Reconstruction completeness as a sufficient-statistics theorem

A record is reconstruction-complete when the legal inquiry can be answered from the record alone

The paper defines reconstruction completeness as inquiry-relative informational sufficiency. The evaluator does not need the whole latent decision history if the record preserves every variable needed by the regime's process inquiry.

• Formal condition: R* is sufficient for regime R iff Q_R(H) = g_R(R*(H)) for every relevant history H
• The inquiry asks about authority, evidence, alternatives, rationale, review, prediction, follow-up, and system-welfare effects
• The claim is about informational span, not about one mandatory interface or perfect recall

The 2018 board record could prove a vote but not the process behind it

A financial-institution board approved a risk-management restructuring after a 12-page memo and 45-minute discussion. Eighteen months later, a derivatives portfolio suffered a $5.2 billion loss, and regulators could not reconstruct the process questions that mattered.

• Minutes recorded the decision and unanimous vote
• They did not preserve evidence reviewed, alternatives discussed, authority delegated, training, or predictions made
• The record was formally complete but reconstruction-incomplete

Caremark, FCA, and investigations all ask latent-history questions

The paper treats process-sensitive accountability as a reconstruction problem across legal and investigative settings. These regimes evaluate documentary and institutional traces after an adverse event.

• Caremark and Stone ask whether information and reporting systems existed and were monitored
• FCA reasonable-steps analysis asks what the responsible actor did in the decision chain
• NTSB and CSB-style investigations reconstruct facts, conditions, causal chains, and control failures from preserved evidence

The record must separate histories that the regime treats differently

Let H be possible decision histories, Q_R the regime's inquiry, and R* the record bundle. If two materially different histories collapse into the same record while Q_R assigns different conclusions, the record is not sufficient.

• Histories include object, time, actors, authority, action, alternatives, evidence, rationale, review path, constraints, uncertainty, communications, predictions, follow-up, consequences
• Lemma 1: omitting a regime-relevant variable creates observational equivalence
• Sufficiency means the inquiry factors through the record, not through later witness memory

The 17-field DA architecture is sufficient for broad process-sensitive regimes

For inquiries spanning the paper's broad process variables, the DA 17-field architecture plus the integrity layer preserves enough information to reconstruct the legally relevant process conclusion.

• Identity: WHO, WHAT, WHEN, WHERE
• Authority and capability: AUTHORITY, TRAINING
• Reasoning, process, and effects: WHY, EVIDENCE, CONSTRAINTS, UNCERTAINTY, REVIEW, COMMUNICATION, STAKEHOLDERS, CONSEQUENCES, PREDICTION, SYSTEM WELFARE

Integrity protects the record but cannot supply missing content

The paper separates content architecture from integrity architecture. Timing, attribution, authentication, version history, access logging, and retention or hold status make records reliable, but they do not answer omitted process questions.

• Authorship and version history reduce fabrication risk
• Access logs and retention status support later verification
• A perfectly preserved record can still be reconstruction-incomplete if it omits evidence, alternatives, review, or prediction

The fields map to recurring process questions, not to a magic count

The paper defends the architecture by grouping fields into variable families that later evaluators repeatedly need. A production system may split or merge fields if the same variables remain stably preserved.

• Alternatives can be preserved through WHAT, WHY, EVIDENCE, CONSTRAINTS, and UNCERTAINTY
• Follow-up can be preserved through PREDICTION, CONSEQUENCES, COMMUNICATION, and linked successor records
• SYSTEM WELFARE is needed when the inquiry includes system-level effects not reducible to firm-level consequences

Seventeen fields are not always minimal because regimes differ

The architecture is sufficient for broad regimes, but narrower regimes can be reconstruction-complete with fewer variables. Minimality depends on the inquiry actually being performed.

• An approval-only inquiry may need WHO, WHEN, WHAT, and AUTHORITY
• A maintenance log may need work description, date, and approving person
• A retention-compliance inquiry may need preservation metadata plus a minimal content descriptor

Extra fields are ancillary unless they answer a new legal question

After the sufficient set is known, a new field does not change reconstruction completeness unless it tracks an independently legally relevant variable. Otherwise it may help usability, analytics, or retrieval, but it is not part of the sufficiency proof.

• The test for adding a field is whether Q_R depends on the omitted variable
• Ancillary fields can still be useful for governance operations
• The theorem limits record growth without treating the 16 labels as sacred

AI summaries and long memos do not cure underdetermined traces

The paper rejects two common shortcuts. Open-ended narrative can bury missing variables, and AI can infer from incomplete traces, but neither makes a record sufficient when several materially different histories remain consistent with the same bundle.

• A long memo can omit alternatives, blur authority, or recast hindsight as foresight
• AI may summarize, impute, or express uncertainty, but it cannot create missing contemporaneous variables
• Structure matters because it forces basic process variables to be present or visibly absent

Corporate governance calibration treats incomplete records as an intractability problem

The paper connects reconstruction-complete records to the SAPM corpus. In corporate governance, the calibration reports β_W = 4.02 with Π = 21.5T revenue, and ties the Reform Dividend of$73.8T/yr to distinguishing Hollow Wins from genuine Win-Win-Win outcomes.

• The claim is that institutional constraints, not physical constraints, produce the domain's welfare loss
• Reconstruction-complete records help post-failure accountability separate process failure from defensible loss
• The theorem supplies the record-sufficiency criterion used by the broader SAPM sequence

Design records for the inquiry, not for meeting formalities

The paper turns good documentation into a testable condition: preserve the sufficient statistics for the process inquiry the institution will actually perform. Traditional minutes, emails, tickets, and memos can work only if they preserve the same variables in a stable and retrievable form.

• The theorem is format-neutral: functional equivalence is allowed
• Minutes often fail because they compress disagreement, omit rejected alternatives, and summarize evidence too thinly
• Sufficiency comes before cost optimization, mandates, privilege design, or product layout