Anticipating a Vehicle’s Next Move: Perceived Autonomy of the Lead Vehicle Increases Following Distance

Authors:
Preprint
Publisher website
Google Scholar
Abstract
Human-machine interaction is becoming increasingly relevant in traffic, where autonomous vehicles (AVs) and manual vehicles (MVs) share the same roads. When encountering AVs, drivers must interpret a machine agent’s behavior rather than a human’s. Does this impact human drivers’ behavior? Prior research suggests that people engage in mentalizing processes toward robots and autonomous agents, resulting in adjusted social behavior. Whether such processes similarly influence how human drivers respond to AVs remains unclear. We tested whether the perceived autonomy of vehicles impacts their perceived predictability and how this influences driving behavior of other drivers in a within-subjects virtual reality (VR) lab study (N = 690,094 observations from 61 participants) and in a field experiment (N = 27,198 observations from 1,414 vehicles). Across both studies, perceived autonomy of the vehicle was associated with greater following distance, though the evidence varied in strength. In the VR study, drivers maintained a significantly larger following distance (9%) to AVs relative to MVs. In the field experiment, manipulating perceived autonomy produced no overall difference in following distance to the same AV presented under different autonomy framings; however, an exploratory growth curve model revealed that following distance developed differently over time depending on perceived autonomy. Overall, the data offers indications that perceived autonomy might impact driving behavior, as the mental model could adjust differently for AVs during the first encounters. We quantify following distances in mixed-traffic, with implications for traffic planning regarding lane capacity and traffic flow at merges and bottlenecks when adopting AVs.