When Takeovers Require Cognitive Control: Effects of Partial Automation and Cognitive Load on Driver-Initiated Responses to Traffic Hazards on Urban Roads

Partially automated driving (PAD) systems are becoming increasingly prevalent in consumer vehicles. While drivers are obliged to monitor the roadway and the system during PAD, they do not always comply with this requirement and may engage in non-driving-related tasks (NDRTs). Literature has emphasized the detrimental effects of visual–manual NDRTs on drivers’ capability to takeover control from PAD while reporting negligible effects of non-visual NDRTs (e.g., hands-free phone calls) and cognitive load. However, most studies focused on alerted scenarios, with takeover requests lowering drivers’ reliance on anticipation and cognitive load, masking the effects of cognitive load. Related work also examined PAD conditions alone, neglecting the combined effects of automation and cognitive load on drivers’ responses. To address the gaps, we conducted a driving simulator study examining the influence of cognitive load, induced by a 2-back task, on driver-initiated responses to urban traffic hazards (e.g., crossing pedestrians, oncoming vehicles) during both manual driving and PAD. The timing and quality of driver-initiated responses were analyzed separately for each hazard with detail. Results showed that (i) cognitive load effects manifested most when drivers encountered hazards with lower perceptual salience (e.g., a pedestrian gradually approaching the roadway); (ii) driver-initiated responses during PAD were less proactive and relied more on cognitive control; and (iii) this shift led to higher situational criticality and heightened susceptibility of responses to cognitive load during PAD. Together, these findings underscore an elevated safety risk associated with cognitive load during PAD, especially in scenarios requiring cognitively demanding, driver-initiated intervention.