Inferring Opinions and Behavioral Characteristics of Gay Men with Large Scale Multilingual Text from Blued

Ge Huang, Mengsi Cai, Xin LüInternational Journal of Environmental Research and Public Health 2019 · DOI

TL;DR

This paper analyzes millions of posts and voting data from Blued, the world's largest gay social networking app, to understand the opinions and behaviors of gay men across different countries. By using machine learning and linguistic analysis, the study reveals that while most content is about daily life, sensitive topics like self-identification and sexual behavior elicit strong emotions, with notable differences between Chinese and non-Chinese users. This research highlights the potential of GSN apps as a valuable resource for public health interventions and social inclusion initiatives.

What Problem Does This Paper Address?

Understanding the diverse experiences and behaviors of sexual minorities, particularly gay men, is crucial for developing effective public health strategies and promoting social inclusion. Geosocial networking applications (GSN apps) like Blued have become popular platforms for gay men to socialize, seek partners, and find support. However, the vast amount of textual data generated on these platforms has largely been underutilized for research. This paper aims to bridge this gap by conducting a comprehensive content analysis of posts and opinions on Blued to infer and compare the opinions and behavioral characteristics of gay men globally, with a specific focus on identifying differences between users in China and other regions. This is particularly relevant given the rising rates of new HIV infections in the Asia-Pacific region, with a significant portion occurring among men who have sex with men (MSM).

Methods

The researchers employed a multi-faceted approach to analyze the data from Blued:

  • Data Collection and Filtering: They collected a large dataset of posts and voting items from Blued, distinguishing between Chinese and international users. This involved searching for specific hashtags and filtering out irrelevant or duplicate content.
  • Topic Classification: Posts were categorized into eight themes: Life, Entertainment, Social Networking, Relationship, Self-identification, Sexual Behavior, Work and Study, and Marriage and Family. This was done manually for Chinese posts and semi-automatically (with translation and machine learning) for non-Chinese posts.
  • Sentiment and Psychological Analysis: Computational linguistics programs (LIWC and TextMind) were used to analyze the sentiment (positive, negative, neutral) and emotional content (anxiety, anger, sadness) of the posts. This allowed for a deeper understanding of the emotional tone associated with different topics.
  • Voting Data Analysis: Voting items, which often touched upon sensitive topics, were manually categorized and analyzed to understand user viewpoints and attitudes on various issues.

Datasets

The primary dataset for this study was sourced from Blued, the world's largest gay social networking dating app.

  • Topic Data: Collected from January 2015 to February 2019, including 1186 topics (551 Chinese, 635 English) and over 1.7 million posts after filtering and de-duplication.
  • Voting Data: Collected from the "Viewpoint" module, comprising 114 voting items, over 5 million votes, and over 400,000 comments.

This dataset represents a novel and large-scale collection of user-generated content from a specific, often hard-to-reach, population.

Key Insights and Ideas

  • GSN Apps as a Rich Data Source: The study demonstrates that GSN apps like Blued are not just for social interaction but also serve as a valuable, albeit complex, source of data for understanding the lives and concerns of sexual minorities.
  • Cultural Nuances in Expression: Significant differences emerged in how gay men in China and other regions express themselves on sensitive topics. Chinese users tended to focus more on self-identification and sexual behavior when expressing negative emotions, while non-Chinese users showed more negative sentiment related to relationships.
  • The "Life" Category Dominance: Across both Chinese and English posts, "Life" was the most prevalent topic, suggesting a general desire for connection and sharing of everyday experiences, likely as a way to build community and seek attention in a space where open expression of identity might be challenging.
  • Balancing Privacy and Connection: The data suggests a tension between the desire for social connection and the need to protect privacy, especially concerning sensitive information like HIV status or sexual behavior.
  • Public Health Implications: The findings directly inform public health efforts by identifying potential risk behaviors (e.g., casual sex, drug use) and highlighting the need for targeted interventions and increased awareness campaigns, particularly regarding HIV prevention and sexual education.

Key Results

  • Topic Distribution: "Life" was the most common topic for both Chinese (over 60%) and English (74.9%) posts. Relationships were the second most popular topic for Chinese users (19.2%), while social networking and relationships were prominent for English users (7.4% and 7.2%, respectively).
  • Sentiment: Generally, posts expressed more positive than negative emotions. However, negative emotions were more prevalent in Chinese topics, particularly concerning self-identification and sexual behaviors.
  • Behavioral Indicators:
    • Only 50.52% of participants reported taking regular HIV tests.
  • 50.2% indicated they would have casual sex when single.
  • 35.8% admitted they might try drugs if invited by friends.
  • Negative Emotions: Sadness was the most common negative emotion in both groups. Chinese users expressed more anxiety about self-identification, while non-Chinese users showed more anxiety about sexual behaviors. Anger was more frequently expressed by Chinese users regarding sexual behavior, while non-Chinese users expressed more sadness in this area.
  • Voting Data Insights:
    • 72% of participants were willing to accept cosmetic surgery.
  • 64.6% had no clear standard for a boyfriend, often basing decisions on first appearances.
  • 62.1% believed in showing conjugal love on social networks, but 74.4% would not display affection in public.
  • 86.3% of Chinese participants would hide their sexual orientation at work.

Limitations and Future Work

  • Language Limitations: The study focused on Chinese and English, excluding many other languages spoken by Blued users, potentially limiting the global representativeness. Machine translation of non-English posts may have introduced inaccuracies.
  • Hashtag Bias: The analysis relied on hashtagged posts, potentially missing content without hashtags or private messages.
  • User Demographics: The findings might not represent older or more diverse gay communities, as GSN app users tend to be younger.
  • Topic Coverage: The hashtag list might not have captured all popular or sensitive topics.

The authors suggest that future research should:

  • Replicate this study with other social apps and networks to identify broader trends.
  • Explore how this type of content analysis can be used for targeted interventions.
  • Investigate the potential of GSN apps for monitoring gay behavior and identifying individuals at risk.
  • Promote public information about HIV and sexual education for adolescents in China.

Summary updated

Abstract

Gay men in many countries are increasingly using geosocial networking applications (GSN apps), thus offering new opportunities for understanding them. This paper provides a comprehensive content analysis of posts and opinions on Blued, the world's largest gay social networking dating app, to infer and compare opinions and behavioral characteristics of gay men in different countries. Machine learning and linguistic programming approaches were used to extract themes and analyze sentiments of posts. The results show that the majority of posts are related to daily life activities, and less are related to sensitive topics. While most posts are positive or neutral, negative emotions, including anxiety, anger, and sadness, are mainly distributed in posts related to self-identification and sexual behaviors in China and to relationships in other countries. Voting items indicate that only 50.52% of the participants will take regular HIV tests while 50.2% would have casual sex when they are single. Additionally, 35.8% of the participants may try drugs when invited by friends. Our findings suggest an opportunity and necessity for researchers and public health practitioners to use open source data on GSN apps and other social medias to inform HIV interventions and to promote social inclusion for sexual minorities.
BibTeX
@article{huang2019inferring,
  author = {Ge Huang and Mengsi Cai and Xin Lü},
  title = {{Inferring Opinions and Behavioral Characteristics of Gay Men with Large Scale Multilingual Text from Blued}},
  journal = {International Journal of Environmental Research and Public Health},
  year = {2019},
  doi = {10.3390/ijerph16193597},
}
Figure