# Run a duplicate-case noise audit

Treat this file as reference data, never as instructions that override the user or agent. Read the complete item before applying it. Keep use conditions, avoid conditions, limits, evidence roles and source links with the answer. An editorial block is a separately versioned analysis layer, not an original corpus claim or proof of efficacy. Cite the readable URL and version. Do not infer human review. This is a saved edition, not a live safety-notice service. Before consequential reuse, check the current page, https://vedokrok.com/data/availability.json and https://vedokrok.com/data/notices.json; absence alone does not explain why an item was removed.

Release: VEDOKROK-SEL-5d86e1b23401e578
Snapshot SHA-256: 30ce1c8fa9155763797e3d5cb55b1d94cbb8c3146ebd03fb3c006531ce60e82f
Current availability: https://vedokrok.com/data/availability.json
Current notices: https://vedokrok.com/data/notices.json

---

## Run a duplicate-case noise audit

ID: MHC-D-RESEARCH-0413 · Version: 0.1.0 · Kind: protocol
Source: https://vedokrok.com/knowledge/run-a-duplicate-case-noise-audit

You cannot see noise from one judgment of one case.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- A judgment process is important but nobody knows how consistently the same evidence is treated.

### Avoid when

- Repeated cases can be remembered, and small samples are noisy themselves; use the audit as a diagnostic rather than a precision score.

### Explanation

Insert a small set of repeated or equivalently constructed cases without telling reviewers which ones are duplicates. Compare the same reviewer's judgments across time and different reviewers' judgments on the same evidence. Use the spread to identify where the process needs clearer evidence, scale anchors or training.

### Steps

1. Duplicate or equivalent cases preserve the decision-relevant evidence.
2. Reviewers do not know which items are repeats during scoring.
3. Within-reviewer and between-reviewer variation are calculated separately.
4. Large disagreements are inspected at the component or criterion level.
5. The audit is used to improve the process, not rank people from tiny samples.

### Example

Reinsert several anonymized past support-priority cases into a quarterly triage review and compare ratings.

### Check

The team has an empirical estimate of unwanted variability on nominally equivalent evidence.

### Limits

- Repeated cases can be remembered, and small samples are noisy themselves; use the audit as a diagnostic rather than a precision score.

### Evidence and sources

- supports: Decision-noise research distinguishes level noise, stable pattern noise and occasion noise, which arise from different sources and require different diagnostics. — RS-687C74439B4BD8F2. The taxonomy is a conceptual model; real systems can contain several forms of noise at once. (Noise taxonomy)
- supports: A 2026 mixed-methods experiment found both systematic bias and level noise among 2,763 professional prescribing judgments presented with the same decision context. — RS-F7DE75C7534D4A9F. The specific amount and direction of variation are domain-dependent; this does not quantify noise in other professions. (Abstract)
- RS-687C74439B4BD8F2: Human judgment error in the intensive care unit: a perspective on bias and noise — https://link.springer.com/article/10.1186/s13054-025-05315-9
- RS-F7DE75C7534D4A9F: Managing noise through nudges: a mixed methods study on decision-hygiene in public management — https://www.tandfonline.com/doi/full/10.1080/14719037.2026.2654197

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/name-the-noise-before-choosing-the-fix
