Skip to content

Body-Change Summary Invalidation

Invalidation rule — instantiates Summary-Substance Alignment Audit

Treats any material edit to the substance as automatically making the summary stale — invalid until it is re-derived and re-approved.

A Body-Change Summary Invalidation rule enforces a single coupling: the moment the substance changes materially, its summary is presumed wrong until proven current. It detects nothing about the summary's wording and fixes nothing; it makes staleness the default state on any material edit, flipping the summary to invalid and gating it from being shown as current until an owner re-derives and re-approves it. A materiality threshold decides which edits trip it — a typo or reformat passes, a changed number, scope, or conclusion does not — so the rule couples the two surfaces at exactly the edits that would make a reader act on outdated information.

Example

A company keeps its data-processing policy as a full legal body plus a plain-language "summary for teams." Legal edits clause 7 to narrow which vendors may receive customer data. The edit crosses the materiality threshold — a scope change, not a wording tweak — so the instant it lands, the coupling rule flips the attached team summary to "stale — pending re-derivation" and blocks it from being surfaced as the current policy. Nobody had to notice the summary went out of date; the moment the substance moved, its summary was presumed wrong. It stays gated until a policy owner rewrites the summary against the new clause 7 and re-approves it — at which point the rule clears the flag, tied to that approval and nothing else.

How it works

  • Watch the substance for edits.
  • Classify each edit against the materiality threshold — cosmetic (passes) versus material change to a number, scope, claim, or conclusion (trips).
  • On a material edit, presume the summary stale — set its status to invalid and gate its "current" display.
  • Hold that status until an approved re-derivation clears it — the flag is released by re-approval, not by dismissal.

It is a presumption rule, not a diff viewer or a propagator: its entire content is the enforced link material change ⇒ summary invalid until re-approved.

Tuning parameters

  • Threshold height — what counts as material: every edit (safe, noisy, training people to ignore the flag) versus only conclusion, number, and scope changes (cheaper, risks a borderline edit slipping).
  • Invalidation granularity — flag the whole summary, or only the section coupled to the edited passage.
  • Enforcement strength — a soft "may be stale" banner versus a hard block on presenting the summary as current.
  • Clear condition — what releases the flag; tying it to an approved re-derivation is what stops the alarm being clicked away.
  • Coupling map — which summaries are bound to which body sections, so an edit invalidates only what it actually affects.

When it helps, and when it misleads

Its strength is closing the drift that opens whenever a body is updated and its summary is forgotten — it makes staleness loud and automatic instead of silent and discovered later. The analogy is cache invalidation: a summary is a cached copy of its substance, and knowing when the cache has gone stale is proverbially one of the genuinely hard problems.[1] Its failure modes are threshold-shaped: set too low, it invalidates constantly and people learn to ignore the flag; set too high, real drift slips through unflagged. The classic misuse is clearing the stale flag to unblock publishing without actually re-deriving the summary — dismissing the alarm rather than answering it. The discipline is to tie the clear to an approved re-derivation, never to a click.

How it implements the components

This rule fills the coupling-and-trigger side of the audit:

  • update_coupling_rule — the enforced link binding a material substance edit to summary invalidation; this rule is the mechanism.
  • material_divergence_threshold — the cutoff that separates edits which trip invalidation from cosmetic ones it lets pass.

It only fires the staleness trigger. It does not push the resulting correction out to dependent surfaces (Correction Synchronization Workflow, and Social-Preview Cache Invalidation for machine caches), nor judge the *content of the divergence (the review and check mechanisms). It shares the divergence threshold with Material-Divergence Red Team, which probes that same cutoff adversarially rather than enforcing it on edits.*

Notes

This rule is the trigger; the Correction Synchronization Workflow is the response. Keeping them separate is what lets invalidation stay dumb and reliable — it fires on every material edit without deciding who fixes what — while the workflow handles the messy human coordination of actually reconciling the copies. Its close sibling Summary-Diff Review inspects what changed; this rule only cares that something material did.

References

[1] Borrowed from computing, cache invalidation — knowing when a cached copy has gone stale and must be refreshed from its source — is proverbially one of the hard problems; a summary is a cache of its substance, and the same difficulty is why summaries quietly outlive their bodies. ↩