The safety culture of so-called high-reliability organizations (HROs) encompasses values, routines, and work processes that allow an organization to prevent mistakes and quickly bounce back if unexpected events occur. It is said to provide a model for improving organizational resilience in the offshore oil and gas industry, where small errors can grow into accidents with devastating environmental, social, and economic impacts. To date, such a transfer of successful practices is impeded by a lack of system perspective that would allow researchers and practitioners to fully understand the safety dynamics in HROs, adjust them to the unique setting of offshore oil and gas, plan safety interventions, and anticipate the direct and indirect effects of these interventions.

In this dissertation, I developed and rigorously tested a model of how safety interventions impact interdependent aspects of HROs' characteristics, based on peer-reviewed research, an industry workshop, and a survey of offshore oil and gas industry practitioners. This approach combines the qualitative research method of Thematic Analysis (TA), Thematic Network (TN), Fuzzy Cognitive Maps (FCM) modeling and simulation, and Exploratory Modelling and Analysis (EMA). Furthermore, I developed thematic proximity as a measure for determining edges' weights in FCMs that are based on research texts, thus reducing the need for including subject matter experts in modeling studies.

This work makes several contributions: on a theoretical level, it shows the inherent dynamics of HROs and points to several limitations in existing High Reliability Organizations Theory (HROT) as well as uncertainties regarding the efficacy of some safety interventions. On a practical level, it provides a planning tool for safety decision-makers that can also serve as the foundation of future safety culture training. Finally, it makes several contributions to FCM methodology, namely a model architecture that combines knowledge from the literature with that of human experts, the introduction of thematic proximity coefficient, and the adaptation of model testing strategies from the literature on Systems Dynamics.