Human Risk Perception in Driving
Understanding and modeling how humans perceive risk in dynamic driving environments
This research studies how humans perceive and respond to risk in dynamic driving environments. We investigate how perceived risk changes with traffic conditions and scene dynamics, and how computational models can learn the visual evidence underlying human risk judgments.
Our work spans both human-centered studies and computer vision approaches to risk perception. One line of research examines how transitions in traffic density influence perceived risk and driver responses. Another line, CoRE, studies coarse-to-fine risk understanding from driving videos, learning fine-grained temporal and object-level risk evidence from coarse video-level supervision.
Selected Work
Asymmetric Shifts in Risk Perception: Evaluating Driver Responses to Traffic Density Transitions
Peihang Li, Nade Liang
In Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 2025.
CoRE: Coarse-to-Fine Risk Evidence Learning
Kaiser Hamid, Nade Liang *Ongoing research on weakly supervised learning of temporal and object-level risk evidence from driving videos.