INDEPENDENT POLICY & RESEARCHConceptual framework Edition 0.16.4
Demonstration packs A–H · All synthetic

Practice with disclosed inputs.
Keep the conclusion narrow.

These fictional examples exercise accounting and review rules. They do not describe actual countries, politicians, institutions or measured MAR effects.

DEMO A · Synthetic exercise

Civilian diesel conservation

100 barrels opening stock. Day 1: 90 requested, 20 received, 70 delivery capacity. Day 2: 40 new requests, 10 received, 100 capacity. Toggle backlog to distinguish carried requests from permanently unfilled requests.

Current fixture calculation
  • New demand: 130 barrels
  • Demand served under stated constraints: 130 barrels
  • Closing stock: 0 barrels
  • Final carried backlog: 0 barrels
  • Demand still unserved at the end or lost: 0 barrels
  • Demand coverage: 100 %
  • First day ending with unmet demand: 1 day number

Daily service is calculated from declared constraints. Supply totals and capacity do not verify actual service.

Run CORE-001 and inspect the fixture
DEMO B · Synthetic exercise

Public policy cost

Fictional year-zero expense of $100,000 and two later expenses of $40,000 each. Unknown revenue remains unknown; assumed funding offsets stay separate from measured fiscal flows.

Current fixture calculation
  • Baseline cost: 0 USD minor units; nominal; reference 2026
  • Incremental cost excluding baseline: 18000000 USD minor units; nominal; reference 2026
  • Declared funding offsets: 2000000 USD minor units; nominal; reference 2026
  • Declared program costs less assumed funding offsets: 16000000 USD minor units; nominal; reference 2026
  • Discounted declared program funding need: 16000000 USD minor units; nominal; reference 2026
  • Incremental fiscal balance excluding current-policy baseline: unknown USD minor units; nominal; reference 2026
  • Incremental fiscal NPV (unknown if revenue or transfers unknown): unknown USD minor units; nominal; reference 2026

Annual fiscal inputs are incomplete. Incremental fiscal balance and NPV are not estimated.

Run CORE-003 and inspect the fixture
DEMO C · Synthetic exercise

Copied crisis reports

Articles A, B and C trace to O1; D traces to O2; E has unknown origin. Two traceable origins are not five independent confirmations.

Current fixture calculation
  • Known traceable origin clusters: 2 origin clusters
  • Reports with unknown origin: 1 reports
  • Declared reports: 5 reports
  • Traceable origins without supplied timestamps: 2 origins

Traceable origin clusters do not establish independent corroboration or truth.

Run CORE-005 and inspect the fixture
DEMO D · Synthetic exercise

Independent civilian fallback

A shared portal fails. Service A has a tested paper substitute; B has none; C has an untested substitute. A is retained, B unavailable and C unknown. The portal itself is also a graph node.

Current fixture calculation
  • Retained functions: 1 functions
  • Unavailable functions: 2 functions
  • Unknown functions: 1 functions
  • All-function supplied restoration upper bound: unknown hours after exercise start

Counts include every declared graph node, including shared services. Retained status is conditional on supplied test evidence.

Run CORE-010 and inspect the fixture
DEMO E · Synthetic exercise

Aggregate service coverage

Two disjoint fictional counties each have 100 eligible requests. Serving 70 and 90 yields coverage of 70% and 90%, a 20 percentage-point observed gap. No cause or discrimination finding is inferred.

Current fixture calculation
  • Total eligible (only disclosed disjoint counts): 200 requests
  • Total served (only disclosed disjoint counts): 160 requests
  • Aggregate coverage: 80 %
  • Observed gap among disclosed nonzero-denominator groups: 20 percentage points
  • Suppressed aggregate groups: 0 groups
  • Total disclosed budget burden: unknown unknown currency minor units

Budget burden data are absent or incomplete; burden shares are not estimated.

Run CORE-009 and inspect the fixture
DEMO F · Synthetic exercise

Permission withdrawal

Withdrawal blocks a fictional cross-border activity. Independent local record review remains separately declared. Evidence with two traceable origins remains disputed unless the participants actually verify and acknowledge it.

Current fixture calculation
  • Stopped activities: 1 activities
  • Unresolved activities: 0 activities
  • Exercise-only activities: 0 activities
  • Activities with supplied approval declarations: 1 activities
  • Transaction activities passing supplied declarations: 0 activities

Supplied approval declarations are not a legal-clearance certificate; no permission is inferred from geography, commodity, or source.

Run CORE-014 and inspect the fixture
DEMO H · Synthetic exercise

Electricity interruption accounting

100 customers × 2 hours plus a separately confirmed disjoint group of 50 × 3 hours gives 350 customer-hours. Without the disjoint-group declaration, the combined total remains unknown.

Current fixture calculation
  • Audited false alerts: 1 alerts
  • Audited missed events: 1 events
  • Unknown alert outcomes: 1 alerts
  • Unaudited miss candidates: 0 events
  • Precision among alerts with known outcomes: 0.5 fraction of classified alerts
  • Recall among documented distinct audited events: 0.5 fraction of documented audited events
  • Total measured disjoint interruption: 350 customer-hours

Timing is descriptive. The rates cover the documented classified cases.

Run CORE-012 and inspect the fixture
DEMO G · Synthetic arithmetic illustration

Cost per documented training completion

100 authorized seats, 85 documented completions and $850,000 of recorded eligible expense in the same period and cost scope.

850,000 USD ÷ 85 completions = 10,000 USD/completion

A zero or unknown denominator makes this quotient undefined. This is a hand-replicable arithmetic illustration, not an employment, earnings or training-quality estimate. EDU-002 remains a planned specialized workflow; this example does not make that module complete.

Inspect the planned domain requirements

What would a controlled MAR comparison require?

Repeat the same exercise under fragmented arrangements, a structured human-only process and MAR. Keep resources, source documents, information release dates, staffing and criteria equal. Record actual review labor, independently adjudicated errors, disclosure burdens and failed fallback. A copied false source or extra review effort may make MAR worse. Identical arithmetic establishes no AI benefit.

Compare the approachesRead separately sourced historical cooperation