Understanding hazard recognition behavior using eye-tracking metrics in a VR-simulated environment: Learning from successful and failed conditions

Effective hazard recognition is critical for ensuring safety in complex construction sites. Workers often encounter distinct hazard recognition conditions, influenced by limited cognitive abilities and attentional resources. However, the cognitive processes such as attention allocation, cognitive load, and emotional arousal between successful and failed (misidentified, overlooked) conditions have not been fully investigated. This study employed decision time and eye movement metrics to investigate the cognitive processes among these three recognition conditions. The results revealed that successful hazard recognitions involved increased attentional resources, lower cognitive load, and more emotional arousal, evidenced by proper decision time, a prolonged fixation duration, smaller average pupil diameter, and pupil dilation while gazing at areas of interest (AOI). Conversely, participants in the misidentified condition appeared to have limited information extraction and comprehension abilities with long fixation duration and more saccade count, while those in the overlooked condition demonstrated poor visual search skills with less saccade count. Both misidentified and overlooked conditions exhibited high cognitive load and low emotional arousal, as indicated by larger average pupil size and pupil contraction. These findings offer valuable insights for developing targeted training strategies to address the deficiencies of cognitive ability associated with each recognition condition.