alibi-explainability

Alibi Explain provides explanation algorithms answering: "How do predictions change with feature inputs? Which features matter for a prediction? What minimal changes would alter a prediction? How does each feature contribute to predictions?" per the [Alibi Explain docs].

When to use

• Compliance / regulator inquiry: "explain this denied loan prediction" - generate counterfactual + per-feature attribution.
• Model debugging: a single instance got the wrong answer; explain why.
• High-risk system audit (EU AI Act Annex III): every prediction ships with a stored explanation record.

Step 1 - Install