A revised framework for the assessment of psychological safety in autonomous vehicles

Despite rapid technological advancements in autonomous vehicles (AVs), their societal adoption faces significant roadblocks due to psychological concerns rather than just physical safety risks. Passengers frequently experience psychological discomfort, anxiety, and a lack of trust when interacting with automated driving systems. However, traditional automotive safety standards focus exclusively on physical hazards and lack the formal mechanisms, severity scales, and guidelines needed to assess and mitigate psychological risks. This paper addresses this critical gap by extending a theoretical framework for AV psychological safety risk assessment.

The authors develop an extended risk model that consolidates human-AV psychological experiences into four core components: trust, perceived control, predictability, and perceived support. By integrating these components with objective measurement techniques, the framework enables a pragmatic estimation of psychological loss severity. Furthermore, the model incorporates contextual elements such as the internal vehicle state, external environmental events, and the human driver's baseline psychological state to comprehensively define psychological hazards.

To operationalize the risk model, the authors adapt the Systems-Theoretic Accident Model and Processes (STAMP) and Systems-Theoretic Process Analysis (STPA) frameworks. Unlike classical deterministic safety approaches that treat safety strictly as component failures, STAMP views safety as a dynamic control problem arising from complex interactions. The adapted methodology establishes complete traceability between psychological safety goals, unsafe control actions, responsibilities, and psychological loss scenarios, offering systems designers a repeatable process for hazard analysis.