Detecting Eating Disorders From Social Media Content: What Has Been Done and Where Do We Go Next?

Area of Science:

• Computational psychiatry
• Digital health
• Artificial intelligence in healthcare

Background:

• Eating disorders (EDs) affect millions, yet less than 20% receive treatment, necessitating scalable identification and support methods.
• Social media (SM) platforms host pervasive ED-related content, presenting an opportunity to detect ED symptoms through user-generated data.

Purpose of the Study:

• To examine how artificial intelligence (AI) and computational methods can detect ED symptoms from SM content.
• To explore the potential for timely intervention based on SM data analysis.
• To review current SM-based ED detection research and identify future directions.

Main Methods:

• Review of existing SM-based ED detection methods, including content tags, topic modeling, and natural language processing (NLP).
• Discussion of opportunities for integrating detection with digital interventions.
• Examination of challenges in developing, evaluating, and implementing AI-driven detection tools.

Main Results:

• Research confirms the feasibility of using SM data to identify individuals with ED symptoms.
• Initial exploration into pairing SM-based ED detection with digital interventions is underway.
• Current SM content moderation systems lack transparency, underscoring the need for expert involvement in developing detection methods.

Conclusions:

• Harnessing SM data offers a significant opportunity to identify and support individuals with ED symptoms at scale.
• Interdisciplinary collaboration and ethical considerations are crucial for transforming SM into a supportive resource.
• Continued research and development of robust detection systems, guided by ED experts, are essential.