Modelling transboundary consequences and trade-offs

Published: 16 November 2016
Last edited: 27 January 2020
Through workshops and conference calls, the core team develops a concise influence diagram that represents the key hypothesized relationships between the possible actions, external factors, and ultimate objectives. The coaches use this diagram as a conceptual basis when developing a Bayesian decision network, which allows for assigning stakeholder values and probabilities within the influence diagram. The Bayesian decision network therefore provides a visualization of the quantitative decision model. Within another workshop setting that includes the 8 representative stakeholders and up to 2 experts, the coaches ask each participant to individually provide numerical inputs for the model. There are two types of questions for the elicitation on a scale from 0 to 100%: 1) percent chance that a given external factor or ultimate objective will follow a particular trajectory while accounting other external factors and allocation options; 2) percent satisfaction with each possible combination of outcomes for the three ultimate objectives. During a following discussion, stakeholders agree on set of predictions and satisfaction scores to represent the averages among participants in the decision analysis.

Classifications

Category
Collection of baseline and monitoring data and knowledge
Technical interventions and infrastructure
Scale of implementation
Subnational
Multi-national
Phase of solution
Documentation and dissemination of results

Enabling factors

Face-to-face interactions among core team members are essential for developing and filling in the decision model, considering that many participants are not accustomed to modeling. Reducing categories per variable in the Bayesian decision network to 2-3 ensures that the analysis is feasible. Conducting the analysis requires expertise in workshop facilitation, elicitation of quantitative inputs from stakeholders, multi-criteria decision analysis, and Bayesian belief networks.

Lessons learned

For transparency it is useful to have two versions of the influence diagram: a comprehensive one representing all hypothesized relationships and a concise one representing only the relationships with a high degree of uncertainty and relevance to the decision. To ensure understanding of the elicitation, coaches should provide participants background information and a written guide for providing their independent inputs for the analysis. It is essential that participants provide their inputs individually to avoid a subset of participants driving the outcome of the analysis. The coaches should inform participants that the model inputs only represent perspectives of participants at the workshop and that a forthcoming sensitivity analysis can guide future modeling and estimation work. Participants are more motivated to provide quantitative inputs for the BDN when they are informed that it provides a visual and quantitative justification for how the recommended decision is determined.