KB article
Semantic Contracts: Setting Expectations for Questions and Answers
Semantic contracts define what questions are valid and how answers should be interpreted.
arf-kbai-readiness-interoperabilitysemantic-contractdeterministic-querymetadata-density
TL;DR
- Contracts reduce ambiguity and misunderstanding.
- They define scope, units, and acceptable queries.
The problem
- Users ask questions outside the model’s intended scope.
- AI answers without clear constraints.
Why it matters
- Contracts prevent misuse and reduce errors.
- They make evaluation possible.
Symptoms
- AI answers unsupported questions.
- Stakeholders misinterpret results.
Root causes
- No explicit scope for metrics.
- Lack of documentation for valid queries.
What good looks like
- Defined set of valid questions per KPI.
- Clear scope, units, and exclusions.
How to fix
- Create semantic contracts for top KPIs.
- Publish and enforce valid question patterns.
- Tie contracts to metadata.
Pitfalls
- Contracts too broad to be useful.
- No enforcement or education.
Checklist
- Contracts documented.
- Valid question list published.
- Contracts reviewed with stakeholders.
Framework placement
Primary ARF layer: AI Readiness & Interoperability. Diagnostic bridge: data-movement-reliability, semantic-reliability, execution-reliability, change-reliability.