Self-Critique Analysis Prompt AI Prompt

Design a self-critique prompt pattern where the LLM generates an initial data analysis and then critiques and improves its own output.

Self-critique significantly improves analysis quality by catching errors, unsupported conclusions, and missing context that the initial generation missed.

1. The two-pass pattern:

   Pass 1 — Initial analysis:
   Use a standard analysis prompt to generate an initial response.
   Do not add self-critique instructions yet — let the model generate its natural first response.

   Pass 2 — Self-critique (separate prompt call):
   Feed the initial analysis back to the model with this critique prompt:

   'Review the following data analysis. Critique it on these specific dimensions:

   1. Factual accuracy: Are all numbers and statistics correctly stated? Check each claim against the source data.
   2. Unsupported claims: Are any conclusions drawn that go beyond what the data supports? Flag each one.
   3. Missing context: What important context was omitted that would change the interpretation?
   4. Confounding factors: What alternative explanations were not considered?
   5. Misleading framing: Is any language used that could lead a reader to a wrong conclusion?
   6. Precision: Are confidence levels stated where appropriate? Is uncertainty acknowledged?

   For each issue found: quote the problematic text, explain the issue, and provide the corrected version.'

   Pass 3 — Revised analysis:
   'Now write a revised version of the analysis that incorporates all the corrections from your critique.'

2. When self-critique is most valuable:
   - High-stakes analyses that will be presented to leadership
   - Analyses that will inform a significant business decision
   - Any analysis containing causal claims (correlation ≠ causation)
   - Analyses where the conclusion is surprising — surprising results deserve extra scrutiny

3. Efficiency tip:
   - For most analyses, the two-pass pattern (initial + critique) is sufficient
   - Three passes (initial + critique + revised) adds quality but also cost and latency
   - Use three passes only when the stakes are high enough to justify it

4. Automated critique checklist integration:
   - Convert the critique dimensions into a checklist that runs automatically after every analysis
   - Flag outputs that trigger any checklist item for human review before distribution

Return: the three-pass prompt sequence, a test case showing how critique improved a flawed initial analysis, and a decision guide for when to use 2 vs 3 passes.