This research advances mental health applications of AI by improving how LLMs reframe negative thoughts into positive ones through a multi-step reasoning process.

• Introduces Socratic reasoning to help models better understand underlying issues before reframing
• Shows that breaking down reframing into sequential reasoning steps produces more helpful responses
• Demonstrates significant improvement in quality when evaluated by psychology students and experts
• Creates a new dataset with step-by-step reasoning paths for cognitive reframing

This work bridges AI and cognitive behavioral therapy techniques, potentially making mental health support more accessible while maintaining quality through structured reasoning approaches.

Socratic Reasoning Improves Positive Text Rewriting