Understanding non-motorists' views on automated vehicle safety through Bayesian network analysis and latent dirichlet allocation

Abstract

Automated vehicles (AVs) hold great promise for creating a safer, more efficient, more equitable, and more sustainable transportation system. However, the rapid adoption of AVs requires a thorough understanding in their coexistence with the human environment in the current roadway network, particularly with respect to interactions between AVs and non-motorists. Bike Pittsburgh (BikePGH) conducted a 2019 survey to examine non-motorists' perceptions of AV safety. Using Bayesian network (BN) analysis, the study identified key factors such as safety perception, AV technology knowledge, and real-world interaction experiences that influence non-motorists' overall perception of AV safety using BikePGH survey data. The study also explored several counterfactual scenarios to gain insights into non-motorists' viewpoints on AV safety. Notably, the study found that the differences in the ways of AVs and human-driven vehicles interacted with non-motorists at intersections played a crucial role in shaping survey participants' opinions. By taking into account the key insights identified in this study, policymakers can develop evidence-based strategies to achieve sustainable urban mobility goals while ensuring the safety and well-being of all road users, particularly non-motorists.