Engineering Psychological Safety in Autonomous Vehicles: A Systems-Theoretic Framework for Psychological Safety in Autonomous Vehicles and its Validation in Real-World Scenarios

While traditional autonomous vehicle safety engineering focuses heavily on physical harm prevention, societal adoption remains constrained by psychological barriers such as anxiety, confusion, and loss of trust. Although previous research introduced the conceptual AV-PsySafe framework based on STAMP (Systems-Theoretic Accident Model and Processes), it remained largely theoretical and untested in industrial workflows. This paper addresses this gap by enhancing the framework into an engineering-ready methodology and evaluating its practical applicability through an independent validation study in a real-world autonomous driving project.

The authors operationalized the AV-PsySafe framework by introducing explicit methodological guidance, standardized analysis worksheets, and end-to-end traceability mechanisms. The framework centers on a psychological safety risk model that breaks down human stakes into four components: trust, perceived safety, predictability, and perceived support. It employs a hazard analysis process called Psy-STPA, which guides analysts through system boundary definition, control structure modeling, identification of psychological Unsafe Control Actions, and loss scenario mapping. Additionally, the methodology integrates the Psychological Safety Integrity Level (PsySIL) matrix, allowing practitioners to categorize psychological risks from level A to D based on severity, exposure, and controllability.

To evaluate the industrial feasibility of the methodology, the authors implemented a structured validation strategy where an independent Validation Team—composed of practitioners experienced in physical safety engineering—applied the framework to realistic autonomous driving scenarios. The evaluation combined the technical analysis outputs generated by the team with structured feedback regarding usability and consistency. The results showed that independent analysts could successfully and consistently apply the framework to produce actionable insights into psychological risks, proving that psychological safety can be co-assessed alongside physical safety.