SHAP (Shapley Additive Explanations) is the leading, industry-recognized method for explaining complex AI model predictions.
AAIA specifically identifies SHAP as a robust technique for transparency in high-impact systems such as medical AI.
SHAP provides:
* Feature contribution breakdowns
* Visual explanation of each decision
* Global and local interpretability
* Quantifiable insights into model logic
* Compliance-friendly justification of predictions
Options A and B relate to integration and stress testing, not transparency.
Option C validates outputs but does not explain how the model reached them.
Thus, SHAP is the best approach for achieving explainability in medical AI.
References:
AAIA Domain 5: Explainability Techniques
AAIA Domain 3: Model Validation and Auditability