Constructive Diagnosis

I. Reports that do not repair, reforms that do not diagnose

Much public political analysis collapses into one of three sterile forms.

The sterile pundit form describes dysfunction. The system is broken. Politics is dysfunctional. Trust has collapsed. Polarization has won. This produces salience. It does not produce repair, because it does not name what is missing.

The sterile think-tank form identifies a real failure and stops before the intervention geometry. The output is a 200-page report with a menu of policy options, recommendations for “stronger oversight” or “better coordination,” and a request for further study. This produces analysis. It does not produce a repair specification, because the recommendations are compatible with infinitely many institutional shapes, most of which would not address the named failure.

The sterile consultant form produces motion without owning the diagnosis. Dashboards, key performance indicators, reorganizations, stakeholder workshops, implementation roadmaps. This produces action-shaped theatre. It does not produce repair, because the diagnostic backbone connecting the motion to the underlying failure is absent or imported from the client’s own priors.

The substitution is reliable enough that an alert reader can usually predict the failure mode of a public reform before it begins. A reform announced with no diagnostic backbone will produce motion. A diagnosis announced with no repair geometry will produce reports. A description of dysfunction with no missing primitive named will produce salience and then silence.

The constructive third move is different.

A diagnosis is constructive when it compiles into a repair specification.

“Constructive” here means construction-constraining: the diagnosis specifies the shape of the repair the failure admits, not just that the failure exists.

II. The repair specification

A repair specification is the minimum output a constructive diagnosis owes. It has six fields.

1. Failure mechanism. The specific loop, asymmetry, or division of labor that, while it persists, generates the harm by routine operation. Named in terms of what currently exists; “the system is broken” and “trust is low” are themes, not mechanisms.
2. Missing primitive. The nameable institutional capability that, if added, would interrupt the failure mechanism. “More transparency” and “better coordination” do not name primitives.
3. Owner. Who must hold the primitive? An identifiable institutional seat, real or proposed, that can be named today.
4. Trigger. When must the primitive activate? A condition under which the owner is obligated to produce, decide, or contest. Without a trigger, even the right primitive remains advisory.
5. Wrong-repair warning. What tempting intervention would worsen or merely preserve the failure? Institutions default to repairs that preserve their current power. A diagnosis must name and rule out this attractive wrong shape in advance.
6. Movement test. A specific observable behavior or output that would not be present under the failure regime. “Improved outcomes” is too vague to falsify the model.

This is not a call for “actionable recommendations.” Every report already claims to have those. The standard is stricter. If the diagnosis says “trust is low,” the repair space is infinite — fund a campaign, train officials, restructure media, change leadership, write better speeches. If it says “no actor owns the harm after the permit is issued,” the repair space narrows immediately: assignment, route, deadline, trigger, review. The first is a theme. The second is a repair specification.

This is engineering discipline applied to institutional design. The repair specification adds a constraint that ordinary engineering does not need: the diagnosis itself must identify the primitive whose absence keeps the failure stable. Engineering domains have legibility and feedback that allow trial-and-error from a vague problem statement. Governance domains do not. In governance, the absence of fast feedback means the diagnosis is the only place where constraint propagation can happen, because by the time the institution is built and tested, the political moment has passed.

A constructive diagnosis does not produce the final policy. It narrows the repair class. The repair class still has to survive politics, capture, and degradation — those are downstream problems. But without the repair class being named, the downstream problems are not even visible.

III. CBO: the diagnosis selects the institution

The 1973–1974 impoundment crisis produced one of the cleanest examples of constructive diagnosis at the level of constitutional architecture.

The failure mechanism was an asymmetric information regime. The executive branch’s Office of Management and Budget produced the only authoritative fiscal numbers in Washington. Congress could only argue against those numbers with ad-hoc committee staff and partisan accounting, ensuring the executive’s framing carried more institutional weight in any fiscal dispute regardless of who was substantively right. When President Nixon impounded appropriated funds, the asymmetry made Congress structurally incapable of contesting the executive’s fiscal narrative on shared evidence.

