Bayesian networks as a decision making tool to plan and assess maritime safety management indicators

Today, maritime safety management norms, self-assessment guides and frameworks demand and/or recommend the collection, report, and analysis of indicators to measure the safety performance of shipping companies. However, the characteristic of classic indicators only provide information about the specific evaluated activity. In this paper, a new quantitative and qualitative option to jointly analyze the performance of individual and collective indicators of a maritime safety management system is proposed. For this purpose, the dependencies between the quality of the most representative components of maritime safety management and their designated indicators levels are probabilistically estimated using a Bayesian network model and two expert views. Each component has one or more designated indicators which aim to identify practical values for the performance of those components. Based on the findings of this study, the implementation of the Bayesian network model seem to provide a unique decision support tool to plan and set indicators, and also to evaluate the indicators' performance and the effect on their designated components. Furthermore, the use of the indicators in the model enable detecting their repercussion on other components of an evaluated safety management system, even when those components do not seem to be directly related.