Co-designing and building an expert-elicited non-parametric Bayesian network model: demonstrating a methodology using...

Area of Science:

• Ecology
• Risk Assessment
• Computational Biology

Background:

• Probabilistic models, especially Bayesian networks (BNs), are crucial for risk-based decision-making.
• Balancing model complexity and development ease presents ongoing challenges.
• Codesign and nonparametric Bayesian networks (NPBNs) offer promising solutions for capturing complex relationships.

Purpose of the Study:

• To describe and demonstrate the codesign, building, quantification, and validation of an NPBN for emerging risks.
• To apply this methodology to a case study involving the marine pathogen Bonamia ostreae.
• To highlight the modeling and quantification process, including methodological challenges and solutions.

Main Methods:

• Utilized a codesign approach with experts for developing the BN structure.
• Employed semistructured workshops with extensive expert feedback.
• Quantified the model using field data and structured expert judgment (SEJ) via the IDEA protocol.

Main Results:

• Successfully developed and validated an NPBN model for assessing risks associated with marine pathogen spread.
• Elicited over 100 parameters using a hybrid (remote and face-to-face) IDEA protocol for SEJ.
• Addressed and documented methodological challenges encountered during the modeling and quantification process.

Conclusions:

• Codesigning NPBNs with expert input is an effective strategy for risk assessment.
• The methodology provides a robust framework for managing emerging risks, as demonstrated by the Bonamia ostreae case study.
• This approach facilitates informed decision-making in complex ecological and biological systems.