A dynamic human-factor risk model to analyze safety in sociotechnical systems

The performance of sociotechnical elements varies owing to a wide range of endogenous and exogenous influencing factors. These are called uncoupled variability as per Safety-II. The uncoupled variability has drawn rare attention, despite its vital importance in major accidents analysis as per Safety-I and Safety-II paradigms. Accordingly, as the first attempt, this study proposes a systematic model to analyze performance variability in human, organizational, and technology-oriented functions caused by various variability shaping factors (VSFs). The model contains three main phases. First, a FRAM (Functional Resonance Analysis Method) - driven Human-Organization-Technology Taxonomy is developed. Subsequently, Dempster - Shafer Evidence theory is employed to elicit knowledge under epistemic uncertainty. The proposed causation model is integrated into Dynamic Bayesian Networks to support decision-making under aleatory uncertainty. Finally, a criticality matrix is developed to evaluate the performance of the system functions to support decision-making. The proposed model is built considering the advanced canonical probabilistic approaches (e.g., Noisy Max and Leaky models) that address the critical challenges of incomplete and imprecise data. The proposed dynamic model would help better understand, analyze, and improve the safety performance of complex sociotechnical systems.