A brief and unsupervised online intervention improves performance on a validated test of hazard perception skill used for driver licensing

Abstract

Drivers who have higher levels of hazard perception skill also tend to have fewer crashes. Training designed to improve this skill has therefore been proposed as a strategy for reducing crash risk. To date, however, hazard perception training has only been evaluated in supervised settings. This means that improvements in hazard perception skill resulting from such training may not generalize to unsupervised situations, which may limit opportunities for large scale roll-out via automated delivery methods. In the present study, we investigated whether a brief video-based training intervention could improve hazard perception skill when drivers completed it online without supervision. The training involved drivers watching videos of traffic scenes, while generating a commentary of what they were searching for, monitoring, and anticipating in each scene. Drivers then compared their own commentary to a pre-recorded commentary generated by an expert driver, hence allowing for performance feedback without an instructor present. A convenience sample of 93 drivers (who did not receive any performance-related incentives) participated in a randomized control study. The training was found to significantly improve response times to hazards in stimuli from the official hazard perception test used for driver licensing in Queensland, Australia, which is known to predict crash involvement. That is, the training was effective in improving hazard perception skill (Cohen's d = 0.50), even though participants were aware that no one was monitoring the extent to which they engaged in the intervention. Given that the training could, in principle, be deployed at scale with minimal resources (e.g. via any online platform that allows video streaming), the intervention may represent a practical and effective opportunity to improve road safety.