SUMMARY: A search result can identify records worth comparing without establishing that they describe the same person. Keep that distinction visible when a proposed relationship begins influencing a service decision.
Separate Retrieval From Confirmation
In The Shared Name, a fictional station retrieves a candidate by name and later presents the joined history as settled. The problem is not merely that a match can be wrong. The process loses the status of the evidence as the result moves between screens.
Define which operation retrieves candidates, which establishes a link and which authorizes consequences. A person inspecting the final record should be able to identify where confirmation occurred, or see clearly that it has not.
Do Not Turn Missing Information Into Agreement
Two blank fields do not provide two independent pieces of confirming evidence. A matching rule may specify how missing values affect its result, but a reviewer should not interpret absence as proof that two histories agree.
Likewise, disagreement needs context. An address may have changed; a transcription may be wrong. Neither automatic merging nor automatic separation is a universal answer. State what the available evidence supports and what remains unresolved.
Keep Consequences Proportionate
A candidate useful for a staff search may be insufficient for a restriction. Decide which actions require stronger verification and which should wait. Do not repair an uncertain match by routinely demanding every document a person can produce. Identify the specific unresolved question and the minimum appropriate evidence needed to investigate it.
Make a Link Correctable
Preserve source records and record the basis, status and owner of a link. When a relationship is disputed, identify downstream copies or decisions that rely on it. Removing a link from one display does not establish that every consequence has been reconsidered.
Use Test a Record Link for a synthetic walkthrough and the accountability guide for the wider correction process. This is a design-review framework, not an identity-verification standard or a recommendation to collect additional personal data.