That mechanism dictated the shape of the repair almost entirely. The missing primitive was a permanent source of objective, impartial budget and economic information serving Congress rather than the executive; the owner had to be a body inside Congress with permanent staff and statutory standing; the trigger was the budget cycle; the wrong-repair warning was “demand more transparency from OMB” or “create a partisan counter-shop” — both would have left the structural asymmetry intact. The movement test was that Congress could publicly score and contest executive fiscal claims with its own institutional source on shared analytical ground.

The Congressional Budget and Impoundment Control Act of 1974 created the Congressional Budget Office. The repair shape matches the diagnosis at every field. CBO has survived for half a century because the diagnosis was structurally correct and a legitimacy substrate happened to be available; both conditions had to hold, but the first was the necessary one.

IV. NTSB: the diagnosis specifies the boundaries

A subtler example: the National Transportation Safety Board, and the structural insistence that it have no enforcement power.

The failure mechanism is a recursive capture loop. When the same body that operates a transportation system also investigates its own catastrophic failures, the investigative function is exposed to the operator’s liability incentives, status protection, and regulatory blowback. Each of those forces pushes any investigation toward conclusions that protect the operator. The capture is structural, not bad-faith; an investigator embedded in or accountable to the operator faces a coherence pressure that distorts findings in a predictable direction.

The missing primitive is investigative authority structurally insulated from the consequences of investigation. The owner had to be an independent body. The trigger had to be any major transportation incident. The wrong-repair warning was crucial: an investigator with enforcement power would reintroduce prosecutorial and liability pressures into the investigation, recreating the very capture loop the diagnosis named. The movement test was that published investigation reports would be specific and structural enough that operational organizations and regulators could not publicly ignore them.

The NTSB describes itself as an independent federal agency that investigates transportation accidents, determines probable causes, and issues safety recommendations; it has subpoena and investigative authority but no regulatory power and no punitive power. That boundary is not a political compromise; it is the diagnostic specification. The NTSB design keeps the forensic function separate from regulatory enforcement, which is what allows the safety recommendations to pressure operational and regulatory change without the agency becoming the regulator itself. The repair shape specifies what the institution must not have, not just what it must have. That is constraint propagation working correctly.

V. NEPA: an incomplete specification predictably absorbed

A repair specification can be partially correct and still fail in a way the framework anticipates ex ante. The US National Environmental Policy Act (NEPA) is the case.

NEPA named a defensible failure mechanism: a closed-loop agency decision process in which costs external to the deciding agency’s primary mandate were structurally zero-weighted, so each agency’s project pipeline ran without internalizing the environmental consequences of the projects it advanced. The Act specified a recognizably constructive primitive — a public, contestable environmental impact statement (EIS) before action — and a trigger (major federal action significantly affecting environmental quality).

The Owner field, however, instantiated the failure the spec was meant to repair. NEPA assigned ownership to the lead agency: the same body responsible for advancing the action whose consequences it was now obligated to document. That is structurally identical to the recursive capture loop §IV identified as a Field 1 failure for transportation safety — operator investigating itself. Filling the Owner field with a name does not satisfy the field; the field requires an owner exempt from the loop the failure mechanism describes. By the standard’s own rules, NEPA’s Owner field was not validly filled.

That made field 5 (the wrong-repair warning) and field 6 (the movement test) doubly necessary, and both were absent. NEPA did not name the absorption pattern in which the document’s existence becomes a substitute for the modeling it represented; it did not impose a procedural bound on length, time, boilerplate, or structure; it did not require non-trivial adversarial-review funding that would keep the document compact and contested; it did not specify what observation would distinguish a real consequence model from compliance documentation. The conflicted owner could satisfy the formal duty by producing defensible documentation rather than a live consequence model, and the spec did not name this attractive wrong shape in advance.

