Bayesian modeling of expert elicitation in human reliability analysis: Theoretical foundations and model development

Abstract

Human reliability analysis (HRA) examines human errors or human failure events (HFEs) and their probabilities (Human Error Probabilities, HEPs) under specific task conditions using various models. Given the scarcity of empirical data, expert judgments remain vital in the HRA community. Despite prior efforts to model the expert elicitation process using Bayesian methods, there is still a lack of comprehensive understanding of the overall process and its key components in the Bayesian context, which undermines the effectiveness and applicability of these models. To address this gap, this study starts by investigating the sources of uncertainties in various HRA constructs and elements of the expert elicitation process, aiming to form rigorous definitions in probabilistic terms. Furthermore, practical elicitation approaches are modeled and common misunderstandings about Bayesian modeling in past HRA research are examined and addressed. Based on this foundation, a Bayesian model is developed to integrate the different elements with the HRA constructs, and the probabilistic relationships between them are analyzed and elucidated. Two direct applications of the model are also provided to demonstrate the model’s practical applicability, highlighting its strengths and limitations. The proposed model not only offers a formal approach to rigorously incorporating expert judgments into HRA but also enhances researchers’ and practitioners’ understanding of the uncertainties related to the expert elicitation process, thereby providing a solid theoretical foundation for the development of future Bayesian models.