AI Explainability
The extent to which the internal mechanics of an AI system can be understood and interpreted by humans.
Definition
The suite of methods (surrogate models, feature-attribution, counterfactuals) and processes (documentation, user-friendly dashboards) that make AI decisions transparent, so stakeholders can understand, contest, and trust outcomes.
Real-World Example
A credit-scoring model uses SHAP to highlight which financial factors (e.g., “low credit history”) influenced a denial. Loan officers review these explanations alongside the AI’s recommendation, allowing applicants to correct inaccuracies and request human review.
.svg)
