LLM-based Conversational AI Therapist for Daily Functioning Screening and Psychotherapeutic Intervention via Everyday Smart Devices (2403.10779v1)

Abstract: Despite the global mental health crisis, access to screenings, professionals, and treatments remains high. In collaboration with licensed psychotherapists, we propose a Conversational AI Therapist with psychotherapeutic Interventions (CaiTI), a platform that leverages LLMs (LLM)s and smart devices to enable better mental health self-care. CaiTI can screen the day-to-day functioning using natural and psychotherapeutic conversations. CaiTI leverages reinforcement learning to provide personalized conversation flow. CaiTI can accurately understand and interpret user responses. When the user needs further attention during the conversation, CaiTI can provide conversational psychotherapeutic interventions, including cognitive behavioral therapy (CBT) and motivational interviewing (MI). Leveraging the datasets prepared by the licensed psychotherapists, we experiment and microbenchmark various LLMs' performance in tasks along CaiTI's conversation flow and discuss their strengths and weaknesses. With the psychotherapists, we implement CaiTI and conduct 14-day and 24-week studies. The study results, validated by therapists, demonstrate that CaiTI can converse with users naturally, accurately understand and interpret user responses, and provide psychotherapeutic interventions appropriately and effectively. We showcase the potential of CaiTI LLMs to assist the mental therapy diagnosis and treatment and improve day-to-day functioning screening and precautionary psychotherapeutic intervention systems.

• The paper introduces CaiTI, a system that uses LLMs to screen daily mental health across 37 dimensions based on established assessments.
• It integrates CBT and Motivational Interviewing techniques to deliver personalized, interactive psychotherapeutic interventions.
• The study validates CaiTI through 14-day and 24-week trials, demonstrating reduced anxiety and improved daily functioning.

An Overview of "LLM-based Conversational AI Therapist for Daily Functioning Screening and Psychotherapeutic Intervention via Everyday Smart Devices"

The paper introduces CaiTI, a LLM-based conversational AI therapist that serves as a sophisticated tool for precautionary mental health monitoring and psychotherapeutic intervention. By leveraging smart devices like smartphones and smart speakers, CaiTI offers continuous interaction aimed at improving day-to-day mental health functioning through psychotherapeutic conversations.

Key Features and Contributions

The research provides several significant contributions to the domain of AI-driven mental healthcare, articulated as follows:

1. Day-to-day Functioning Screening: CaiTI employs LLMs to conduct a comprehensive screening of the user’s daily functions across 37 dimensions. These dimensions are rooted in established mental health assessments like the DSM-IV and DLA-20, providing a standardized method to assess mental well-being through interactive conversations.
2. Integrated Psychotherapeutic Interventions: Drawing upon well-established therapeutic processes such as Cognitive Behavioral Therapy (CBT) and Motivational Interviewing (MI), CaiTI ensures that users receive empathetic engagement and relevant psychological interventions. The model is meticulously designed to introduce therapy throughout the conversation, with specific task modules guiding and validating user input through the process.
3. Reinforcement Learning for Personalized Interaction: CaiTI utilizes reinforcement learning to enable adaptive conversation flows, tailoring interactions based on user profiles and historical data. This ensures that the system remains relevant and sensitive to the user's evolving mental health context.
4. Performance Testing and Evaluation: The paper includes detailed comparative analyses of various LLMs, notably GPT-4 and Llama-2 variants, evaluating their efficacy in handling distinct psychotherapeutic tasks. This paper provides insights into the strengths and limitations of different LLM configurations in accurately executing the system's objectives.

Experimental Validation

The research showcases the successful deployment of CaiTI through both a 14-day trial and an extended 24-week paper. These periods allowed for the evaluation of the system’s capability to engage users daily and its effects on their mental health parameters. The studies found that CaiTI's interactions were well-received, indicating notable improvements in day-to-day functioning and reductions in anxiety and depression symptoms among participants.

Implications and Future Directions

The implications of deploying such a system are wide-ranging:

• Enhanced Accessibility: CaiTI democratizes access to mental health resources, providing a therapeutic touchpoint for users who might otherwise face barriers to traditional therapy settings.
• Scalability of Mental Health Interventions: By integrating with commonly available smart devices, CaiTI foregrounds a scalable model for disbursing mental health interventions to a broader audience.
• Incremental Advancements in AI Healthcare Applications: The methodology of utilizing specific LLM modules for task-specific interventions represents an advancement in the design of AI healthcare solutions, reducing concerns over potential misinformation or bias typically associated with LLMs.

Looking ahead, the authors propose further expansion of CaiTI to incorporate a more refined integration of IoT devices, thus extending the biofeedback loop to improve the precision and personalization of mental health interventions.

In summary, the paper documents a well-rounded endeavor in merging state-of-the-art AI with practical psychotherapy practices. Through methodical research and feedback from real-world deployments, it underscores the capacity and promise of LLM-powered AI in supporting and potentially transforming mental health care practices.