Scholarlink

<p>Mental health disorders and substance use disorders (SUDs) frequently co-occur, yet digital interventions typically address them separately. This paper presents a novel graph-based chatbot framework that simultaneously addresses these co-occurring conditions through a unified system. Our framework employs a multi- agent architecture implemented with LangGraph, where specialized agents handle distinct conversation aspects while collaborating to provide comprehensive support. The system integrates validated screening tools directly into the conversation flow, uses Retrieval-Augmented Generation (RAG) to deliver evidence-based therapeutic content, and implements human-in-the-loop safety protocols that automatically alert therapists when crisis indicators are detected. Our comprehensive evaluation across 492 synthetic test cases demonstrates 96.88% screening accuracy, 93.33% crisis detection recall with 97.22% precision, significantly exceeding existing system performance. Our approach aims to overcome limitations of existing chatbots by combining systematic assessment, knowledge graph-based reasoning, therapeutic content retrieval, and explicit escalation protocols across both mental health and substance use domains.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

IntegraMind: An Intelligent Framework for Unified Assessment and Intervention in Dual Diagnosis

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月28日

查看全文
  • Arun Agarwal,
  • Ramanarayan Ransingh

摘要

Mental health disorders and substance use disorders (SUDs) frequently co-occur, yet digital interventions typically address them separately. This paper presents a novel graph-based chatbot framework that simultaneously addresses these co-occurring conditions through a unified system. Our framework employs a multi- agent architecture implemented with LangGraph, where specialized agents handle distinct conversation aspects while collaborating to provide comprehensive support. The system integrates validated screening tools directly into the conversation flow, uses Retrieval-Augmented Generation (RAG) to deliver evidence-based therapeutic content, and implements human-in-the-loop safety protocols that automatically alert therapists when crisis indicators are detected. Our comprehensive evaluation across 492 synthetic test cases demonstrates 96.88% screening accuracy, 93.33% crisis detection recall with 97.22% precision, significantly exceeding existing system performance. Our approach aims to overcome limitations of existing chatbots by combining systematic assessment, knowledge graph-based reasoning, therapeutic content retrieval, and explicit escalation protocols across both mental health and substance use domains.