Causal Structural Priors
Overview
The Causal Structural Priors tool is a Hellixia feature that you apply to arcs that are already present in a network — typically after you have performed Machine Learning to learn the structure from data. Its purpose is to analyze each selected relationship, determine whether it is causal, define an appropriate Structural Prior, and provide an explanation for its judgment.
This tool operationalizes the From Arcs to Causal Structural Priors workflow. In that workflow, you start with a data-driven structure (learned by BayesiaLab’s algorithms), then select arcs and ask Hellixia to vet them for causality. The resulting priors can then steer a subsequent round of Machine Learning to refine the network, adding causal interpretation to what was originally an associative, data-derived structure.
When you use the tool, and the corresponding option is checked, Hellixia supplies a Causal Explanation, i.e., a natural-language justification for the proposed cause-effect direction. This makes the LLM-derived judgment auditable, letting you decide whether the direction is plausible before committing it to the next round of learning.
The Causal Structural Priors tool is part of Hellixia’s broader LLM-Enhanced Network Documentation toolkit, which also includes features like Assess Causality in Learned Networks, Import Arcs (SP), Verbalize Relationships, and Define Node Concepts. Together, these tools bridge statistical learning and causal inference by adding semantic meaning and causal direction to networks learned purely from data.
Workflow Walkthroughs
- Workflow 1Export the Structural Prior Dictionary, then re-import it as an Arc Dictionary to apply the causal directions
- Workflow 2Apply the Causal Structural Priors directly during structure learning, without the export/import step