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

TL;DR

CaiTI addresses the gap in accessible daily-functioning screening and psychotherapeutic care by combining large language models with reinforcement learning to personalize conversations across $37$ dimensions and deliver MI and CBT via everyday smart devices. The system architecture segregates tasks into a Questioner, Response Analyzer, Reflection-Validation chain, and CBT Reasoner/Guide/Validator, all trained with therapist-labeled datasets and evaluated through 14-day in-lab/home studies and a 24-week longitudinal trial with real users. Microbenchmarks compare GPT-4, GPT-3.5 Turbo, and Llama-2, showing GPT-based models generally outperforming Llama across screening and psychotherapy components, while RL-driven personalization enhances engagement. The results demonstrate CaiTI’s capability to converse naturally, interpret responses accurately, and provide appropriate interventions, supporting its potential to augment traditional psychotherapy and broaden daily mental-health screening in privacy-conscious settings.

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 large language models (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.

Figures (13)

• Figure 1: The system architecture of CaiTI consists of two main functionalities: 1) day-to-day functioning screening through natural conversation and 2) precautionary psychotherapeutic conversational interventions.
• Figure 2: The workflow of CaiTI, including the connections between the main components and the locations where different functionalities take place.
• Figure 3: The process to converse with the user naturally to screen day-to-day functioning and provide psychotherapies through the conversation. In particular, MI therapies are conducted throughout the conversation, while CBT proceeds at the end of the conversation session.
• Figure 4: The example process of how follows the sequence based on the Q-table, segments the user response, classifies the Segment into the format of (Dimension, Score), and proceeds with reflection-validation (R-V) process.
• Figure 5: An example conversation using 's customized smartphone platform. The MI and CBT embedded in the conversation are annotated.