You are here:
A Modular Approach to Risk Assessment with Structural Models
Citation:
Carriger, John F. A Modular Approach to Risk Assessment with Structural Models. SETAC NA 44th Annual Meeting, Louisville, KY, November 12 - 16, 2023.
Impact/Purpose:
This presentation provides an overview of applying structural models, especially causal structural models, to environmental risk assessment. An approach is described for developing and maintaining a library of reusable sub-models to facilitate construction of complex models to specific risk assessment applications.
Description:
Structural models based on Bayesian networks are powerful knowledge representation tools that can capture diverse problem analysis structures and provide a basis for quantitative analysis and weight of evidence. They have only recently been developed and applied in fields ranging from law to medicine to product safety. They have also been applied to environmental risk assessment and their uptake is increasing. Conceptual models in risk assessment are frequently used to capture the causal information in a risk problem. Bayesian networks provide an additional means for including uncertainty in the relationships of the variables in the conceptual models through conditional probabilities. Their development with a conceptual model also allows additional variables that can influence the fate and transport of chemicals to be included, including intervention decisions, and modifying factors. This presentation will overview past applications of Bayesian network and probabilistic model fragments for organizational risk management and demonstrate how these tools can facilitate advanced modeling and more useful assessment and management models. An approach using structural causal models will be emphasized for building a library of risk model structures for re-use and adaptation for individual risk problems in an organizational context. Various model types will be discussed but causality will be emphasized for capturing key understanding and components of the risk problem.