Explainability

Kind: quality-attribute

Layer: Correctness Core

Record: architecture:explainability

Severity: contextual

Scope: model, model-backed system, decision flow

Canonical: Ontology

The degree to which a model's decision comes with the evidence or attribution a reviewer needs to understand it.

Listed in Architecture principles, after Knowledge Graphs and before Model Safety.

Requires

• Traceability
• Rationale/Evidence

Reinforces

• Governance
• Trust

Enables

• Audit and Debugging

Conflicts with

• Opaque Black-Box Decisions

In tension with

• Model Complexity
• Artificial Intelligence Architecture
• Artificial Intelligence Architecture / Explainability

Tensions

• Explainability / Model Complexity

Violated by

• Opaque Black-Box Decisions

Refactored by

• Evidence Citation

Severity

• contextual

Category

• Model Architecture

Reinforced by

• Retrieval-Augmented Generation
• Knowledge Graphs
• Agentic Architecture

Linked from

• Systems built around a model
• Model Architecture
• Observability / Auditability / Traceability
• Security / Privacy / Compliance / Governance
• Model Architecture
• Severity levels
• The resolutions