> ## Documentation Index
> Fetch the complete documentation index at: https://wuweism.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Claim governance

> How Wu-Weism records, structures, and manages every hypothesis and inference you produce.

Every time Wu-Weism produces a hypothesis, observation, or counterfactual finding, it writes a record to the **Claim Ledger** — automatically, without any action on your part. This page explains what that record contains, how to navigate and filter your claims, and how to export them.

## The Claim Ledger

The Claim Ledger is a persistent, append-only log of every claim produced during your sessions. You cannot silently overwrite or discard a claim — you can only reconcile it with newer evidence. This design ensures that your reasoning trail is always auditable, even when your conclusions change.

<Note>
  The Claim Ledger records claims from all three Wu-Weism surfaces: Causal Workbench, Hybrid Synthesis, and Legal Causation. Each claim carries a `source_feature` field so you always know where it originated.
</Note>

To open the Claim Ledger, navigate to **/claims** in the workbench. Claims are shown in reverse chronological order by default.

***

## Claim structure

Each claim in the ledger exposes the following fields.

```json theme={null}
{
  "id": "clm_01j9k2rz4m8xp7qw3n5vb6",
  "claim_text": "Increasing advertising spend by 20% in Q3 would have increased revenue by 8–12%, holding production costs constant.",
  "claim_kind": "counterfactual",
  "status": "active",
  "confidence_score": 0.74,
  "uncertainty_label": "medium",
  "source_feature": "chat",
  "trace_id": "trc_9xm2pq7kl4rz8w1n3v5b",
  "session_id": "ses_4kw9l2pz7m1xqr3n5v8b",
  "evidence_links": [
    { "type": "pdf", "ref": "smith2021_advertising_elasticity.pdf" },
    { "type": "web", "ref": "https://example.com/study" }
  ],
  "gate_decisions": [
    {
      "gateName": "falsifiability",
      "decision": "pass",
      "rationale": "Claim specifies measurable outcome (revenue %) and a testable intervention (ad spend %).",
      "score": 0.91
    },
    {
      "gateName": "identifiability",
      "decision": "pass",
      "rationale": "Causal effect is identifiable given the provided confounder adjustment set.",
      "score": 0.88
    }
  ],
  "created_at": "2026-03-14T11:42:07Z"
}
```

### Field reference

| Field               | Description                                                                                                                                                 |
| ------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `claim_text`        | The full hypothesis, observation, or finding as a natural-language statement.                                                                               |
| `claim_kind`        | One of `hypothesis`, `observation`, or `counterfactual`. Reflects where on Pearl's causal ladder the claim sits.                                            |
| `status`            | `active` means the claim stands. `reconciled` means newer evidence has updated or superseded it.                                                            |
| `confidence_score`  | A value from `0.0` to `1.0` representing the model's confidence in the claim given available evidence and the SCM structure.                                |
| `uncertainty_label` | A human-readable summary of confidence: `low` (score \< 0.4), `medium` (0.4–0.74), or `high` (≥ 0.75).                                                      |
| `source_feature`    | The surface that produced the claim: `chat` (Causal Workbench), `hybrid` (Hybrid Synthesis), or `legal` (Legal Causation).                                  |
| `trace_id`          | Links this claim to the counterfactual trace that produced it. Follow this ID in [Provenance](/governance/provenance) to see the full reasoning chain.      |
| `session_id`        | The workbench session that produced this claim. Use this to group all claims from a single research session.                                                |
| `evidence_links`    | The sources backing the claim — PDFs you uploaded, company data, or web sources extracted during analysis.                                                  |
| `gate_decisions`    | The pass/fail outcomes of the falsifiability, identifiability, and novelty gates for this claim. See [The Falsifiability Gate](/governance/falsifiability). |

***

## Navigating your claims

Open **/claims** in the workbench to see your full claim history. Each row shows the claim text, kind, status, confidence score, and source.

### Filtering by source

Use the **Source** filter to narrow claims by the surface that produced them:

* **Chat** — claims from structured causal dialogue sessions in the Causal Workbench.
* **Hybrid** — claims from Hybrid Synthesis multi-document analysis.
* **Legal** — claims from Legal Causation sessions.

