LLM-Empowered Social Support for Eating Disorders: A SEM-Based Correlational Study of Usage Behavior and Support Effectiveness

Abstract

The global incidence rate of eating disorders is on the rise, and traditional prevention and treatment face issues such as uneven resource distribution and stigma. Large Language Models (LLMs) have emerged as a new carrier of social support for patients. Based on the Social Support Theory, this study employed structural equation modeling (with 134 valid questionnaires) and semi-structured interviews to explore the usage behavior and effects of LLM-supported social support. The results show that both usage motivation and usage intensity are significantly positively correlated with informational, emotional, and practical support (p<0.01), with motivation exerting a stronger influence. LLMs have prominent advantages in integrated information provision and round-the-clock emotional support, but their ability to connect with practical resources is insufficient. This study confirms the supplementary value of LLMs in the digital intervention of eating disorders and also points out their shortcomings in practical support.