Can Large Language Models Replace Therapists? Evaluating Performance at Simple Cognitive Behavioral Therapy Tasks.

Area of Science:

• Artificial Intelligence in Mental Health
• Clinical Psychology Applications
• Digital Therapeutics

Background:

• Large language models (LLMs) like ChatGPT are emerging technologies with potential applications in various fields.
• Psychological therapies, particularly cognitive behavioral therapy (CBT), rely on specific cognitive restructuring techniques.
• The integration of AI into mental healthcare presents both opportunities and challenges.

Purpose of the Study:

• To systematically evaluate the capability of LLMs in performing key cognitive restructuring tasks within CBT.
• To assess LLMs' effectiveness in recognizing, examining, and reframing unhelpful thoughts.
• To determine the current limitations and potential role of LLMs in supporting CBT delivery.

Main Methods:

• LLMs were presented with scenarios requiring the identification of unhelpful thoughts.
• The models' ability to analyze the validity of these thoughts was tested.
• LLMs were prompted to generate more helpful reframings of the identified thoughts.

Main Results:

• LLMs demonstrated a capacity to provide reasonable suggestions for identifying unhelpful thoughts.
• The models showed potential in offering reframed, more helpful thought alternatives.
• Current LLM performance suggests they can assist but not independently lead CBT processes.

Conclusions:

• LLMs show preliminary efficacy in supporting specific cognitive restructuring techniques in CBT.
• The findings suggest LLMs can be valuable tools for augmenting therapeutic interventions.
• Further research is needed to establish the safety and efficacy of LLMs in leading mental health delivery.