Under this standard, a 3-of-6 repair specification (mechanism, primitive, trigger correct; owner structurally invalid; warnings and movement test absent) has a predictable absorption surface. The Council on Environmental Quality’s own data show final EISs averaging 661 pages and 447 pages by median in 2013–2018, well above the original 150–300-page expectation; recent CEQ data give a median of 2.2 years from notice of intent to final EIS. The instrument now functions partly as litigation defense and compliance documentation rather than as agile model-update infrastructure. Power’s Audit Society names the general pattern: when verification becomes mandated, capture answers with rituals. The ritualization is exactly the failure mode that a non-conflicted Owner plus a present field 5 plus a present field 6 would have anticipated.

NEPA is better read not as a right-shaped repair that later degraded, but as a 3-of-6 repair specification with a structurally invalid owner that degraded through its missing fields. The framework anticipates this in advance: a repair spec assigning the primitive to a conflicted owner and missing fields 5 and 6 will absorb into compliance documentation; identify it before the institution is built.

VI. The Statistics Finland case: applying the spec to a corpus case

The Statistics Finland 2022 reclassification of state-subsidized housing loans into the EDP public-debt aggregate added about six percentage points to Finland’s headline debt ratio overnight. Every actor was acting within role; the classification was procedurally legitimate; no institution owned the consequence model. The Legitimacy Came Before Cognition essay treats this as the legitimacy/cognition gap. Read through the repair specification, the same case becomes a worked example.

• Failure mechanism. A division of labor in which the statistical arbiter (Statistics Finland) is decoupled from the fiscal optimizer (ministries, fiscal council, parliament), forcing downstream macro-policy to absorb raw technical classifications without consequence-mediation. The decoupling is functional and largely correct as a descriptive principle; the failure mechanism is the unmediated absorption pathway by which technical classifications with macro-policy effects propagate into political and fiscal decision-making without any seat producing a consequence model along the way.
• Missing primitive. A cognition organ with the standing to produce, publish, and defend consequence models when a technical decision’s modeled effect crosses a defined macro-policy threshold.
• Owner. A protected mechanism-cognition operator with authority to publish consequence models for technical decisions crossing macro-policy thresholds. Not Statistics Finland, whose descriptive independence must be protected; not the ministry proposing or absorbing the policy, which has a conflict of interest; not the existing fiscal council unless its mandate is expanded.
• Trigger. A technical decision whose modeled effect crosses a defined threshold (debt ratio, policy instrument disablement, fiscal-rule exposure).
• Wrong-repair warning. Demand consequence-weighting from Statistics Finland (which would corrupt the descriptive function); demand more independent statistics offices (we already have one); demand a parliamentary debate (the legitimacy stack already provides this and it was used and produced no consequence model).
• Movement test. A published consequence model is produced, contested, and attached to the decision before it locks in; opposition staffers can cite specific predicted consequences against the decision in real time.

The legitimacy essay names the function. This essay produces the repair specification. The Mechanism Authority work spells out one institutional design downstream of the spec. Each piece carries a different field of the standard.

VII. Wrong-repair warnings, in the grammar of bad equilibria

The wrong-repair warning is the field most commonly omitted, and the field that most often determines whether a constructive diagnosis survives contact with institutions. Bad Equilibria Are Not One Thing sorts the failure modes; the wrong-repair warning attaches the typical defensive intervention each failure mode invites.

