AutoCBT: An Autonomous Multi-agent Framework for Cognitive Behavioral Therapy in Psychological Counseling (2501.09426v1)

Abstract: Traditional in-person psychological counseling remains primarily niche, often chosen by individuals with psychological issues, while online automated counseling offers a potential solution for those hesitant to seek help due to feelings of shame. Cognitive Behavioral Therapy (CBT) is an essential and widely used approach in psychological counseling. The advent of LLMs and agent technology enables automatic CBT diagnosis and treatment. However, current LLM-based CBT systems use agents with a fixed structure, limiting their self-optimization capabilities, or providing hollow, unhelpful suggestions due to redundant response patterns. In this work, we utilize Quora-like and YiXinLi single-round consultation models to build a general agent framework that generates high-quality responses for single-turn psychological consultation scenarios. We use a bilingual dataset to evaluate the quality of single-response consultations generated by each framework. Then, we incorporate dynamic routing and supervisory mechanisms inspired by real psychological counseling to construct a CBT-oriented autonomous multi-agent framework, demonstrating its general applicability. Experimental results indicate that AutoCBT can provide higher-quality automated psychological counseling services.

• The paper introduces AutoCBT, an autonomous multi-agent framework that uses dynamic routing and supervisory mechanisms to adapt CBT responses.
• It employs integrated short- and long-term memory systems and a bilingual dataset evaluation to maintain context and improve response quality.
• The research demonstrates that AutoCBT outperforms existing LLM-based approaches, suggesting scalable and accessible automated psychological support.

AutoCBT: An Autonomous Multi-agent Framework for Cognitive Behavioral Therapy in Psychological Counseling

The paper "AutoCBT: An Autonomous Multi-agent Framework for Cognitive Behavioral Therapy in Psychological Counseling" addresses the integration of Cognitive Behavioral Therapy (CBT) with advanced computational models to offer psychological counseling. This research highlights the development of AutoCBT, an autonomous multi-agent system that leverages LLMs and agent technology to address the limitations of traditional CBT delivery and extend its applicability in online psychological counseling.

Key Aspects of the AutoCBT System

AutoCBT introduces a dynamic routing and supervisory mechanism inspired by real-life psychological counseling to enable more adaptive and context-aware therapeutic systems. This system comprises a Counsellor Agent and multiple Supervisor Agents, each operating over LLMs. The framework employs a flexible topology and routing strategies that mimic the complexity of human cognitive processes, enhancing the ability to generate high-quality responses in single-turn consultation scenarios.

1. Dynamic Routing and Supervisory Mechanism: The system's Counsellor Agent decides whether to consult a Supervisor or respond directly to a user query through dynamic routing processes. This capability enables the system to adapt and refine its responses based on the input, thus improving response quality and relevance.
2. Memory Mechanisms: AutoCBT incorporates both short-term and long-term memory for each agent, allowing for the extraction and storage of summaries, which aids in maintaining context during interactions.
3. Experimental Evaluation: Utilizing a bilingual dataset, the quality of machine-generated responses was evaluated across several metrics: empathy, identification of cognitive distortions, reflection, strategy provision, encouragement, and relevance. Results indicate that AutoCBT outperforms existing LLM-based approaches by achieving superior scores, illustrating its potential in automated psychological counseling.

Practical and Theoretical Implications

The practical implications of AutoCBT are considerable, as it demonstrates the capacity to deliver CBT-oriented autonomous counseling services effectively. It reflects substantial advancements in making mental health support scalable and accessible beyond traditional settings, potentially reducing barriers for individuals reluctant to seek in-person therapy due to stigma or resource limitations.

Theoretically, AutoCBT presents a significant step towards understanding how multi-agent systems based on LLMs can be organized to enhance cognitive and emotional support systems. It underscores the importance of dynamic adaptability in AI systems designed for complex human-centric tasks such as psychological counseling.

Challenges and Future Developments

Despite the promising results, the research identifies challenges such as managing the complexity of simultaneous routing and role assignments among agents, which requires further exploration. Furthermore, the paper notes the Llama model's over-protective nature in handling sensitive topics, which impacts user experience and requires careful handling in future implementations.

This paper opens avenues for future research to enhance AI systems' interpretability and adaptability in human-centric applications. It suggests that further refinements in dynamic routing and memory mechanisms can lead toward more sophisticated and effective AI counseling systems.

Concluding Remarks

The development of AutoCBT represents a meaningful stride in leveraging AI to augment traditional therapeutic approaches such as CBT. As research in this domain progresses, AutoCBT's framework and methodologies could inform the design of future AI systems in diverse areas of psychological support, advancing the interface between technology and mental health care.