Comparing Human and AI Therapists in Behavioral Activation for Depression: Cross-Sectional Questionnaire Study.

Area of Science:

• Artificial Intelligence in Healthcare
• Clinical Psychology
• Mental Health Technology

Background:

• Large language models (LLMs) are increasingly used for mental health support due to therapist shortages.
• The clinical effectiveness of LLMs in therapeutic settings is not well-established.

Purpose of the Study:

• To compare the therapeutic knowledge and response capabilities of LLMs and psychotherapy trainees in behavioral activation (BA) therapy.
• To assess the impact of structured training materials on both LLMs and human participants.

Main Methods:

• Six LLMs and eight human participants completed a depression and BA questionnaire.
• Human participants underwent training; LLMs received training content as context.
• Responses were evaluated by experts for empathy, validation, and theory of mind.

Main Results:

• LLMs initially scored higher in knowledge and were rated superior in empathy, validation, and anticipation of cognition/emotion.
• LLMs maintained their performance advantage after training, outperforming human participants.

Conclusions:

• LLMs can generate high-quality, empathetic therapeutic responses.
• LLMs show potential as tools in mental health care, pending further clinical validation.