A New Bayesian Method for Estimating Drivers' Risk Perception Thresholds and Agreement in Simulated Curve Scenarios

Risk perception is a critical topic in cognitive psychology, given the serious consequences that misjudgements in this regard could lead to. It is crucial to develop reliable methods for assessing how individuals perceive risk and whether their evaluations differ from expert standards. In this study, we focused on risk perception in the driving context: fifty participants classified 32 simulated curve scenarios as risky or non-risky, with variations in the level of risk obtained by manipulating Available Sight Distance (ASD) and lateral friction coefficient. We applied a Bayesian procedure (Nucci et al., 2021) to estimate each participant' Belonging Threshold (BT) – the risk level at which a curve is deemed risky. Afterwards, we estimated several interrater agreement indices between the drivers and between the drivers and the objective curve risk evaluations. These indices are more reliable as they rule out spurious agreements and disagreements due to differences in the drivers' BT. Results showed a slight tendency to underestimate the risk level of curve scenarios, but nevertheless a relatively high agreement with the road design formula. This approach allows for identifying drivers who would benefit from a training aimed at standardizing their risk perception thresholds and curve scenarios that are problematic to assess.