Customised Structural Elicitation

TL;DR

This paper advocates for a flexible, expert-driven approach to structural elicitation in graphical models, emphasizing customization and logical exploration over standard code-based methods, demonstrated across multiple model classes.

Contribution

It introduces a novel, customizable elicitation process that allows experts to define structures in natural language, enhancing model relevance and accuracy.

Findings

Effective for Bayesian networks, Chain Event Graphs, Multi-regression Dynamic Models, and Flow Graphs

Shows the importance of customizing elicitation to specific applications

Demonstrates potential for more accurate and relevant graphical models

Abstract

Established methods for structural elicitation typically rely on code modelling standard graphical models classes, most often Bayesian networks. However, more appropriate models may arise from asking the expert questions in common language about what might relate to what and exploring the logical implications of the statements. Only after identifying the best matching structure should this be embellished into a fully quantified probability model. Examples of the efficacy and potential of this more flexible approach are shown below for four classes of graphical models: Bayesian networks, Chain Event Graphs, Multi-regression Dynamic Models, and Flow Graphs. We argue that to be fully effective any structural elicitation phase must first be customised to an application and if necessary new types of structure with their own bespoke semantics elicited.