driving

A team led by Ashwin George (TU Delft) investigated how to empirically evaluate meaningful human control (MHC) in partially automated driving systems. The study, published on arXiv, addressed the tension where drivers remain legally responsible but have reduced active control. Twenty-four participants completed a simulator experiment with silent automation failures in two driving modes: haptic shared control (HSC), where automation provides continuous force feedback, and traded control (TC), where control switches entirely between driver and automation. The researchers derived behavioral metrics (steering torque conflict, reaction times) from telemetry and subjective perception scores from post-trial surveys to test hypothesized relations based on MHC properties.

Confirmatory analysis revealed a significant negative correlation between drivers' perception that the automated vehicle (AV) understands them and the conflict in steering torques. Surprisingly, an exploratory analysis found a positive correlation between reaction times and the perception of having sufficient control. Qualitative feedback from open-ended questionnaires identified three key factors reducing perceived MHC: mismatched intentions between driver and automation, lack of safety, and resistance to driver inputs. Conversely, subtle haptic guidance that aligned with driver intent enhanced the sense of control. The authors conclude that future designs should prioritize effortless driver interventions, transparent communication of automation intent, and context-sensitive authority allocation.