1. Problem: How to Make Online Psychological Counseling More Intelligent and Humanized
With the growing demand for mental health, traditional offline psychological counseling is difficult to popularize due to its limitations, while online automated psychological counseling has become a new solution. Cognitive Behavioral Therapy (CBT) is a widely used form of psychological treatment that mainly improves mental states by identifying and challenging users’ cognitive biases. However, current CBT systems based on Large Language Models (LLM) often have fixed structures, lack dynamic adaptability, or generate hollow, unhelpful responses. This flaw severely impacts the quality of counseling.
2. Method: How the Multi-Agent Framework of AutoCBT Solves the Problem
The paper proposesAutoCBT, an autonomous multi-agent framework for CBT. AutoCBT enables multiple agents to collaborate on psychological counseling tasks through dynamic routing and supervisory mechanisms. Its main features include:
Counsellor Agent: Serves as the user interface, responsible for receiving and providing feedback on user information.
Supervisor Agents: Assist in analyzing user input, providing multi-level support.
Dynamic Routing Strategy: Flexibly assigns tasks to appropriate agents based on user needs, ensuring a personalized and efficient counseling process.
Unlike traditional single models, AutoCBT not only incorporates the core principles of CBT but also optimizes the identification and challenge of users’ cognitive biases through multi-agent collaboration. This flexible topology allows AutoCBT to adapt to different psychological therapy needs while enhancing the system’s self-optimization capabilities.
3. Effect: Higher Quality Psychological Counseling Experience
Experimental results show that AutoCBT performs excellently on bilingual datasets in Chinese and English. On six core metrics, AutoCBT outperformed the baseline PromptCBT method. For instance, in terms of emotional support, AutoCBT demonstrated stronger empathy through a gentler tone and contextual adaptability, providing users with better emotional validation, whereas PromptCBT was relatively too academic and lacked flexibility.
Human evaluations further confirmed this: in over 70% of psychological counseling tasks, psychology professionals preferred AutoCBT’s responses, especially in identifying and challenging cognitive biases. AutoCBT not only provides clear guidance but also enhances users’ emotional resonance through delicate expression, making it more aligned with user needs in practical applications.
4. Summary: From Intelligence to Empathy, The New Future of Psychological Counseling
AutoCBT pioneeringly integrates multi-agent technology into CBT psychological counseling, significantly enhancing the handling of users’ cognitive biases and the quality of dialogue through dynamic routing and collaboration mechanisms. This framework offers a new paradigm for mental health services, applicable not only to mild psychological issues but also providing warmth and encouragement for users needing deeper emotional support.
The future of mental health technology has arrived, and AutoCBT makes psychological therapy more caring and efficient. Want to know more technical details? Stay tuned!
Paper Title: AutoCBT: An Autonomous Multi-agent Framework for Cognitive Behavioral Therapy in Psychological Counseling
Paper Link:https://arxiv.org/pdf/2501.09426