Skip to main content
The graph is the set of all claims and the relationships between them: supports[] links (what this claim rests on) and contradicts[] links (what this claim is in tension with).
Trust is derived from this structure, and read off two derived axes: Status per content_id (the state of the answer) and question_status per frame_id (the state of the question), both computed on every read from graph.proposition_status(prop). The more independent paths converge on the same upstream evidence, the stronger a proposition’s status. A single agent asserting the same claim a thousand times stays at the bottom of the scale: independence of provenance paths is what matters. The stored support_level ladder (PRELIMINARY / REPLICATED / ESTABLISHED) shown above is the legacy per-claim axis; its REPLICATED / ESTABLISHED public labels are deprecated for v0.4.0.

​
supports[]

supports[] is the provenance chain. It records what a claim rests on: upstream claim_ids or reference strings (e.g. DOIs).
Leaving supports[] empty for a DERIVED claim makes the chain unverifiable. For ANALYTICAL and INFERRED claims, supports[] is optional but records what the finding is explicitly grounded in.

​
contradicts[]

contradicts[] documents explicit tension. When a new finding is in conflict with an existing one, both coexist in the graph; neither is silently overwritten.
Science advances by documented contestation. ESTABLISHED means human-validated evidence, not settled truth.

​
Reading a chain of origins

A conclusion is only as grounded as the weakest step behind it. A short chain of ANALYTICAL steps close to raw data is more trustworthy than a long chain of INFERRED steps, even if each step looks locally valid. One INFERRED step means the chain as a whole rests on inference: origin does not average, and a reader should take the weakest link.
This is a way to read a chain, not a label Mareforma computes for you. The graph does not collapse a chain to a single origin. It records each claim’s origin at assertion time, permanently and unchangeably, so the weakest link is visible when you walk the supports[] edges.

​
Graph fragmentation

The most common failure mode: two agents assert the same finding in different words without linking to a common upstream in supports[]. REPLICATED never fires because the graph cannot detect convergence without a shared anchor. The mitigation is the shared anchor itself: both agents cite the same ESTABLISHED upstream in supports[] and sign with distinct keys, and REPLICATED fires. Sharing an idempotency_key is not an alternative. It raises rather than merging two authors into one row, since the merge would discard the second contribution; see the API reference assert_claim → Idempotency.