Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

MAGI: Multi-Agent Guided Interview for Psychiatric Assessment

Guanqun Bi, Zhuang Chen, Zhoufu Liu, Hongkai Wang, Xiyao Xiao, Yuqiang Xie, Wen Zhang, Yongkang Huang, Yuxuan Chen, Libiao Peng, Yi Feng, Minlie Huang

TL;DR

MAGI addresses the misalignment between LLM-based interviews and psychiatric protocols by operationalizing the MINI into a four-agent collaborative workflow with four specialized components. It introduces PsyCoT, a DSM-5-aligned reasoning trace that maps symptoms to criteria, and demonstrates how the navigation, question, judgment, and diagnosis agents work together to maintain protocol fidelity while enabling adaptive and empathetic dialogue. In a large school-based study with 1,002 real interviews across depression, GAD, social anxiety, and suicide risk, MAGI achieves high diagnostic agreement with expert clinicians and improved explainability relative to baseline LLM approaches. The work demonstrates scalable, ethical, and transparent AI-assisted psychiatric assessment with potential to improve access to structured mental health care.

Abstract

Automating structured clinical interviews could revolutionize mental healthcare accessibility, yet existing large language models (LLMs) approaches fail to align with psychiatric diagnostic protocols. We present MAGI, the first framework that transforms the gold-standard Mini International Neuropsychiatric Interview (MINI) into automatic computational workflows through coordinated multi-agent collaboration. MAGI dynamically navigates clinical logic via four specialized agents: 1) an interview tree guided navigation agent adhering to the MINI's branching structure, 2) an adaptive question agent blending diagnostic probing, explaining, and empathy, 3) a judgment agent validating whether the response from participants meet the node, and 4) a diagnosis Agent generating Psychometric Chain-of- Thought (PsyCoT) traces that explicitly map symptoms to clinical criteria. Experimental results on 1,002 real-world participants covering depression, generalized anxiety, social anxiety and suicide shows that MAGI advances LLM- assisted mental health assessment by combining clinical rigor, conversational adaptability, and explainable reasoning.

MAGI: Multi-Agent Guided Interview for Psychiatric Assessment

TL;DR

MAGI addresses the misalignment between LLM-based interviews and psychiatric protocols by operationalizing the MINI into a four-agent collaborative workflow with four specialized components. It introduces PsyCoT, a DSM-5-aligned reasoning trace that maps symptoms to criteria, and demonstrates how the navigation, question, judgment, and diagnosis agents work together to maintain protocol fidelity while enabling adaptive and empathetic dialogue. In a large school-based study with 1,002 real interviews across depression, GAD, social anxiety, and suicide risk, MAGI achieves high diagnostic agreement with expert clinicians and improved explainability relative to baseline LLM approaches. The work demonstrates scalable, ethical, and transparent AI-assisted psychiatric assessment with potential to improve access to structured mental health care.

Abstract

Automating structured clinical interviews could revolutionize mental healthcare accessibility, yet existing large language models (LLMs) approaches fail to align with psychiatric diagnostic protocols. We present MAGI, the first framework that transforms the gold-standard Mini International Neuropsychiatric Interview (MINI) into automatic computational workflows through coordinated multi-agent collaboration. MAGI dynamically navigates clinical logic via four specialized agents: 1) an interview tree guided navigation agent adhering to the MINI's branching structure, 2) an adaptive question agent blending diagnostic probing, explaining, and empathy, 3) a judgment agent validating whether the response from participants meet the node, and 4) a diagnosis Agent generating Psychometric Chain-of- Thought (PsyCoT) traces that explicitly map symptoms to clinical criteria. Experimental results on 1,002 real-world participants covering depression, generalized anxiety, social anxiety and suicide shows that MAGI advances LLM- assisted mental health assessment by combining clinical rigor, conversational adaptability, and explainable reasoning.

Paper Structure

This paper contains 35 sections, 4 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Automatic Mental Health Assessment
  4. MINI: A Structured Clinical Interview
  5. Methodology
  6. Navigation Agent
  7. Question Agent
  8. Judgment Agent
  9. Diagnosis Agent
  10. Experiments
  11. Data Collection
  12. Dialogue Quality Evaluation
  13. Experimental Setup
  14. Baselines
  15. Metrics
  16. ...and 20 more sections

Figures (4)

  • Figure 1: Example dialogue flow from Magi. Our multi-agent framework Magi guides participants through structured psychiatric interviews following MINI protocol.
  • Figure 2: Overview of Magi framework. The framework consists of a navigation agent for interview management, a question agent for dynamic utterance generation, a judgment agent for symptom validity analysis, and a diagnosis agent for DSM-5 compliant conclusions, ensuring adherence to psychiatric protocols and conversational adaptability.
  • Figure 3: Heatmap illustrating participant interaction patterns with interview nodes across different disorders, with darker blue indicating higher percentage at different assessment nodes.
  • Figure 4: Case of the demo of MAGI.