UTE is not just “answering queries”. It is also about keeping evidence and surviving change:
new measurements, new assumptions, and conflicting sources.
What exists today
- Proof-real direction: DS19
- Runtime contradiction checks (engine-side consistency checks)
- Explainability surfaces (query traces and proofs)
UTE requirement: provenance is a first-class object
Provenance
Track which theory facts/rules contributed to an answer, which engines ran, and which assumptions were required.
Provenance must be serializable and auditable.
Evidence
Evidence is broader than a proof: it can include measurements, data sources, experimental outputs, and confidence models.
UTE needs a schema for evidence and a way to attach it to beliefs.
Contradictions
Contradictions should produce actionable outputs: minimally inconsistent subsets, “which assumption breaks”, and suggested
revision actions.
Key research question: revision is not merely “delete facts until consistent”.
UTE needs a policy language (or at least a policy module): source trust, recency, causal priority, and domain-specific override rules.
Proposed extensions (research direction)
- Belief objects: assertions with metadata (source, time, confidence, evidence links).
- Revision operators: explicit rewrite/retract/override operations that preserve a history.
- Contradiction reports: return “what conflicts” and “what would fix it”, not only a boolean.
Spec
This page is summarized and formalized in DS34.