Perception Differences between the Depressed and Non-Depressed Users in Twitter

Minsu Park, David W. McDonald, Meeyoung Cha

Abstract

One's state of mind will influence her perception of the world and people within it. In this paper, we explore attitudes and behaviors toward online social media based on whether one is depressed or not. We conducted semi-structured face-to-face interviews with 14 active Twitter users, half of whom were depressed and the other half non-depressed. Our results highlight key differences between the two groups in terms of perception towards online social media and behaviors within such systems. Non-depressed individuals perceived Twitter as an information consuming and sharing tool, while depressed individuals perceived it as a tool for social awareness and emotional interaction. We discuss several design implications for future social networks that could better accommodate users with depression and provide insights towards helping depressed users meet their needs through online social media.

Introduction

While most research focuses on detecting depression through social media data, this paper explores how depression fundamentally alters how users perceive and interact with platforms like Twitter. The study reveals a stark divide: non-depressed users view Twitter as an information-sharing tool, while depressed users utilize it as a vital space for social awareness and emotional consolation.

Problem Statement

Depression is a leading cause of disability worldwide, and researchers have spent years developing NLP models to identify depressive signals in user-generated text. However, there is a significant gap in understanding the user experience of depressed individuals. The authors argue that we cannot build effective digital interventions if we don't understand how depression changes a person's "state of mind" regarding social information. This paper moves beyond mere detection to explore the psychological relationship between the user and the platform.

Methods

The researchers employed a mixed-methods approach to gain both depth and scale. They conducted 90-minute semi-structured, face-to-face interviews with 14 active Twitter users in South Korea. Participants were screened using the CES-D (Center for Epidemiologic Studies Depression Scale) to form two groups: 7 clinically depressed and 7 non-depressed individuals.

Table 1: Participant demographics and primary social networks.

To triangulate interview findings, they analyzed the "consumption" habits of the participants by collecting over 3 million tweets from the accounts they followed. Using LIWC (Linguistic Inquiry and Word Count), they measured the "affectiveness" (emotionality) of the content consumed by both groups.

Key Insights

  • Information vs. Emotion: Non-depressed users are "information hunters"—they follow authoritative sources and news outlets to stay informed. Depressed users are "emotion seekers"—they follow people who share "everyday life episodes" to feel a sense of normalcy and empathy.
  • The "Void" as a Feature: Depressed users described Twitter as "shouting into a void." Paradoxically, this was a positive. The perceived indifference of a loosely connected network lowered the barrier for self-disclosure. They felt safer sharing dark thoughts on Twitter than on Facebook, where close offline friends might "overreact."
  • Social Awareness as Therapy: For depressed users, seeing mundane tweets (e.g., "I'm eating," "Goodnight") served as a "social awareness stream." It helped them synchronize their biological clocks with society and reduced "brooding" by reminding them they are part of a larger world.
  • Volatility vs. Archiving: Non-depressed users often used third-party tools to archive tweets. Depressed users loved the volatility; the fact that tweets "flow away and disappear" made it easier to disclose temporary emotional states without them becoming a permanent digital record.

Key Results

Quantitative analysis confirmed the interview findings: the accounts followed by depressed users produced content with significantly higher emotional affectiveness (mean = 4.67) compared to the non-depressed group (mean = 3.42, p < 0.01). Furthermore, unfollowing logic differed: non-depressed users unfollowed based on volume, while depressed users unfollowed based on sentiment—specifically "downers" or highly critical content that might trigger a negative spiral.

Design Implications and Future Work

The authors suggest several design directions to better accommodate depressed users:

  • Social Squelch: Filtering tools to automatically hide "gloomy" or triggering content.
  • Sentiment-Based Recommendations: Suggesting friends based on "how they feel" (sentiment similarity) rather than just "who you know."
  • Adjustable Volatility: Allowing users to choose how long a message persists based on its emotional weight.

The study’s primary limitation is its small sample size (14 participants), which is standard for qualitative work but limits generalizability. Future research should explore these behaviors across different cultural contexts.

Summary updated

Abstract

One's state of mind will influence her perception of the world and people within it. In this paper, we explore attitudes and behaviors toward online social media based on whether one is depressed or not. We conducted semi-structured face-to-face interviews with 14 active Twitter users, half of whom were depressed and the other half non-depressed. Our results highlight key differences between the two groups in terms of perception towards online social media and behaviors within such systems. Non-depressed individuals perceived Twitter as an information consuming and sharing tool, while depressed individuals perceived it as a tool for social awareness and emotional interaction. We discuss several design implications for future social networks that could better accommodate users with depression and provide insights towards helping depressed users meet their needs through online social media.
BibTeX
@inproceedings{park2021perception,
  author = {Minsu Park and David W. McDonald and Meeyoung Cha},
  title = {{Perception Differences between the Depressed and Non-Depressed Users in Twitter}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2021},
  doi = {10.1609/icwsm.v7i1.14425},
}
Figure