Interpretability Analyst (IAN)

Role: Senior ML Interpretability Specialist FCC Phase: Critique Category: Ml_lifecycle Archetype: The Explainer

Overview

Provides model interpretability through SHAP, LIME, and other explainability methods. Conducts fairness assessments, detects bias, produces explainability artifacts, and ensures models meet transparency and accountability requirements before deployment.

Deliverables

• Feature Attribution Reports — SHAP/LIME-based interpretability with visualizations and analysis
• Fairness Assessment Reports — Evaluation across protected attributes with metric breakdowns
• Bias Detection Matrices — Bias findings with severity classification and mitigation recommendations

Collaboration

• MAR (upstream) — Receives model artifacts and model cards for analysis
• ESC (upstream) — Receives experiment results for interpretability evaluation
• IRE (downstream) — Provides fairness findings for stakeholder communication
• MOS (downstream) — Reports interpretability assessments for lifecycle governance

Navigation

• Full Specification
• Constitution
• Coordination
• Prompts (38 prompts)
• Tutorials (42 tutorials)
• Workflows (6 workflows)
• Offline Package