• Destination failure (no institution receives the harm; the ownerless / telos-gap case). The attractive wrong repair is to forward the signal to more inboxes — more reports, more dashboards, more notifications — leaving the assignment unsolved. The wrong-repair warning says: the address is missing, not the visibility.
• Payoff failure (the Campbell–Goodhart case: a measure becomes a target and distorts behavior). The attractive wrong repair is to add more metrics. The wrong-repair warning says: more metrics multiply the gaming surface unless the metric is redesigned, audited, and tied back to the underlying target through evidence the optimizer cannot cheaply manipulate.
• Execution failure (the institution lacks staff, tools, authority, or implementation path to do the thing it is asked to do). The attractive wrong repair is punishment: fines, censure, leadership change. The wrong-repair warning says: punishment redistributes blame, not capability; the repair must build the capacity that is absent.
• Bonded failure (epistemic, status, material, and coordination axes hold the failure in place together). The attractive wrong repair is awareness campaigns inside the captured cluster. The wrong-repair warning says: the cluster will absorb awareness without changing behavior; repair must decouple the axes or build alternative substrate.
• Absorption pattern (formal success is preserved by breaking an unowned variable; bad-equilibria treats this as a cross-cutting audit move rather than a primitive family). The attractive wrong repair is to celebrate compliance with the formal target. The wrong-repair warning says: the formal compliance is the absorption mechanism; the movement test must be on the substantive variable, not the formal one.

Institutions default to repairs that preserve their current power. A diagnosis must name and rule out this attractive wrong shape in advance.

The wrong-repair warning is the field that anticipates that choice and refuses it before the institution gets to make it.

VIII. The repair specification as a strategic weapon

A constructive diagnosis selects the repair shape. It does not guarantee political survival. The CBO survived because the impoundment crisis produced an intra-elite institutional threat, technocratic mediation was available, and a legislative vehicle was passing. The Office of Technology Assessment, which had a right-shaped diagnosis behind it, was defunded in 1995 because its constituency was diffuse and its faction-specific indispensability was thin. A repair specification is necessary for repair; it is not sufficient against political capture, defunding, mission drift, or the absorption pattern documented in §V.

Constructive diagnosis is preparatory infrastructure for political windows. Windows open suddenly and close fast. When intra-elite leverage produces an opening for institutional change, the side that arrived with a pre-compiled repair specification gets the institution; the side that did not gets consultants writing the spec on contract during the window’s narrow duration, with the result shaped by veto players rather than by the diagnostic. CBO emerged because the impoundment crisis exposed a concrete asymmetry, and the repair shape matched that asymmetry closely enough to legislate. With only the wedge: ad-hoc committee staff. With only the diagnosis: another generation of think-tank reports. The window does not stay open long enough to write the spec from scratch, and the spec written under window pressure is the spec the strongest faction in the room dictates.

The diagnostic act has to commit before the politics opens. The specification is the strategic resource the moment requires. Without it, the moment is captured.

IX. The corpus standard

The same standard reads backwards across the corpus. Legitimacy Came Before Cognition names the failure mechanism and the missing primitive class. Bad Equilibria Are Not One Thing supplies the wrong-repair warnings vocabulary. The Mechanism Authority spells out one institutional spec for the missing primitive. Together they supply the components from which a complete repair specification is built. If a kunnas.com essay does not contribute at least one field of a repair specification, it has failed its own standard. That is the falsifiability anchor for the corpus.

X. Close

Constructive diagnosis is a debt the diagnostic act owes to the repair act. The debt is met when the diagnosis names the failure mechanism in terms of what currently exists, identifies the primitive whose absence keeps the mechanism stable, names the owner who could hold it, specifies the trigger that would activate it, warns against the attractive wrong shape, and proposes the observation that would show the system moved.

A diagnosis that meets the debt is a precondition for repair, not a guarantee of it. Most repair-side political analysis fails to meet the debt and produces salience, reports, or motion. The third move is rarer because it is harder, and it is harder because it requires the diagnostic act to commit to the structural shape of the repair before the politics opens.

A repair-side diagnosis has not finished until it names the primitive that would repair it.

Related:

• Legitimacy Came Before Cognition — the historical setup: why a cognition layer never integrated
• The Legitimacy Gate — the applied piece: signals need procedural attachment to bind
• Powerless Intelligence — the AI-era variant: receivers exist, response duty does not
• Bad Equilibria Are Not One Thing — the taxonomy this essay’s wrong-repair warnings map to
• The Fourth Branch — the institutional seat for cross-domain consequence cognition
• Mechanism Realism — the foundational ontology the corpus runs on