Uncertainty on Display: The Effects of Communicating Confidence Cues in Autonomous Vehicle-Pedestrian Interactions

2025-07-24

AI Summary

Autonomous vehicles (AVs) fail to effectively communicate decision uncertainty to pedestrians, undermining interaction transparency and trust. This study conducts the first systematic comparison of two uncertainty communication modalities—explicit (onboard confidence percentage display) versus implicit (kinematics-based deceleration/pausing patterns)—within a VR environment using a within-subjects experimental design. We evaluate their effects on pedestrian perceived safety, trust, and overall user experience. Results indicate that explicit communication significantly improves all subjective metrics and is strongly preferred; implicit cues, particularly under low-confidence conditions, often induce ambiguity. Methodologically, this work innovatively reframes algorithmic confidence—not as an internal system variable but as a designable external interaction dimension. It proposes the “explainability-as-interface” paradigm, establishing an empirical foundation and actionable design guidelines for AV-pedestrian cooperative interaction. (149 words)

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Reasoning under Uncertainty: Uncertainty Representations
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Security and Privacy: Data transparency and provenance
• User Modeling, Personalization and Recommendation: Accountability, Transparency, and Ethics for personalization

Abstract

Uncertainty is an inherent aspect of autonomous vehicle (AV) decision-making, yet it is rarely communicated to pedestrians, which hinders transparency. This study investigates how AV uncertainty can be conveyed through two approaches: explicit communication (confidence percentage displays) and implicit communication (vehicle motion cues), across different confidence levels (high and low). Through a within-subject VR experiment (N=26), we evaluated these approaches in a crossing scenario, assessing interface qualities (visibility and intuitiveness), how well the information conveyed the vehicle's level of confidence, and their impact on participants' perceived safety, trust, and user experience. Our results show that explicit communication is more effective and preferred for conveying uncertainty, enhancing safety, trust, and user experience. Conversely, implicit communication introduces ambiguity, especially when AV confidence is low. This research provides empirical insights into how uncertainty communication shapes pedestrian interpretation of AV behaviour and offer design guidance for external interfaces that integrate uncertainty as a communicative element.

Problem

• How to communicate AV uncertainty to pedestrians effectively
• Comparing explicit vs implicit confidence cues in AV interactions
• Impact of uncertainty display on pedestrian safety and trust

Innovation

• Explicit confidence percentage displays enhance transparency
• Implicit vehicle motion cues introduce ambiguity
• VR experiment evaluates safety, trust, user experience