Preserve Source Disagreement About an Event Time

Preserve conflicting event-time claims until their meanings and evidence can be reconciled. Choosing the earliest timestamp can make information seem available before it reached the source. Choosing the latest can hide real prior publication. We would keep each source's assertion and declare the policy used to admit the event into an agent feature.

X's data dictionary identifies created_at as a post's creation time. A linked document may describe a separate event time. A post timestamp therefore establishes the social record's clock, while the underlying announcement needs its own evidence.

A five-minute disagreement affects one decision

Assume a fictional issuer release. A news vendor labels its event time 10:00 UTC. An issuer page displays 10:05 UTC. A social link is created at 10:06 UTC. A research decision is evaluated at 10:03 UTC. For this fixture, local receipt records are absent and all three observations were acquired later.

Source	Claimed time	What the fixture can establish
News vendor	10:00	Vendor's event-time assertion
Issuer page	10:05	Displayed issuer time
Social post	10:06	Creation of the linking post

The vendor assertion precedes the decision and the issuer time follows it. The social time follows both. These values alone cannot prove what the decision collector knew at 10:03. The appropriate historical availability label remains unresolved unless another record supplies that evidence.

First ask whether the sources describe the same clock

One source may label scheduled release time, another first publication, and a third an update. Preserve the field name and source explanation before calculating a disagreement. Converting all three to UTC solves representation differences; it leaves semantic differences intact.

Seek a primary publication artifact, a source revision record, or a captured response with receipt evidence. Keep the resolved interpretation linked to the original claims. If the difference remains unresolved, expose an interval or competing timestamps rather than averaging them into a false precision point.

The agent-facing summary can say that the event is supported while its precise publication time is uncertain. That separates confidence in the fact from confidence in its timing. A strategy whose decision depends on whether the release preceded a cutoff should receive that uncertainty directly.

Run the ambiguous cases as sensitivity rows

A proposed audit would evaluate the 10:03 decision once under the early-time assumption and once under the late-time assumption, keeping all other inputs fixed. Report how many decisions change eligibility under the competing interpretations. Their outcomes can be shown as sensitivity results with the source uncertainty attached.

Our event and receipt time note defines local availability when it is captured. Our failed and missing-turn denominator note helps preserve ambiguous cases in reporting. The narrower contribution here is a disagreement record and an admission-sensitivity procedure. It supports a bounded reconstruction, without inventing an exact source history or attributing market behavior to a disputed minute.