Understanding Patterns of Anorexia Manifestations in Social Media Data with Deep Learning

Abstract

Eating disorders are a growing problem especially among young people, yet they have been under-studied in computational research compared to other mental health disorders such as depression. Computational methods have a great potential to aid with the automatic detection of mental health problems, but stateof-the-art machine learning methods based on neural networks are notoriously difficult to interpret, which is a crucial problem for applications in the mental health domain. We propose leveraging the power of deep learning models for automatically detecting signs of anorexia based on social media data, while at the same time focusing on interpreting their behavior. We train a hierarchical attention network to detect people with anorexia, and use its internal encodings to discover different clusters of anorexia symptoms. We interpret the identified patterns from multiple perspectives, including emotion expression, psycho-linguistic features and personality traits, and we offer novel hypotheses to interpret our findings from a psycho-social perspective. Some interesting findings are patterns of word usage in some users with anorexia which show that they feel less as being part of a group compared to control cases, as well as that they have abandoned explanatory activity as a result of a greater feeling of helplessness and fear.

Gender (Dis)identification in Anorexia: An Analysis of Tumblr and the Anonymity of Skinny Bodies

“…Paired with lower values for 'we' and "you" could indicate self-isolation and feel distanced from one's social environment. Uban et al (2021) and De Choudhury (2015) found similar results in their analyses. This interpretation is also supported by the social measures, which are again lower in the anorexia groups.…”

“…Prior versions of LIWC have been successfully applied on posts concerning eating disorders. The single features showed significant differences between two communities within eating disorders (De Choudhury, 2015) and between anorectic social media authors and those of a control group (Uban, Chulvi, & Rosso, 2021). LIWC-22 has been successfully applied on Tumblr posts to distinguish between eating disorder related trend hashtags (#Thinspiration and #Fitspiration), also in relation to Korean popular music (Achilles et al, 2023a).…”

A New Model That Figures out Sub-Emotionality and the Changes in It over Time

“…Now a days the cost of detecting or diagnosing mental illness have improved so by that people can't invest more on that and that cases remain undetected for that the author in this used social media as a platform to detect mental illness [1] This makes the claim that user personality profiles can be used to tailor content, optimize campaigns, and enhance online advertising. [2] They present analysis of 2,34,735 supporter messages is to discover different strategies which correspond with clinical outcomes [3] Non-linear methods for regression was applied for achieving strong correlation between predicted and actual user income [4] Here survey responses and status updates from 24,904 Face book are used to develop a regression model that predicts users degree of depression based on their Face book status update [5] The author in this detect and refer the user's as soon as possible to professional help so they used BoSE representation which represent social media documents by a set of fine grained emotions automatically using a lexical resource [6] social media provides a great opportunity to increase the available data to researchers for better information in health field for a new approach is done by this the data is gathered quickly and cheaply and inform the necessary ethical discussion [7] Eating disorder are complex in mental disorders and they are responsible for highest mortality rate among mental illness for this a snowball sampling method is developed for automatically gathering of individuals who self identify as eating disorder.…”

Detection of Anorexic Girls-In Blog Posts Written in Hebrew Using a Combined Heuristic AI and NLP Method

“…In 2020 and 2021, eRisk did not suggest any task or dataset relevant to Anorexia. Uban et al (2021) proposed a DL model to detect signs of people suffering from anorexia in social media. They also tried to explain the behavior of their model.…”