The Gold Standard · Experimental Design
Make the decision path visible
Independent approval is valuable because it can say no, but independence also creates honest disagreement. One reviewer may accept a bounded replication while another sees purpose drift, incomplete logs or an export path that reaches current decisions. A tie cannot be resolved by letting the old exception continue.
The gold standard classifies the conflict, preserves expired status, identifies what evidence could narrow disagreement and escalates the remaining risk decision to a named authority. Conditional reauthorisation is used only when its safeguards are testable and enforceable.
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 · FRAME THE CONFLICT
1. Keep authority suspended during disagreement
When an exception expires, execution and export remain disabled until a valid decision is recorded. Reviewer conflict is not an implicit extension.
Preserve evidence-only access where justified so reviewers can inspect the historical environment without recreating operational authority.
CHAPTER 2 OF 12 · FRAME THE CONFLICT
2. Classify what the reviewers disagree about
Conflicts may concern missing facts, evidence quality, scope interpretation, control effectiveness, consequence or risk appetite. Record each claim and the evidence cited.
A technical gap may be resolved by testing; a value trade-off belongs to the authority accountable for the decision.
Evidence check
Check whether the scope, owner and decision consequence are explicit before continuing.
CHAPTER 3 OF 12 · FRAME THE CONFLICT
3. Establish roles and conflict rules in advance
Name the first-line owner, primary second-line reviewer, specialist challengers and final escalation authority. Define recusals, quorum, dissent and decision deadlines.
Clear rules prevent applicants from shopping for the easiest reviewer or treating silence as consent.
| Class | Resolution |
|---|---|
| Missing fact | Obtain evidence |
| Method dispute | Run test |
| Scope dispute | Interpret condition |
| Risk choice | Escalate authority |
CHAPTER 4 OF 12 · CHALLENGE THE CASE
4. Build a shared evidence record
Use the same manifest, purpose statement, run history, access list, export map, incidents, validation results and proposed conditions. Version later additions and keep reviewer annotations distinct from facts.
A common record makes differences visible. Parallel private packs can create disagreement that is really inconsistent evidence.
CHAPTER 5 OF 12 · CHALLENGE THE CASE
5. Ask what evidence would change each position
For every objection, state a test, missing document or boundary demonstration that would address it. Identify objections that would remain even if the test passes.
This separates resolvable uncertainty from a genuine decision about whether the residual risk is acceptable.
Evidence check
Ask a second reviewer to reconstruct the example from the preserved evidence and stated assumptions.
CHAPTER 6 OF 12 · CHALLENGE THE CASE
6. Worked example: reviewers split over a monthly export
An exception began as on-demand historical replication. The owner requests renewal with a monthly chart sent to a shared planning folder. One reviewer accepts because the code and weights are reproducible; another rejects because the new destination lets current teams rely on superseded analysis.
The exception remains suspended. Reviewers agree that historical runs may occur in isolation, while current export requires a different authorised model. Governance conditionally reauthorises on-demand replication only. Changed-context check: if the chart is deposited in an access-controlled archive clearly labelled non-authoritative and no current workflow consumes it, export risk may be acceptable after the boundary is tested.
| Element | Evidence |
|---|---|
| Purpose | Bounded request |
| Manifest | Coherent environment |
| Run history | Actual use |
| Export map | Downstream reach |
CHAPTER 7 OF 12 · CONDITION OR ESCALATE
7. Write conditions that answer the objection
Conditions may limit datasets, users, frequency, outputs, destinations and duration, while requiring labels, logs, dual approval or isolation. Each needs an owner and breach response.
A condition is useful only if the system or review process can observe it. Aspirational wording does not reduce risk.
CHAPTER 8 OF 12 · CONDITION OR ESCALATE
8. Escalate the unresolved decision, not the whole history
Give the escalation authority the shared facts, competing interpretations, consequence ranges, options and dissent. Ask for a bounded decision about the disputed authority.
Do not ask senior governance to repeat technical validation. Make the remaining choice clear enough to own.
Evidence check
Verify that the control still reaches the intended decision under a changed but plausible context.
CHAPTER 9 OF 12 · CONDITION OR ESCALATE
9. Record a reasoned outcome and effective time
State approve, condition, reject or retire; identify version, purpose, evidence, dissent, expiry and when authority changes. Notify repositories and decision gates.
A decision is incomplete until executable systems and users receive the same authoritative status.
| Control | Observed by |
|---|---|
| Dataset limit | Manifest gate |
| User limit | Access role |
| Destination limit | Export block |
| Expiry | Automatic suspension |
CHAPTER 10 OF 12 · MONITOR THE OUTCOME
10. Monitor both conditions and renewed conflict
Review actual runs, exports, downstream references and control failures. Reopen approval when facts change or a condition proves impractical.
Repeated disagreement can reveal unclear policy or a control boundary that needs redesign, not merely difficult reviewers.
CHAPTER 11 OF 12 · MONITOR THE OUTCOME
11. State the limits of escalation
A senior decision cannot create missing evidence or make an unenforceable condition effective. Consensus can still be wrong, and logs can miss informal use.
Preserve limitations and a route for later evidence to reopen the record. Authority makes the decision accountable, not infallible.
Evidence check
Confirm that versions, notices and unresolved limits remain visible in the final record.
CHAPTER 12 OF 12 · MONITOR THE OUTCOME
12. Audit conflicts as governance evidence
Track disagreement types, time in suspension, added tests, conditional outcomes, breaches, escalations, reversals and policy changes.
The gold standard protects independence without allowing paralysis or silent continuation. Conflict becomes a documented route to better evidence and accountable authority.
| Measure | Warning |
|---|---|
| Reviewer shopping | Independence weak |
| Long suspension | Decision stalled |
| Repeated condition breach | Scope unworkable |
| Frequent reversal | Policy unclear |
Frequently asked questions
Does reviewer disagreement allow the old exception to continue?
No. Expired authority stays suspended until a valid reauthorisation decision is recorded.
Must every disagreement go to senior governance?
No. Resolve factual and testable disputes first; escalate only the remaining interpretation or risk choice.
When is conditional reauthorisation appropriate?
When the permitted use is bounded and every material safeguard can be observed, owned and enforced.
Should dissent remain in the record?
Yes. It explains residual uncertainty and gives later reviewers a clear basis for reopening the decision.
Continue through verified routes
Core routes
Experimental Design subject root · Gold Standard of Education · Gold Standard directory · How X Works hub
Specialist routes
- Exception reauthorisation conditions, purpose drift and second-line approval
- Exception register expiry, orphaned owners and reauthorisation evidence
- Supersession compliance exceptions, historical replication and preventing exception creep
- Mixed-finding detection, decision-gate rejection and supersession compliance