# Thigh gaps and filtered snaps: a qualitative study exploring opportunities to mitigate social media harm through content moderation for people with eating disorders

## TL;DR

This study explores how harmful social media content affects people with eating disorders and suggests ways to moderate content to reduce harm.

## Contribution

The paper introduces an eight-category framework for identifying harmful social media content related to body image and eating disorders.

## Key findings

- Participants identified a spectrum of harmful and ambiguous content on social media.
- Algorithms contribute to an 'echo chamber' that amplifies harmful content for people with eating disorders.
- Shared responsibility among users, creators, platforms, and policymakers is needed to mitigate social media harm.

## Abstract

The ubiquity of social media has increased exposure to idealised beauty standards, often unrealistic and harmful. Repeated exposure has been linked to body dissatisfaction, harmful behaviours, and potentially the development of eating disorders (ED). Given the volume of content produced daily, effective harm mitigation strategies (automated or user-driven) are essential, requiring an informed understanding of the contexts and nuances surrounding harmful content.

The study has two key aims: (1) to understand the perspectives of experts by profession and people with lived experience of ED, on what makes social media content harmful in the context of body image and ED, including why and how this harm occurs; and (2) to explore how technology might help mitigate these effects.

We engaged n = 30 participants, including 12 interviews with experts by profession (n = 2 ED support service providers and n = 10 body image and ED experts), and five focus groups with experts by lived experience (n = 18 people with lived experience of ED).

Using the Framework Method guided by inductive thematic analysis, we developed six prominent themes: (1) Spectrum of harmful and ambiguous content on social media, (2) The “echo chamber” of harmful content amplified by social media algorithms, (3) Balancing safety, freedom and responsibility in social media moderation, (4) Shared responsibility and collaboration for safer social media environments, (5) The role of representation and diversity in social media recovery and support, and (6) Harnessing digital innovation to reduce harm on social media. We developed an eight-category framework of harmful social media content, offering an underlying contextual understanding of harmful content and guidance for harm-reducing technologies.

Manual safeguards place significant responsibility on users. This work supports informed distinctions between harmful, ambiguous and safe content and provides design insights for classification systems and adaptable automated moderation.

The online version contains supplementary material available at 10.1186/s40337-025-01504-7.

Constant exposure to idealised beauty standards online can lead to negative body image, unhealthy behaviours and eating disorders (ED). To reduce these harms, we need effective ways to identify and moderate harmful content. However, first, we must understand why and how this content causes harm. We consulted n = 30 participants, including interviews with n = 12 professionals in ED support and research, and five focus groups with n = 18 individuals with lived experience of ED. Six major themes were developed using the Framework Method guided by inductive thematic analysis: the range of harmful and ambiguous content, social media’s contribution to an ED “echo chamber,” challenges in content moderation, the need for shared responsibility in creating safer spaces, the importance of diverse representation and the potential for technology to help mitigate harm. From this analysis, we developed eight categories of harmful social media content related to body image and ED, clarifying how individuals may be negatively affected and providing structure for future interventions. Participants highlighted the powerful influence of algorithms in promoting harmful content and called for shared responsibility among users, content creators, platforms, and policymakers. The findings offer guidance for designing technologies that mitigate social media harm.

The online version contains supplementary material available at 10.1186/s40337-025-01504-7.