The fastest way to misuse a forecast is to accept the summary before inspecting the structure.
Review the question first
A vague question creates polished ambiguity. Before you read the report, check that the prompt answers a bounded problem:
- what event is being simulated,
- what time horizon matters,
- what kind of outcome the operator actually needs.
Review the graph second
The graph should include the major actors, pressures, and incentives. Missing pressure is one of the most common reasons a forecast feels smart but fails in practice.
Review the evidence line
Ask what assumptions are carrying the forecast. If the answer depends on one weak interpretation in the source material, confidence should fall accordingly.
Review the confidence by consequence
High-confidence language does not make a scenario high-confidence in reality. The stricter the downstream consequence, the stricter your human review should be.
Prompt template
Re-run this scenario with emphasis on the weakest assumptions and show
which missing actor or constraint would most change the final report.
A useful review checklist
Use this quick pass:
- Is the scenario boundary clear?
- Are the core actors present?
- Is one missing pressure distorting everything?
- What evidence would reverse the result?
Related guides: What Is MiroFish? and Can AI Agents Predict Human Behavior?.