Cognitive Verifier

What Is This Pattern?

The Cognitive Verifier pattern asks the AI to critically examine its own reasoning process, identify potential errors or assumptions, and verify the correctness of its conclusions before presenting them. This pattern improves accuracy by making the AI double-check its work, similar to how a human would review their solution before submitting it. It's especially powerful when combined with Chain-of-Thought, as the AI can verify each step of its reasoning.

How It Works

You instruct the AI to: (1) Provide its initial answer or reasoning, (2) Review its own work critically, checking for logical errors, incorrect assumptions, or missing considerations, (3) Verify the correctness of each step or conclusion, and (4) Present the verified answer with confidence. The AI acts as its own quality control mechanism.

When To Use This Pattern

• Mathematical problems or calculations where accuracy is critical
• Logical reasoning tasks where errors could cascade
• Code review or debugging where correctness matters
• Analysis tasks where assumptions need verification
• Any complex problem where self-checking improves quality
• Situations where you want to catch errors before they propagate

Best Practices

• Explicitly ask the AI to verify its reasoning
• Specify what aspects to check (logic, math, assumptions, completeness)
• Combine with Chain-of-Thought for step-by-step verification
• Use for critical tasks where errors are costly
• Instruct the AI to show its verification process
• Ask for confidence level after verification

Common Mistakes to Avoid

• Not specifying what to verify (too vague)
• Using for simple tasks where verification adds no value
• Expecting perfect verification (AI can still make errors)
• Not combining with other patterns (works best with Chain-of-Thought)
• Asking for verification but not using the results

Related Patterns