Designing for Human Systems: Integrating Organizational Psychology into Systems-Theoretic Safety Analysis for Extreme Teams

Abstract

As the complexity of high-risk domains such as space exploration and energy production grows, safety analysis must account for cognitive, organizational, and team-based factors that shape human performance. System-Theoretic Process Analysis is a leading method for identifying unsafe control actions and loss scenarios in sociotechnical systems, yet its representation of complex human teams in extreme environments is limited. This thesis investigates whether STPA with the Engineering for Humans extension can capture important decision-making dynamics of a land-based drilling rig crew. The study draws on seven semi-structured interviews, 127 excerpts coded into 215 tags, and structured surveys over a 14-day work rotation with 32 of 42 responses. Three team-level themes emerged: performance demands, shared mental models, and hierarchical challenges. Applying STPA to bottom hole assembly handling produced four representative unsafe control actions and two detailed loss scenarios that link operator cognition and context to hazardous states. To address identified gaps, the thesis proposes an enhanced control structure that explicitly represents distributed cognition and cross-level information flows. This visualization layer revealed additional loss mechanisms related to delayed feedback, cognitive load, and cross-boundary misalignment, and supported three actionable mitigations in the case context. The findings show that STPA offers value beyond traditional methods, while indicating opportunities to adapt it to meaningfully represent distributed cognition and social dynamics in extreme teams, without altering core STPA steps. This work also contributes actionable insights for the case context examined here, as well as other high-risk extreme domains facing similar challenges. Disclosure: The author takes full responsibility for the content provided in this thesis. While the ideas and writing are the author’s own, an AI-based tool was utilized to develop and refine the material.