Multi-Engine Causal Relationship Finder
Overview
The Multi-Engine Causal Relationship Finder is a Hellixia tool that detects causal relationships among a set of selected nodes by using several independent evaluators (engines) simultaneously. It is the more sophisticated counterpart of the Causal Relationship Finder, designed to produce more robust results by aggregating evidence from multiple sources.
Function
The tool is accessed via Menus > Hellixia > Multi-Engine Causal Relationship Finder. It works on nodes you have already defined (like the Causal Relationship Finder) but differs in its internal evaluation process.
Each Causal Engine independently evaluates every possible causal relationship between the selected nodes. For each candidate relationship, every engine returns two values:
- Direct Causal Effect: A signed strength in the range -1 (inhibiting) to +1 (promoting), indicating how strongly the cause influences the effect.
- Confidence Level: A certainty score reflecting how confident that engine is in the identified relationship.
Aggregation and Arc Selection
The tool then aggregates the results from all engines:
- For each candidate arc, the sum of confidence levels across all engines is computed. This sum becomes the arc weight.
- All candidate arcs are ranked from the highest arc weight to the lowest.
- Arcs are added to the network incrementally from highest weight to lowest, but any arc that would create a cycle is skipped. This incremental cycle-free arc addition guarantees the resulting network remains a Directed Acyclic Graph (DAG), which is required by BayesiaLab.
Output
The result is a Constraint-Free Propositional Causal Network, i.e., a fully-specified Bayesian network where every node is a Boolean proposition. The arcs carry signed direct effects, and the Conditional Probability Tables are written using the DualNoisyOr() function.