Glossary
Updated
Refers to the inherent transparency of a model’s structure (e.g., linear models or decision trees where feature impacts map directly to outputs). Interpretability governance encourages interpretable models for high-risk use cases, documents model logic clearly, and restricts opaque models to lower-risk domains or pairs them with post hoc explanation methods.
Example
A credit-scoring team opts for a decision-tree model for initial loan approvals because each split can be directly interpreted (“income > $50K”). They publish the tree logic to stakeholders, ensuring full interpretability and facilitating regulatory reviews.