Collapse repeated information
Repeated values on a resolved entity consolidate into its existing record.
02 / Consolidation
Let memory change with the facts.
A new observation should improve the record. ConceptMem checks incoming information, resolves updates under your policy, and preserves the history behind every change.
In practice
One update. A clear record of what changed.
Example using a single-value field with supersede policy. Other policies can reject a change or hold it for review.
- Previous value
- October 12 · superseded
- New observation
- The launch has moved to October 19.
- Current value
- October 19 · after consolidation
- Provenance
- Source, actor, confidence, timestamp, extraction method
Built into the workflow
Make every update accountable.
Handle disagreements deliberately
Use deterministic updates or opt into confidence-weighted arbitration. Require review for changes that need a human decision.
Explain the outcome
Applied operations, skipped writes, and ontology violations tell you what happened to each proposal.
Keep the evidence behind the answer.
Superseded and retracted facts remain separate from current knowledge. The changelog records mutations with provenance so you can inspect how a record evolved.
Explore current state and historyCOMMON QUESTIONS
Consolidation, answered.
What happens to an observation after it is submitted?
It is validated against the ontology, resolved against the existing entity, and turned into typed operations under the effective policy. The result reports what was applied, what was skipped, and what did not conform.
Why are writes called proposals?
Because an observation never edits stored state directly. The consolidation and policy engines decide whether it supersedes a current value, collapses into an existing one, is rejected, or is held for a human decision.
What happens when two observations disagree?
The effective class and predicate policy decides. The default is deterministic supersede, where the newer value wins. Confidence-weighted arbitration is opt-in, and a review threshold can hold the change instead of applying it.
Where do superseded values go?
They stay in memory with a superseded status and a closed validity interval, separate from current state. The changelog records the mutation with its provenance, so you can see how a record reached its present value.
Is anything lost when duplicates collapse?
No. A repeated value on a resolved entity consolidates into the existing fact rather than creating a second one, and the observation is still recorded with its own provenance.
Make your next update count
Start with one useful memory.
Create a workspace, define a type, and see your first observation become a traceable record.