Write your schemas first, then ask AI to extract assumptions and find patterns — AI cannot inspect the unwritten
When using AI to assist schema inspection, first externalize your thinking in writing, then request assumption extraction, pattern detection across entries, or adversarial questioning—because AI can only inspect articulated schemas, not internal ones.
Why This Is a Rule
AI can only operate on text you provide — it cannot introspect your mind. Asking AI "what are my assumptions?" without first externalizing your thinking produces the AI's generic assumptions about people in your situation, not your actual assumptions. The AI doesn't know your schemas because your schemas live in your head, not in the prompt.
The prerequisite — writing first — converts internal schemas into inspectable text. Once externalized, AI becomes a powerful schema inspection tool through three modes: Assumption extraction ("list every assumption embedded in this text"), Pattern detection ("what themes recur across these ten journal entries?"), and Adversarial questioning ("argue against each of my stated rules — what's the strongest case that each one is wrong?").
Each mode leverages AI's strengths while requiring human externalization as input. The human does what AI can't (introspecting and articulating internal schemas); the AI does what humans can't (processing large volumes of text for patterns, generating adversarial arguments without emotional investment, extracting assumptions with no loyalty to the author's perspective).
When This Fires
- Using AI to assist with self-examination, schema inspection, or mental model analysis
- After completing a schema audit (Five-step schema audit: list rules, source origins, find successes/failures, rate confidence, check evidence) and wanting AI to find patterns across entries
- When you want an adversarial perspective on your operating rules but don't have a thinking partner
- Any AI-augmented self-reflection where the input is your own thinking
Common Failure Mode
Asking AI to analyze your schemas without writing them first: "What patterns do you see in how I think?" The AI has no access to how you think — it can only analyze how you write. Without written externalization, the AI produces generic personality analysis that feels insightful but isn't personalized. Write your schemas first; then AI analysis becomes genuinely personalized because it's operating on your actual articulated thinking.
The Protocol
For AI-assisted schema inspection: (1) Write first: externalize your thinking — operating rules, decision patterns, recurring reactions, assumptions — in text. Don't ask AI to generate this; generate it yourself. (2) Then request one of three AI modes: "Extract every assumption embedded in this text" (assumption extraction), "What patterns do you see across these [N] entries?" (pattern detection), or "Argue against each of my stated rules — what's the strongest case each one is wrong?" (adversarial questioning). (3) Review AI output against your lived experience — AI finds patterns in text; you validate whether they match reality.