Research Brief · When Correction Cannot Catch the System

A correction was accepted. Does it still reach what depends on the old state?

Micheal Charles Preble · Independent Researcher · Release manuscript, September 23, 2026

The Question

When is human corrective authority over consequential personal state actually effective — not just recognized — across the processes where the superseded state continues to matter?

The Problem

AI governance increasingly recognizes correction, deletion, contestation, and revocation. But recognizing an intervention at its point of origin does not establish that it remains effective once personal state has been copied, summarized, inferred from, exported to another system, or already used to initiate a consequential action. A correction can be validly accepted while superseded state continues to shape materially relevant processes elsewhere.

The Contribution

Separates correction validity (the corrective act is recognized at its source) from correction effectiveness (materially consequential reliance on the superseded state is also governed). Develops temporal adequacy as the governance test for that gap, and four possible outcomes after correction — invalidation (stop future reliance), revalidation (ask if a dependent result remains justified), retirement (historical state remains stored without operating as current authority), and remedy (address a consequence already produced).

How the Framework Works

The formal test: a system preserves temporally adequate human authority only when an accepted corrective act can be traced to identified materially consequential reliance paths and, within an interval proportionate to the risk of continued reliance, brings paths within the operator's governance boundary into an appropriate governed state. “Corrective latency” is the interval between a correction becoming authoritative and a reliance path reaching that governed state — explicitly not database-replication time. The paper grounds this in a bounded legal analogue: under the Fair Credit Reporting Act, a furnisher's internal knowledge of a correction does not necessarily satisfy its statutory duty once notice-and-investigation triggers arise (citing Saunders v. BB&T, Gorman v. Wolpoff, and Seamans v. Temple), while Frazier v. Dovenmuehle and Sprague v. Salisbury Bank mark the limits of that duty.

Why It Matters

A system does not preserve meaningful corrective authority merely because it allows a person to change the source. It preserves that authority only when the change can still govern the consequences that depend on it. This reframes evaluation of correction mechanisms away from the interface (does a delete button exist?) toward causal reach (does the deletion actually govern what relied on the deleted state?).

Evidence Status

Evidence status: Conceptual governance analysis. It reports no empirical result. The related experimental program is in design.

How evidence status works across this site →

What This Does Not Claim

  • Correction is not deletion.
  • States no universal descendant-update obligation.
  • States no universal synchronization mandate.
  • Does not require automatic reversal of all downstream decisions.
  • Temporal adequacy is not database replication time; corrective latency is an observed interval, not a predictive equation.
  • Does not claim propagation, temporal validity, persistence, or contestability are newly discovered problems — it names six neighboring literatures (machine unlearning, temporal-authorship scholarship, authorization propagation, capability-coherence analysis, privacy policy, consumer-reporting law) it does not claim to replace.
  • Does not depend on Operant Dyad's stronger Relationship State claims — this paper requires only that computational systems can maintain, derive, transfer, and act upon personal state over time.

Open Questions

The related experimental program that would test corrective latency empirically is still in design (see the site’s Research page for the current empirical priority). Whether specific technical mechanisms can reliably achieve temporal adequacy at scale is left open by design — the paper does not prescribe one.

Read / Cite the Research

Read the canonical paper record → · Full paper PDF · Its public SSRN record is not yet available.

Suggested citation
Preble, Micheal Charles. “When Correction Cannot Catch the System: Temporal Adequacy of Human Authority Over Consequential Personal State” Release manuscript, September 23, 2026. SSRN 7514120, submitted to SSRN, 2026.