FoodCensor: Promoting Mindful Digital Food Content Consumption for People with Eating Disorders

Title of the Paper

FoodCensor: Promoting Mindful Digital Food Content Consumption for People with Eating Disorders

Paper Information

• Research Domain: Human-Computer Interaction, Digital Intervention Design, Eating Disorder Behavior Improvement
• Keywords: Food content, eating disorders, intervention, binge eating, bulimia

Research Background and Problem Statement

• Identified Issues:

  • Food content displayed on digital media (e.g., cooking tutorials, mukbang videos) may exacerbate eating disorder behaviors, including binge eating and bulimia.
  • The enticing visual and auditory stimuli of digital food content can trigger unhealthy eating habits and worsen symptoms for certain populations with eating disorders.
  • Current solutions provide limited support for improving eating disorder behaviors, particularly in the context of digital content consumption.

• Significance:

  • The widespread prevalence of food content may lead to repeated exposure to triggers of eating disorders, potentially worsening mental health for affected individuals.

• Research Motivation and Related Work:

  • Many individuals with eating disorders consume digital food content in an automated and unconscious manner.
  • Dual-process theory offers a framework to distinguish between automated behaviors (System 1) and reflective behaviors (System 2), which can be applied to improving digital media content consumption.
  • Digital intervention tools have demonstrated research value in other domains of behavior change, but there has been insufficient exploration of their application to digital food content consumption among individuals with eating disorders.

Solution

• Proposed Approach: The FoodCensor intervention system assists individuals with eating disorders in improving their digital content consumption through two dimensions:

  1. Hiding passive exposure to digital food content: Food videos on media platforms like YouTube are covered with black overlays, encouraging users to consciously choose whether to view them.
  2. Active intervention during user searches: Reflective questions and warnings about the negative consequences of eating disorders are presented when users search for food-related content.

• Innovative Features:

  • The system is the first to detect and regulate digital content consumption in real time from the user side, rather than relying on the content model of media platforms.
  • By combining dual-process theory, the system hides stimuli and prompts reflection to foster more conscious and healthier usage behaviors.

• Implementation Steps and Key Technologies:

  • Technical Implementation: Development of an Android application and Chrome extension that track UI component changes using Accessibility API and DOM tree technologies to identify and overlay food content.
  • Content Detection Method: A dictionary-based approach (keyword matching) is used to detect food-related content.
  • Intervention Mechanisms:
    • Reflective questions are presented via pop-ups when users search for food content.
    • Image warnings highlight the potential negative psychological and physiological consequences of habitual behaviors.

Research Outcomes

• Specific Results:

  • Reduced exposure to digital food content: During the three-week experimental period, the number of suggested food videos and their view counts on YouTube significantly decreased for the experimental group.
  • Enhanced user reflection and behavioral awareness: Black overlays and reflective questions helped participants in the experimental group become more aware of whether their media consumption habits were healthy.
  • Improvement in eating disorder symptoms: Participants in the experimental group reported reductions in binge eating and vomiting episodes, along with greater decreases in EDE-Q (Eating Disorder Examination Questionnaire) scores compared to the control group.

• Advantages Over Existing Solutions:

  • Provides a user-side intervention model for content regulation, enabling personalized application and support for specific populations.
  • Beyond reducing exposure to food content, the system actively encourages users to weigh the potential pros and cons of their behaviors.

• Experimental Evaluation Results:

  • Food Content Exposure Detection: The food detection algorithm achieved an accuracy of 79-83%.
  • User Behavior Analysis: Experimental group participants expressed significant satisfaction with the intervention's effectiveness and willingness to use it actively.

• Limitations and Future Directions:

  • Limitations:
    • The system currently only covers the YouTube platform; other social media platforms (e.g., Instagram, TikTok) require further exploration.
    • The food content detection algorithm may encounter errors when identifying complex cases (e.g., ambiguous keywords in titles).
    • The study population was limited to Korean women, and further research is needed to address international and gender diversity.
  • Future Directions:
    • Design intervention systems that adapt to individual behavioral change stages, supporting transitions from reflection to long-term habit formation.
    • Explore user self-regulation features for content recommendation algorithms.
    • Expand content regulation to more platforms and devices, incorporating large-scale user feedback to improve algorithm detection capabilities.