Survey and Experiments on Mental Disorder Detection via Social Media: From Large Language Models and RAG to Agents

TL;DR

The paper tackles the challenge of detecting mental disorders from social media by surveying how large language models, retrieval-augmented generation, and agentic systems can be combined to improve reliability, interpretability, and autonomy. It outlines a taxonomy of disorders, data modalities, and methodological paradigms, then conducts benchmarked zero-shot, few-shot, and RAG-based experiments across multiple datasets. Key contributions include a unified benchmarking framework, systematic analysis of RAG and agentic systems for mental-health tasks, and guidance on when retrieval-grounded or autonomous reasoning approaches are most beneficial. The findings highlight that while LLMs excel at understanding unstructured text, performance significantly benefits from domain grounding and multi-step reasoning, particularly on long, structured narratives, with implications for safer, more explainable digital mental-health interventions.

Abstract

Mental disorders represent a critical global health challenge, and social media is increasingly viewed as a vital resource for real-time digital phenotyping and intervention. To leverage this data, large language models (LLMs) have been introduced, offering stronger semantic understanding and reasoning than traditional deep learning, thereby enhancing the explainability of detection results. Despite the growing prominence of LLMs in this field, there is a scarcity of scholarly works that systematically synthesize how advanced enhancement techniques, specifically Retrieval-Augmented Generation (RAG) and Agentic systems, can be utilized to address these reliability and reasoning limitations. Here, we systematically survey the evolving landscape of LLM-based methods for social media mental disorder analysis, spanning standard pre-trained language models, RAG to mitigate hallucinations and contextual gaps, and agentic systems for autonomous reasoning and multi-step intervention. We organize existing work by technical paradigm and clinical target, extending beyond common internalizing disorders to include psychotic disorders and externalizing behaviors. Additionally, the paper comprehensively evaluates the performance of LLMs, including the impact of RAG, across various tasks. This work establishes a unified benchmark for the field, paving the way for the development of trustworthy, autonomous AI systems that can deliver precise and explainable mental health support.

Figures (12)

• Figure 1: Global Population with Mental Disorders (2021) who-mental-disorders, where ‘Dep.’ stands for depression, ‘Anx.’ for anxiety, ‘Bip.’ for bipolar, ‘Sch.’ for schizophrenia, and ‘Con.’ for conduct-dissocial disorder.
• Figure 2: Taxonomy and Symptom Introduction of Mental Disorders.
• Figure 3: Representation of Different Data Types on Social Media.
• Figure 4: Overview of LLMs for Mental Disorder Detection. LLMs take a structured prompt that combines role, task, and contextual instructions with user profile information and historical posts, and then output a diagnosis.
• Figure 5: Illustration of the Basic Retrieval-Augmented LLMs framework. LLM retrieves external medical knowledge and then generates the final response.