[Feature] Build the institutional digital twin from governed projections #43

New Issue

2026-08-05T16:21:20+02:00

zemion commented

2026-08-05 16:21:20 +02:00

Outcome

Authorized users can understand the institution as an explainable, time-aware projection of units, functions, mandates, services, cases, decisions, records, resources, risks, controls, workflows, and evidence without creating a second source of truth.

Scope

Versioned projection contracts and stable typed references to owning modules.
Valid-time and recorded-time graph queries with current authorization and purpose constraints.
Provenance, freshness, quality, uncertainty, policy, retention, and source-authority explanations on every projection.
Current-state, historical, planned/scenario, and evidence views kept semantically distinct.
Continuous-assurance indicators, process/conformance analysis, bottleneck/risk views, and governed drill-through.
Federated projections that disclose only profile-approved, locally authorized facts.
AI/assistance only over cited, policy-filtered projections; no hidden institutional memory or automatic authoritative mutation.

First proof

For one reference journey, show responsibility, open work, governing policy, source data, decisions, records, control evidence, and temporal changes in one projection, with drill-through to the owning product surfaces.

Acceptance criteria

Projection rebuild and correction are deterministic and do not mutate owner data.
Every displayed edge identifies owner, source, effective/recorded interval, and authorization explanation.
Scenario state cannot be confused with current or historical fact.
A reference package includes quality tests and a documented institutional question it can answer.

Architecture: docs/INSTITUTIONAL_DIGITAL_TWIN.md.

## Outcome Authorized users can understand the institution as an explainable, time-aware projection of units, functions, mandates, services, cases, decisions, records, resources, risks, controls, workflows, and evidence without creating a second source of truth. ## Scope - Versioned projection contracts and stable typed references to owning modules. - Valid-time and recorded-time graph queries with current authorization and purpose constraints. - Provenance, freshness, quality, uncertainty, policy, retention, and source-authority explanations on every projection. - Current-state, historical, planned/scenario, and evidence views kept semantically distinct. - Continuous-assurance indicators, process/conformance analysis, bottleneck/risk views, and governed drill-through. - Federated projections that disclose only profile-approved, locally authorized facts. - AI/assistance only over cited, policy-filtered projections; no hidden institutional memory or automatic authoritative mutation. ## First proof For one reference journey, show responsibility, open work, governing policy, source data, decisions, records, control evidence, and temporal changes in one projection, with drill-through to the owning product surfaces. ## Acceptance criteria - Projection rebuild and correction are deterministic and do not mutate owner data. - Every displayed edge identifies owner, source, effective/recorded interval, and authorization explanation. - Scenario state cannot be confused with current or historical fact. - A reference package includes quality tests and a documented institutional question it can answer. Architecture: `docs/INSTITUTIONAL_DIGITAL_TWIN.md`.