Subtle motion cues by automated vehicles can nudge human drivers’ decisions: Empirical evidence and computational cognitive model

Automated vehicles (AVs) can bring about numerous benefits for society but they are unprepared to enter our roads yet, in large part because of the difficulties in interacting with human road users. Addressing this issue requires not only developing novel interaction planning approaches for AVs, but also understanding human behavior in interactions with them. Existing investigations of such behavior have predominantly focused on situations in which AV a priori needs to take action because the human has the right of way. However, future AVs would likely need to proactively manage interactions even if they have the right of way over humans, e.g., a human driver taking a left turn in front of the approaching AV. Yet it remains unclear how AVs could behave in such interactions and how humans would react to them. To address this issue, here we investigated behavior of human drivers (N=19) when interacting with an oncoming AV during unprotected left turns in a driving simulator experiment. We measured the outcomes and timing of participants' decisions when interacting with an AV which, initially moving with the constant speed, a) briefly decelerates for 1s and then accelerates back to its original speed (deceleration "nudge"), b) briefly accelerates for 1s and then decelerates back to its original speed (acceleration "nudge"), c) decelerates for 2s; or d) accelerates for 2s. We compared human participants' behavior in these conditions with responses to a constant-speed AV. We found that participants were sensitive to the deceleration nudge but not acceleration nudge. Most importantly, we compared the obtained data to predictions of several variants of a drift-diffusion model of human decision making and investigated the effect of different model assumptions on its behavior. The most parsimonious model that captured observed decision outcomes and response time distributions hypothesized noisy integration of dynamic time-to-arrival and distance information to a fixed decision boundary, with an initial accumulation bias towards the Go decision. Finally, we illustrated how this cognitive model, even if fitted to a limited set of experimental conditions, can flexibly generate predictions of human responses to arbitrary longitudinal AV maneuvers. These predictions are quantitative and testable, making the model a potentially useful tool for both informing future studies of human behavior and incorporating insights from such studies into computational frameworks for AV interaction planning.