Reflection

Purpose

The Reflection pattern encourages the AI to evaluate its own responses, helping to identify and correct potential errors or inaccuracies. This is useful for refining outputs and improving accuracy over multiple iterations.

How it Works

After generating an initial response, prompt the AI to reflect on its accuracy, tone, and content. This can lead to improved output quality through self-review.

Example

• Prompt: “Review your response for accuracy and provide any corrections.”
• AI Response: Reflects on the initial output and makes adjustments as needed.

How to Use

1. Request Self-Review: Ask the AI to evaluate its output for any inaccuracies.
2. Analyze Corrections: Review the AI’s suggested corrections and approve or modify them.
3. Iterate as Needed: Repeat the reflection process until the response meets your standards.

Practical Tips

• Combine with Fact Checking: Use Reflection to enhance fact-checking in information-heavy tasks.
• Use for Tone Adjustment: Reflecting on tone can help align the response with the desired style.
• Encourage Specificity: Ask the AI to reflect on specific aspects, like language clarity or detail accuracy.

Watch Outs

• Limited Self-Insight: The AI’s self-review may miss nuanced errors, so always verify the final response.
• Redundant Feedback: The AI might suggest unnecessary changes; focus on meaningful corrections.
• Time-Intensive: Multiple reflections can be time-consuming, so limit their use to critical tasks.

Creative Use Cases

• For Writers: Refine drafts by having the AI review and suggest improvements.
• For Editors: Use reflection to improve clarity and consistency in editorial content.

Related Patterns

• Fact Check List: Use Fact Check List for targeted verification alongside Reflection.
• Cognitive Verifier: Pair with Cognitive Verifier for detailed reviews of complex content.

Sources

• A Prompt Pattern Catalog to Enhance Prompt Engineering with ChatGPT