Ask AI to enumerate assumptions, not answer directly — then verify completeness
When presenting compound statements to AI systems, explicitly ask for assumption enumeration rather than direct answers, then critically verify the decomposition's completeness since the AI may introduce its own hidden assumptions.
Why This Is a Rule
When you present a compound statement to AI and ask "is this right?" or "what should I do?", the AI answers your question directly — accepting all embedded assumptions as given. "Should I hire two more engineers to meet the Q3 deadline?" The AI discusses hiring strategies without questioning whether two engineers is the right number, whether Q3 is the right deadline, or whether hiring is the right intervention.
Asking for assumption enumeration instead forces a different cognitive operation. "List every assumption embedded in this statement before answering" produces: (1) the deadline is Q3, (2) more engineers will accelerate delivery, (3) two is the right number, (4) the bottleneck is engineering capacity rather than scope or process, (5) hiring can be completed in time to matter. Now you can evaluate each assumption independently before acting on the compound plan.
The verification step is critical because AI introduces its own hidden assumptions during enumeration. The model might enumerate 8 assumptions while silently inheriting 3 others from its training distribution. After the AI enumerates, ask: "What assumptions did you make in producing this list?" This second-order check catches assumptions the model treated as too obvious to mention.
When This Fires
- Presenting a plan, strategy, or complex question to an AI system
- Using AI to evaluate a decision or proposal
- Asking AI for advice on any compound problem
- Any AI interaction where the input contains multiple embedded assumptions
Common Failure Mode
Accepting the AI's assumption list as complete without verification. AI produces confident, well-organized enumerations that feel exhaustive. But the model has no way to signal what it doesn't know it's assuming. The clean, numbered list creates an illusion of completeness that stops your own assumption-hunting. Always add at least 2-3 assumptions the AI missed — if you can't think of any, you're not trying hard enough.
The Protocol
When presenting compound statements to AI: (1) Don't ask for a direct answer. Instead: "Before answering, list every assumption embedded in this statement." (2) Review the enumeration. For each assumption, assess: do I know this is true? (3) Ask the AI: "What assumptions did you make in producing this list?" (4) Add your own missing assumptions — the ones specific to your context that the AI can't know. (5) Only after validating the assumption set, ask for the analysis or answer.