Navigating uncertainty: Safety solutions for autonomous vehicle integration into mixed-mode mobility

Abstract

Uncertainty fundamentally shapes human perception and decision-making, a factor that is becoming increasingly critical as autonomous vehicles (AVs) begin to share physical and social spaces with humans. This perspective paper synthesizes insights from neuroscience, robotics, and behavioral science represent uncertainty and act under it. We identify a significant disconnect between low-level mathematical concepts of uncertainty and risk, the uncertainty-aware prediction and planning methods used in robotics, and the high-level psychological uncertainty experienced by humans. Here, we conceptualize an integration pathway for mixed-mode mobility: we show that all three treat control under partial information as the same two-stage problem – probabilistic inference of the current and future state, followed by a risk-weighted choice of action – and we make this shared structure explicit. We thus argue that AVs should quantify and calibrate their own uncertainty, select risk metrics that reflect both individual and collective safety, model how their actions shape human uncertainty and behavior, and communicate intent and confidence in ways that support predictable interaction. This synthesis positions psychological uncertainty as a behaviorally relevant variable for AV design and outlines a research agenda for safer, more understandable transport systems.