AI 101

Self-reflection

Also known as: self-critique · self-evaluation

Agents

Self-reflection is a model turning its own output back on itself: checking claims, spotting gaps, or writing a critique before finalizing.

  • “Does this answer actually address the question?”
  • “Which of my claims need verification?”
  • A cheap way to catch errors before they ship.

Use it deliberately: a second pass that checks the first is one of the highest-value, lowest-cost reliability improvements available. It's not magic, but it catches a real slice of mistakes.