Dossier 6 — Intervention Doctrine

Purpose: state the conditions under which World Fabric may say what should be built, and the conditions under which it must refuse to say it.

1. Why this dossier exists

Dossiers 1 to 5 describe the world as it is and as it is being changed by projects other people have decided to undertake. They are descriptive by construction.

The question that motivates the observatory, however, is not only descriptive. It is:

Which works would actually improve the capacity of the systems that carry global flows, and on what evidence?

That question is legitimate, and refusing it altogether would be an evasion. But answering it badly is worse than not answering it: a ranked list of desirable infrastructure, published without traceable evidence, is exactly the kind of artefact this project was built to distrust. Such lists already circulate, and their weakness is always the same — they enumerate ambitions rather than relieve identified constraints.

This dossier makes the prescriptive layer possible without letting it contradict the charter.

2. The unit is a bottleneck, never a wish

An intervention is a reasoned proposal to relieve a documented bottleneck in a corridor that has a canonical record.

The order is not negotiable, and it is the whole design:

Corridor record → bottleneck register → intervention

An intervention that cannot name the bottleneck it targets is not a weak intervention. It is not an intervention at all. This single constraint is what separates a research output from a wish list: the work must be earned by a constraint that was established independently, before anyone thought about the remedy.

Interventions are not restricted to concrete. On the border_coordination and digital families the binding constraint is frequently procedural, contractual or informational, and a doctrine that recognised only physical works would systematically misdescribe those systems.

3. A proposed effect is a scenario, permanently

The claimed effect of an intervention carries epistemic_status: scenario, and the schema fixes the value as a constant. It cannot be authored otherwise.

Two consequences follow, and both are deliberate.

Building the work does not convert the claim. When an intervention is implemented, its status becomes implemented. Its effect remains a scenario until new evidence is collected on the corridor metric itself. Construction is evidence that something was built; it is not evidence that the system improved.

A sponsor's projection is not evidence of effect. It may be recorded as expected, attributed to its issuer with its issue date, exactly as Dossier 1 requires of any announced capacity. It never becomes the intervention's own expected effect.

4. Displacement is mandatory

Every intervention must name the constraint that becomes binding once the targeted one stops binding, and explain why.

This field is required by the schema. An intervention with no named successor constraint is treated as incomplete, not as optimal — the absence of a successor is almost always a gap in the analysis rather than a property of the system.

The reasoning is the same one that motivates the whole observatory: widening a segment usually moves the bottleneck. A proposal that does not say where it moves has not been thought through to the point where it could be acted on.

5. No priority score

There is no composite priority index, and none may be stored. This follows directly from the charter's measurement doctrine and ADR-006.

Priority is argued in words, against documented criteria, and selected by a human — the same mechanism Dossier 5 already uses for research priority. The criteria a priority_rationale is expected to engage with:

• how binding the targeted bottleneck is, for which flow families;
• how many segments or functional links the relief would reach;
• whether the displacement is smaller than the relief;
• the strength and independence of the supporting evidence;
• feasibility preconditions — financing, jurisdiction, permitting, operator agreement — stated rather than assumed;
• whether a cheaper coordination or regulatory change would achieve the same relief.

A reader who disagrees with a ranking must be able to see which criterion the argument turned on. A number would hide precisely that.

6. Two tiers, and why the register is not empty

A strict gate — publish nothing until a corridor record is validated — would be defensible, and it would also make the register useless during the entire period when the method is being built. The register therefore has two tiers.

candidate. May be authored before any corridor record is validated. Must carry evidence_sufficiency: insufficient, and is published under an explicit non-citable notice. A candidate is a structured hypothesis about where a constraint lies and what would relieve it. It is the analytical to-do list, and it is honest about being one.

validated. Requires a validated corridor record, at least one evidence reference, and evidence_sufficiency of sufficient or sufficient_for_qualified_publication. The schema enforces all three conditionally on the status value, so the tiers cannot be blurred by an author in a hurry.

The distinction must be visible wherever interventions are displayed. A reader who cannot tell a hypothesis from a finding has been misled, whatever the prose around it says.

7. Falsification

Every intervention states what would show it to be the wrong intervention for its bottleneck. Not what would make it expensive or politically difficult — what would make it incorrect: evidence that the targeted constraint is not binding for the relevant flow families, that the relief is absorbed by the successor constraint, or that an untreated adjacent constraint dominates the effect.

An intervention that cannot fail is not a finding.

8. What this dossier does not authorise

• Ranking works across corridors whose records are not comparable in scope.
• Publishing an intervention whose only evidence is its sponsor's own material.
• Aggregating interventions into a global investment figure. The evidence table does not support that arithmetic, and a total would be read as a measurement.
• Presenting a candidate without its non-citable notice.

9. Current state

No corridor record is validated, therefore no intervention is validated. Any entry in the register today is a candidate, and the register renders as such. Interventions become answerable in Phase 6, once the Middle Corridor pilot has produced a validated record and a bottleneck register.