Methodology
Every simulation on the platform — the pilot, the built-in scenarios, and each hot-topic page under What-If Simulations — runs on the same cost / benefit / risk engine, under the same pipeline, with the same guardrails. This page documents that methodology, the reasons simulations matter before committing to a decision, and the non-bias algorithm that keeps the engine honest.
Every simulation in this section follows the same four-step pipeline used across the platform:
- 1Declared environment
Before the engine scores anything, you state the world it should assume — buffer-zone status, strike regime, FX/debt state, chokepoint transit rate. Missing a required prerequisite blocks the run. Without declared assumptions, scoring is not falsifiable.
- 2Neutral restatement
The proposal or decision is restated in neutral language. Psychologically loaded framing — 'opportunity', 'pragmatism', 'stability' — is flagged in the Narrative Bias panel, not silently accepted.
- 3Seven-dimension scoring
Economic Upside (20), Political Cost (15), Legal / Regulatory Cost (15), Sovereignty Impact (20), Security Impact (10, bidirectional), Social Impact (10, bidirectional), Historical Precedent (10). Each score carries a rationale cited to the proposal, the environment, or a concrete precedent.
- 4Net Impact, Hidden Cost, reversibility, 1 / 5 / 10-year outlook
Benefit minus Risk gives a Net Impact Score. A Hidden Cost Index (0–100) aggregates sovereignty, legal, political, weak precedent, and a reversibility penalty. A red-flag matrix, best / middle / worst scenarios over three horizons, and a plain-language final judgment complete the output.
Most strategic decisions are sold on their visible benefits. A simulation forces the structural costs — lock-in, legal drift, sovereignty erosion, reversibility loss — out of the footnotes and into the same dashboard as the upside.
The same proposal can look favorable at 1 year, inconclusive at 5, and structurally harmful at 10. Scoring across best, middle, and worst cases over three horizons makes that asymmetry visible rather than hidden behind a single headline number.
Thresholds are declared before the run: sovereignty above 12/20, legal cost above 10/15, hidden-cost index above 60/100, low reversibility. When four or more cross, the verdict is structurally high-risk — regardless of narrative.
The engine is not instructed to decide whether a proposal is good or bad. It is instructed to expose the full structural impact across seven dimensions over time, given the declared environment. The following guardrails keep the algorithm non-partisan:
- No verdict-first principle. The system prompt explicitly forbids advocacy. Its role is to distinguish visible short-term gains from hidden long-term structural costs — not to recommend.
- Declared environment gate. The scenario cannot run until required prerequisites are filled. The model is not allowed to invent the world it scores against.
- Fixed category weights. Weights (20/15/15/20/10/10/10) are pre-registered in code. The model cannot silently reweight dimensions to favor an outcome.
- Bidirectional security and social scoring. Security and Social Impact are not monotonic benefits. A promised peace dividend can score as false stability; a promised social good can score as fragmentation. The model is explicitly forbidden from defaulting these to 10.
- Narrative bias detector. Psychologically loaded terms are surfaced with their concern, independently of the score — so reframing a proposal does not change its structural cost.
- Reversibility as a first-class variable. Low reversibility adds a 25-point penalty to the Hidden Cost Index. Lock-in is scored regardless of which side benefits from it.
- Cited rationales. Every category score must cite the proposal text, the declared environment, or a concrete precedent. Unsupported assertions fail the schema.
- Pre-registered red flags. Structural thresholds are defined in code, not inferred by the model. The auto-verdict is a count of crossed thresholds, not a narrative.