Activities on Facebook Reveal the Depressive State of Users

Sungkyu Park, Sang Won Lee, Jinah Kwak, Meeyoung Cha, Bumseok JeongJournal of Medical Internet Research 2013 · DOI

Abstract

Our results using EmotionDiary demonstrated that the more depressed one is, the more one will read tips and facts about depression. We also confirmed depressed individuals had significantly fewer interactions with others (eg, decreased number of friends and location tagging). Our app, EmotionDiary, can successfully evaluate depressive symptoms as well as provide useful tips and facts to users. These results open the door for examining Facebook activities to identify depressed individuals. We aim to conduct the experiment in multiple cultures as well.

Introduction

Depression is a leading cause of disability worldwide, yet early detection remains a massive hurdle. Traditional screening campaigns often suffer from "participation bias"—the individuals most in need of help are often the least likely to attend a physical screening due to the very symptoms of the disorder, such as lethargy or social withdrawal.

The Facebook application interface for EmotionDiary, showing the integration within the social platform.

This research explores whether the "digital footprints" left on Online Social Networks (OSNs) like Facebook can provide a low-cost, non-intrusive way to identify individuals at risk. While previous research focused heavily on linguistic analysis (text mining), this study shifts the focus toward social features (friend counts, tagging) and in-app behavioral patterns (information-seeking behavior).

Methodology: The EmotionDiary Platform

To investigate these digital markers, researchers developed EmotionDiary, a Facebook Web application designed to serve as a proxy for mental health screening. The study followed a multi-step pipeline:

  1. Data Harvesting: With user consent, the app culled demographic data and social activity (likes, groups, friends, location tags) via the Facebook API.
  2. Psychometric Assessment: Users completed two gold-standard self-report scales: the CES-D (Center for Epidemiological Studies-Depression) and the BDI (Beck’s Depression Inventory).
  3. Gamified Engagement: The app included a points system and a library of 100 "tips and facts" about depression. The researchers tracked how many of these a user viewed.
  4. Statistical Validation: Spearman rank correlations linked Facebook features to depression scores, and Mann-Whitney U tests compared "probably depressed" vs. "nondepressed" groups.
  5. Clinical Ground Truth: A psychiatrist conducted face-to-face interviews with a small subset of high-scoring participants to validate the app's findings.
The EmotionDiary interface welcoming a user and explaining the study's purpose to assess emotional levels.

Data and Participants

The study utilized a primary dataset of 55 Facebook users recruited from KAIST (a major research university in South Korea). The cohort consisted of 40 males and 15 females, with a mean age of 24.43. The dataset included 11 social activity features (e.g., relationship status, pending requests, location tags) and 4 app-generated features (CES-D/BDI scores, tips viewed, and app points). To avoid skewing behavioral data, participants with more than 1,000 friends were excluded as "microcelebrities."

Results: Behavioral Markers of Depression

The study identified several key digital biomarkers that correlate with depressive states:

  • Information Seeking: One of the most significant findings was that depressed individuals were much more likely to read the "tips and facts" about depression. The number of app tips viewed and total app points were positively correlated with the CES-D scale ($P=.04$ for both). When comparing the "probably depressed" group to the nondepressed group, the difference in tips viewed was highly significant ($P=.01$).
  • Social Withdrawal: The social withdrawal associated with depression manifested physically in the digital realm. Depressed users had significantly fewer location tags ($P=.045$), suggesting they were visiting fewer new places or lacked the motivation to broadcast their location.
  • Friendship Networks: The number of friends also showed a marginally significant negative correlation with depression ($P=.08$). Nondepressed users had a mean of ~338 friends and ~14.5 location tags, while depressed users averaged ~254 friends and ~9.6 location tags.

Discussion and Future Directions

The authors argue that while standard Facebook features like total friend count are cumulative and change slowly, app-specific behaviors (like points earned or tips viewed) act as acute state markers, reflecting the user's current emotional state.

However, the study faces limitations, including a small sample size (N=55) skewed toward male engineering students in South Korea, which limits generalizability. Future work should focus on longitudinal tracking to see if sudden drops in activity can predict the onset of depressive episodes. Additionally, combining these social activity features with natural language processing (NLP) of status updates could create a more robust predictive model for mental health screening.

Summary updated

Abstract

Our results using EmotionDiary demonstrated that the more depressed one is, the more one will read tips and facts about depression. We also confirmed depressed individuals had significantly fewer interactions with others (eg, decreased number of friends and location tagging). Our app, EmotionDiary, can successfully evaluate depressive symptoms as well as provide useful tips and facts to users. These results open the door for examining Facebook activities to identify depressed individuals. We aim to conduct the experiment in multiple cultures as well.
BibTeX
@article{park2013activities,
  author = {Sungkyu Park and Sang Won Lee and Jinah Kwak and Meeyoung Cha and Bumseok Jeong},
  title = {{Activities on Facebook Reveal the Depressive State of Users}},
  journal = {Journal of Medical Internet Research},
  year = {2013},
  doi = {10.2196/jmir.2718},
}
Figure