<p>In human–automated vehicle cooperative driving, automated vehicles still need to collaborate with human drivers in ambiguous situations. Due to the complexity of such scenarios, automated vehicles cannot fully handle all tasks, and drivers may initiate take-over of vehicle control based on the current context and their assessment of the system’s capability. Compared with passive take-over after system-initiated takeover request, the decision processes underlying driver-initiated take-over have received less research attention. This study used a driving simulation experiment to examine the effects of automated vehicle driving strategies (passing-priority vs yielding-priority) and different time-to-arrival (TTA) gaps (4, 5, and 6&#xa0;s) on driver-initiated take-over decisions at bottleneck sections; the binary decision process was modeled using the drift diffusion model (DDM). Results showed that, among 2281 takeover decision trials, participants initiated 1107 active takeovers, with the passing-priority strategy associated with a higher frequency of active takeovers. Smaller gaps corresponded to shorter decision reaction times. DDM results indicated that the automated driving strategy influenced drivers’ decision caution and risk sensitivity: a yielding-priority driving strategy increased caution and reduced risk sensitivity, while a pass-priority driving strategy had the opposite effect. The TTA of oncoming vehicles affected the rate of evidence accumulation for driver decisions, with shorter TTAs leading to faster evidence accumulation. These findings provide new insight into the cognitive mechanisms underlying driver-initiated take-over in human–vehicle cooperation and offer a basis for optimizing autonomous driving algorithms based on human driver decision models.</p>