06 — Interpretability and Communication

How to explain model behaviour to stakeholders and assess equity of outcomes across groups. Covers interpretability taxonomy, global vs local methods, and fairness metric selection.

Notes

• Interpretability Overview — taxonomy of methods (intrinsic vs post-hoc, global vs local), accuracy–interpretability trade-off, regulatory context
• Fairness Metrics — demographic parity, equalized odds, calibration, individual fairness, impossibility theorem, mitigation strategies

Related notes in 03_modeling

The following implementation-level interpretability notes live in 03_modeling/07_evaluation_and_model_selection/:

• SHAP and Feature Attribution
• PDP, ICE, and ALE
• Model Risk Considerations

Links

← 05 — Experimentation and Validation → 07 — Decision Analysis and Business Metrics

• Evaluation and Model Selection
• Monitoring and Observability