### Filtering by session

If you want to review all claims from a single research session, use the **Session** filter or pass a `sessionId` query parameter to the API (see [Exporting claims](#exporting-claims) below).

### Filtering by status

Toggle the **Status** filter to show only `active` claims, or include `reconciled` ones to see the full history including superseded findings.

***

## Claim reconciliation

When new evidence challenges a previous claim — for example, a new PDF upload contradicts an earlier finding — Wu-Weism can flag the drift.

<Steps>
  <Step title="New evidence is processed">
    You upload a new document or run a new analysis that produces a claim conflicting with an existing `active` claim.
  </Step>

  <Step title="Drift is detected">
    Wu-Weism compares the new claim against existing active claims in the same session. If the new claim materially contradicts an earlier one, both claims are annotated with a drift warning.
  </Step>

  <Step title="You review and reconcile">
    In the Claim Ledger, claims flagged for drift appear with a **Reconcile** prompt. You can mark the older claim as `reconciled`, which preserves it in the ledger but marks it as superseded by the newer finding.
  </Step>

  <Step title="Ledger is updated">
    The reconciled claim retains its original record (including its original confidence score and evidence links) alongside a reference to the claim that superseded it. Nothing is deleted.
  </Step>
</Steps>

<Warning>
  Reconciliation changes a claim's `status` to `reconciled` but never deletes the original record. Your audit trail is always complete.
</Warning>

***

## Exporting claims

Retrieve claims programmatically via the REST API.

```bash theme={null}
GET /api/claims
```

### Query parameters

| Parameter       | Type    | Description                                                   |
| --------------- | ------- | ------------------------------------------------------------- |
| `limit`         | integer | Number of claims to return. Range: 1–100. Default: `20`.      |
| `sourceFeature` | string  | Filter by source: `chat`, `hybrid`, or `legal`.               |
| `sessionId`     | string  | Return only claims from a specific session.                   |
| `traceId`       | string  | Return only claims linked to a specific counterfactual trace. |

### Example request

```bash theme={null}
curl -X GET "https://wuweism.com/api/claims?sourceFeature=hybrid&limit=50" \
  -H "Authorization: Bearer YOUR_API_KEY"
```

### Example response

```json theme={null}
{
  "claims": [
    {
      "id": "clm_01j9k2rz4m8xp7qw3n5vb6",
      "source_feature": "hybrid",
      "claim_kind": "hypothesis",
      "status": "active",
      "confidence_score": 0.82,
      "uncertainty_label": "high",
      "trace_id": "trc_9xm2pq7kl4rz8w1n3v5b",
      "session_id": "ses_4kw9l2pz7m1xqr3n5v8b",
      "created_at": "2026-03-14T11:42:07Z"
    }
  ],
  "total": 1
}
```

<AccordionGroup>
  <Accordion title="What fields does the API return?">
    The `GET /api/claims` endpoint returns: `id`, `source_feature`, `claim_kind`, `status`, `confidence_score`, `uncertainty_label`, `trace_id`, `session_id`, and `created_at`. To retrieve full claim detail including `claim_text`, `evidence_links`, and `gate_decisions`, fetch an individual claim by ID.
  </Accordion>

  <Accordion title="Can I retrieve claims from multiple sessions at once?">
    Yes. Omit the `sessionId` parameter to retrieve claims across all sessions, subject to the `limit`. For large exports, paginate using the `limit` parameter and filter by `created_at` on your side.
  </Accordion>

  <Accordion title="Can I delete a claim via the API?">
    No. The Claim Ledger is append-only. You can reconcile a claim (marking it as `reconciled`) but you cannot delete records. This preserves the integrity of your audit trail.
  </Accordion>

  <Accordion title="Are claims shared across team members?">
    Claims are scoped to your account. Session isolation and team-level sharing depend on your plan configuration. Contact support for workspace sharing options.
  </Accordion>
</AccordionGroup>

***

## Related

<CardGroup cols={2}>
  <Card title="Provenance and audit trails" icon="circle-nodes" href="/governance/provenance">
    Understand how every claim traces back to its SCM model, intervention, and evidence sources.
  </Card>

  <Card title="The Falsifiability Gate" icon="shield-check" href="/governance/falsifiability">
    Learn how Wu-Weism evaluates whether your claims are scientifically valid before recording them.
  </Card>
</CardGroup>