Abstract:Mental health issues significantly impact individuals' daily lives, yet many do not receive the help they need even with available online resources. This study aims to provide diverse, accessible, stigma-free, personalized, and real-time mental health support through cutting-edge AI technologies. It makes the following contributions: (1) Conducting an extensive survey of recent mental health support methods to identify prevalent functionalities and unmet needs. (2) Introducing SouLLMate, an adaptive LLM-driven system that integrates LLM technologies, Chain, Retrieval-Augmented Generation (RAG), prompt engineering, and domain knowledge. This system offers advanced features such as Risk Detection and Proactive Guidance Dialogue, and utilizes RAG for personalized profile uploads and Conversational Information Extraction. (3) Developing novel evaluation approaches for preliminary assessments and risk detection via professionally annotated interview data and real-life suicide tendency data. (4) Proposing the Key Indicator Summarization (KIS), Proactive Questioning Strategy (PQS), and Stacked Multi-Model Reasoning (SMMR) methods to enhance model performance and usability through context-sensitive response adjustments, semantic coherence evaluations, and enhanced accuracy of long-context reasoning in language models. This study contributes to advancing mental health support technologies, potentially improving the accessibility and effectiveness of mental health care globally.

What problem does this paper attempt to address?

### Problems the Paper Aims to Solve The paper aims to address several key issues in current mental health services by providing diverse, accessible, stigma-free, personalized, and real-time mental health support through advanced AI technology: 1. **Insufficient Mental Health Resources**: Globally, many regions lack adequate mental health service resources, especially in economically underdeveloped areas. 2. **Limitations of Online Mental Health Support**: Although online mental health resources exist, factors such as cost, stigma, and language barriers limit people's ability to access these resources. 3. **Need for Personalized and Real-Time Support**: Existing mental health support systems often lack effective personalized and real-time support functions, failing to meet the diverse needs of users. 4. **Risk Detection and Intervention**: Current systems are inadequate in identifying and intervening in high-risk situations (such as suicidal tendencies). To address these issues, the paper proposes the **SouLLMate** system, an adaptive system based on large language models (LLM), integrating technologies such as LangChain, Retrieval-Augmented Generation (RAG), and prompt engineering. The main contributions of this system include: - **Comprehensive Survey**: Conducted an extensive survey of existing mental health support methods, identifying common features and unmet needs. - **System Design**: Introduced the **SouLLMate** system, providing functions such as risk detection, proactive guided conversations, and utilizing RAG for personalized data upload and conversation information extraction. - **Evaluation Methods**: Developed new evaluation methods, using professionally annotated interview data and actual suicidal tendency data to preliminarily assess and detect risks. - **Performance Enhancement Methods**: Proposed three new methods to improve model performance and usability: Key Indicator Summarization (KIS), Proactive Questioning Strategy (PQS), and Stacked Multi-Model Reasoning (SMMR). Through these contributions, the paper hopes to advance mental health support technology, enhancing the accessibility and effectiveness of mental health care.

SouLLMate: An Application Enhancing Diverse Mental Health Support with Adaptive LLMs, Prompt Engineering, and RAG Techniques

SouLLMate: An Adaptive LLM-Driven System for Advanced Mental Health Support and Assessment, Based on a Systematic Application Survey

Advancing Mental Health Care: Intelligent Assessments and Automated Generation of Personalized Advice via M.I.N.I and RoBERTa

Advancing Conversational Psychotherapy: Integrating Privacy, Dual-Memory, and Domain Expertise with Large Language Models

InterMind: A Doctor-Patient-Family Interactive Depression Assessment System Empowered by Large Language Models

MindGuard: Towards Accessible and Sitgma-free Mental Health First Aid via Edge LLM

Enhancing Psychological Counseling with Large Language Model: A Multifaceted Decision-Support System for Non-Professionals

MindScape Study: Integrating LLM and Behavioral Sensing for Personalized AI-Driven Journaling Experiences

Supervised Learning and Large Language Model Benchmarks on Mental Health Datasets: Cognitive Distortions and Suicidal Risks in Chinese Social Media

Supporting the Demand on Mental Health Services with AI-Based Conversational Large Language Models (LLMs)

Distant Supervision for Mental Health Management in Social Media: Suicide Risk Classification System Development Study

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

Psy-LLM: Scaling up Global Mental Health Psychological Services with AI-based Large Language Models

Can AI Relate: Testing Large Language Model Response for Mental Health Support

SMILE: Single-turn to Multi-turn Inclusive Language Expansion via ChatGPT for Mental Health Support

Leveraging Social Media for Real-Time Interpretable and Amendable Suicide Risk Prediction with Human-in-The-Loop

Construction of real-time mental health early warning system based on machine learning

Enhancing Suicide Risk Detection on Social Media through Semi-Supervised Deep Label Smoothing

Enhancing Mental Health Support through Human-AI Collaboration: Toward Secure and Empathetic AI-enabled chatbots

Towards Understanding Emotions for Engaged Mental Health Conversations