Practitioner playbooks¶

These playbooks assemble the shared foundations that practitioners reach for when promoting calibrated explanations into production pipelines.

Core Capability Playbooks¶

Mondrian (Conditional) Calibration Playbook – Conditional calibration for fairness-aware deployments, heterogeneous data, and domain-specific groupings.
Ensured Explanations Playbook – Reduce epistemic uncertainty in alternative explanations for high-stakes decisions.
Decision Policies Guide – Convert uncertainty outputs into systematic accept/reject/defer decision policies.

Integrate with scikit-learn pipelines – Batch and realtime integration recipes, including configuration notes for the WrapCalibratedExplainer.
Export explanations – Persistence patterns for dashboards, audits, and downstream analytics.

Satisfying EU AI Act Requirements with calibrated_explanations – Article-by-article mapping of EU AI Act (Regulation (EU) 2024/1689) obligations to calibrated_explanations capabilities, including a compliance checklist, limitation analysis, and a quick-start integration recipe.
Satisfying EU Data, Liability, and Algorithmic Regulations with calibrated_explanations – Companion compliance reference covering GDPR, the Product Liability Directive (2024/2853), the proposed AI Liability Directive, the Digital Services Act (2022/2065), and Equal Treatment Directives – with code patterns, a cross-regulation checklist, and a gaps analysis.

Release checklist – Calibrated explanations release gates and documentation updates.
Documentation section owners – Ownership map for runtime and documentation components.