The Gold Standard · Archaeology
Make the decision path visible
A redesigned bridge model may be allowed back only in shadow mode: it can generate comparisons, but it cannot authorise joins or releases. Over time, reviewers may begin consulting its suggestions, difficult cases may disappear from the denominator, or a temporary exception may connect its output to a live workflow. A nominally shadow system then gains practical influence before it has earned authority.
The gold standard measures effective use, not labels. It preserves the appeal record, contains affected joins, reconstructs the true validation population and re-revokes the redesigned route when breached conditions make continued evidence gathering unsafe or misleading.
Choose the route closest to your decision
Chapter contents
Chapters 1–3
Chapters 4–6
Chapters 7–9
Chapters 10–12
CHAPTER 1 OF 12 · READ THE CONDITIONS
1. Translate re-entry conditions into tests
List the permitted population, prohibited release effects, validation denominator, reviewer separation, duration and stopping triggers. Attach an evidence source and owner to each condition.
A condition that cannot be observed becomes a hopeful instruction rather than a control.
CHAPTER 2 OF 12 · READ THE CONDITIONS
2. Define shadow authority precisely
State where the redesign may read data, where results may be displayed and which decisions must remain independent. Prohibit copied scores, annotations or queue priorities from steering release.
A result can influence archaeology even when it never writes the final catalogue field.
Evidence check
Check whether the scope, owner and decision consequence are explicit before continuing.
CHAPTER 3 OF 12 · READ THE CONDITIONS
3. Set breach severity and immediate action
Classify breaches by affected contexts, scientific consequence, duration and reversibility. Predefine when to pause, quarantine or re-revoke.
This prevents commercial or schedule pressure from redefining the response after a breach occurs.
| Condition | Observable evidence |
|---|---|
| Shadow only | No release effect |
| Full population | Denominator reconciliation |
| Independent review | Decision locked first |
| Limited term | Expiry enforced |
CHAPTER 4 OF 12 · DETECT SHADOW DRIFT
4. Measure the real validation denominator
Reconcile all eligible cases with cases actually scored, reviewed and resolved. Examine exclusions, timeouts, missing features and cases sent directly to manual review.
Performance among easy completed cases cannot stand for the full archaeological population.
CHAPTER 5 OF 12 · DETECT SHADOW DRIFT
5. Trace indirect influence on reviewers
Inspect screen placement, copied notes, queue order, meeting records and disagreement resolution. Interview reviewers about when a shadow suggestion changed their attention or conclusion.
Effective authority often appears first as convenience rather than a formal permission.
Evidence check
Ask a second reviewer to reconstruct the example from the preserved evidence and stated assumptions.
CHAPTER 6 OF 12 · DETECT SHADOW DRIFT
6. Worked example: difficult joins disappear from shadow evidence
A redesign is authorised for shadow validation only. Its suggestions become visible beside reviewer records, while joins with incomplete geometry time out and are excluded from the reported denominator. Reviewers start using the visible score to prioritise releases, breaching both independence and coverage conditions.
Governance freezes affected releases, reconstructs the full denominator and re-revokes the shadow route while preserving its logs for redesign. Changed-context check: if an independent classification is displayed only after the reviewer records a locked decision, it can support disagreement analysis without influencing the original judgement.
| Signal | Risk |
|---|---|
| Score copied | Indirect authority |
| Timeout excluded | Biased denominator |
| Queue reordered | Attention steering |
| Gate connected | Direct release effect |
CHAPTER 7 OF 12 · DECIDE RE-REVOCATION
7. Assess archaeological consequence
Identify released joins, assemblage interpretations, publications and downstream datasets exposed to shadow influence. Re-examine high-consequence and disagreement cases using primary context.
A condition breach is not only a governance event; it may change the evidential status of archaeological claims.
CHAPTER 8 OF 12 · DECIDE RE-REVOCATION
8. Distinguish implementation error from design failure
Test whether access configuration, workflow presentation, capacity or model behaviour caused the breach. Examine shared dependencies with the previously revoked route.
A narrow configuration repair may be possible, but repeated or structural breaches weaken the basis for continued re-entry.
Evidence check
Verify that the control still reaches the intended decision under a changed but plausible context.
CHAPTER 9 OF 12 · DECIDE RE-REVOCATION
9. Make a scoped re-revocation decision
State which redesign version, population and uses lose permission, the effective time and the status of affected outputs. Keep unrelated manual or independently valid routes available.
Re-revocation should remove unsafe authority without stopping legitimate archaeological work that does not depend on it.
| Use | Question |
|---|---|
| Catalogue | Was state changed? |
| Assemblage | Was grouping affected? |
| Publication | Was claim relied on? |
| Partner data | Was output propagated? |
CHAPTER 10 OF 12 · PRESERVE THE NEXT ROUTE
10. Preserve a clean evidence route
Retain raw inputs, shadow outputs, reviewer decisions, exclusions and timestamps in a controlled record. If redesign continues, require a new plan and new approval rather than informal resumption.
Failure evidence is valuable when it cannot leak back into current authority.
CHAPTER 11 OF 12 · PRESERVE THE NEXT ROUTE
11. State limits and misconceptions
Reconstructed denominators may still miss undocumented manual cases, and interviews cannot prove every instance of reviewer influence. Re-revocation also does not show that every model output was scientifically wrong.
Record uncertainty separately from the authority decision and its protective scope.
Evidence check
Confirm that versions, notices and unresolved limits remain visible in the final record.
CHAPTER 12 OF 12 · PRESERVE THE NEXT ROUTE
12. Audit the appeal-to-revocation chain
Connect original revocation, appeal outcome, re-entry conditions, observed breaches, affected cases and the re-revocation decision. Assign surveillance for any surviving route.
The gold standard allows redesign evidence to develop without letting shadow validation become undeclared authority. When conditions fail, the response remains proportionate, traceable and scientifically useful.
| Field | Purpose |
|---|---|
| Version and scope | Bounds decision |
| Effective time | Bounds exposure |
| Affected outputs | Directs review |
| Future route | Requires new approval |
Frequently asked questions
Can a shadow model influence work without release permission?
Yes. Visible scores, queue order and copied annotations can steer reviewers even when the final field is locked.
Does every condition breach require full revocation?
No. Response should follow consequence, recurrence and whether a reliable bounded repair exists.
Why reconstruct excluded cases?
The missing cases may be exactly where the redesign is least reliable and where archaeological consequence is greatest.
Can redesign resume after re-revocation?
Only through a new recorded plan that addresses the breach, defines evidence and receives explicit authority.
Continue through verified routes
Core routes
Archaeology subject root · Gold Standard of Education · Gold Standard directory · How X Works hub
Specialist routes
- Conditional approval revocation appeal, residual release holds and redesign re-entry criteria
- Second-suspension recovery deadlock, evidence escalation and conditional approval revocation
- Renewed conditional monitoring drift, restoration-condition breaches and second suspension
- Control redesign validation disputes, competing success criteria and governance acceptance