Examining the Unique Online Risk Experiences and Mental Health Outcomes of LGBTQ+ versus Heterosexual Youth

Objective

Social media platforms like Instagram have become essential for LGBTQ+ youth, providing spaces for identity exploration, social support, and community engagement. However, LGBTQ+ youth face heightened risks of harassment, cyberbullying, and other online threats compared to their heterosexual peers. This study aimed to investigate the following research questions (RQs):

Do LGBTQ+ youth experience more frequent and/or severe online risks in their private social media conversations compared to heterosexual youth?
How do negative online experiences impact the mental health of LGBTQ+ versus heterosexual youth?
What are the unique characteristics of online risks encountered by LGBTQ+ youth in private messaging spaces?

Data Collection and Risk Flagging

Participant Demographics and Mental Health Assessment:
- Youth aged 13–21 completed a web-based survey reporting demographic details (gender and sexual identity).
- Mental health and well-being assessed using validated scales:
  - Short Warwick-Edinburgh Mental Well-Being Scale (SWEMWBS)
  - Patient Health Questionnaire (PHQ-9) for depression
  - Inventory of Statements about Self-Injury (ISAS)
- All constructs were measured on a 5-point Likert scale.
Data Upload and Review Process:
- Participants uploaded their Instagram data via a secure PHP-based system with AWS for data encryption and storage.
- Data was uploaded as JSON files, converted, and securely stored on AWS.
Message Review and Risk Classification:
- Participants reviewed their Instagram direct messages (DMs) in reverse chronological order through a web interface.
- Messages marked as risky were categorized by type (e.g., sexual content, harassment, self-injury) and severity:
  - Low: Discomfort without harm
  - Medium: Potential harm
  - High: Actual harm
Dataset Overview:
- Included 32,055 DM conversations, totaling over 6 million messages.
- 2,515 conversations flagged as risky, with 3,023 messages categorized by risk type and severity:
  - 402 high-level risks
  - 821 medium-level risks
  - 1,813 low-level risks.

Methodology

We employed a mixed-methods approach, combining both quantitative and qualitative analyses to comprehensively examine the online risk experiences and mental health outcomes of LGBTQ+ versus heterosexual youth.

Chi-Square Tests

To compare online risk experiences between LGBTQ+ and heterosexual youth, specifically regarding both the types and severity of flagged risks, we used chi-square tests of independence. This test is ideal for examining relationships between categorical variables, enabling us to capture distributional differences in flagged risk types and severity levels between LGBTQ+ and heterosexual youth. By examining standardized residuals, we could isolate the most impactful differences, revealing whether LGBTQ+ youth experienced higher or more severe types of risks.

Variables:

Sexual identity: LGBTQ+ or heterosexual.
Types of flagged messages: Categories such as sexual content, harassment, self-injury.
Risk severity levels: Low, medium, and high, based on participants' perceptions of each flagged message.

Multiple Linear Regression Analysis

To understand how different types of online risks affected mental health outcomes, we used multiple linear regression models. This approach was essential for analyzing the predictive relationship between flagged message types and mental health measures, while also accounting for the moderating role of sexual identity. We built separate regression models for each mental health outcome, adding flagged message types as predictors. Including sexual identity as an interaction term allowed us to see if message types impacted LGBTQ+ youth differently. This approach helped us understand the differential effects of online risks on mental health across groups.

Variables:

Predictor variables: Types of flagged messages, including sexual messages, harassment, and self-injury.
Dependent variables: Three key mental health outcomes:
- Depression: Measured using a modified version of the Patient Health Questionnaire (PHQ-9).
- Self-Harm and Injury: Assessed with the Inventory of Statements about Self-Injury (ISAS).
- Mental Well-Being: Measured using the Short Warwick-Edinburgh Mental Well-Being Scale (SWEMWBS).
Interaction term: Sexual identity (LGBTQ+ or heterosexual) to explore whether the relationship between risk types and mental health varied by group.

Welch’s Two-Sample t-Test

To compare mental health scores between LGBTQ+ and heterosexual youth, we used Welch’s two-sample t-test. This variation of the t-test is suited for comparing means between two groups with unequal variances, which was a condition present in our dataset. The test compared the mean scores of mental health outcomes between the two groups.

Variables:

Group: LGBTQ+ or heterosexual.
Mental health measures: Mean scores on depression, self-harm, and mental well-being.

Qualitative Analysis: Thematic Analysis of Flagged Messages

For our third research question, we adopted a qualitative approach to understand the specific nature of online risks faced by LGBTQ+ youth. We conducted a thematic analysis on flagged Instagram direct messages, allowing us to identify patterns and themes within the content. This method extended beyond quantitative frequency counts to capture the deeper, lived experiences of LGBTQ+ youth in private online spaces.
We began by familiarizing ourselves with 1,468 flagged messages across 808 conversations, creating initial codes based on message content. Two researchers refined these codes iteratively, organizing them into thematic categories that reflected LGBTQ+ youth experiences.

Findings

RQ1: Frequency and Severity of Online Risks for LGBTQ+ Youth
- Higher Severity but Not Frequency: LGBTQ+ youth did not report more online risk experiences overall compared to heterosexual youth. However, the risks they flagged were of significantly higher severity, often involving sexual or self-injurious content.
- Greater Exposure to Sexual and Self-Injury Content: LGBTQ+ youth faced a higher prevalence of messages with sexual content or self-injurious themes, contrasting with heterosexual youth, who frequently reported low-risk content like general harassment.
RQ2: Mental Health Implications of Negative Online Experiences
- Worse Mental Health Outcomes for LGBTQ+ Youth: LGBTQ+ youth reported higher depression and self-harm levels and lower well-being than heterosexual peers, confirmed by regression analysis and Welch’s t-tests, indicating their heightened vulnerability to online risks.
- Stronger Impact of Harassment: Harassment was more strongly linked to self-harm behaviors among LGBTQ+ youth, suggesting identity-based abuse intensifies mental health challenges for this group.
- Complex Role of Sexual Content: Surprisingly, some LGBTQ+ youth associated sexual content with slightly improved well-being, indicating that certain interactions, despite risks, may offer affirming experiences for identity exploration.
RQ3: Unique Nature of Online Risks in LGBTQ+ Youth’s Private Conversations
- Sexual Undertones and Identity Exploration: Many risky conversations flagged by LGBTQ+ youth had sexual undertones, even when intent was harassment, posing both threats and opportunities for identity exploration.
- Supportive Peer Interactions: LGBTQ+ youth frequently used DMs to share struggles with peers, reflecting the dual role of social media in amplifying risks while offering supportive networks.

These findings emphasize the need for targeted interventions that balance mitigating harmful content with preserving positive online spaces for identity exploration and peer support for LGBTQ+ youth.

CDT Blog post

Read paper

Empirical Evidence for Reciprocal Radicalization: Conformity of User Comment, Peer Feedback, and Community Emotion in Extreme Misogynistic Social Media Groups

Objective
This study explores the theory of reciprocal radicalization, which posits that extremist communities and individual users mutually influence each other’s beliefs, emotions, and actions. The research seeks to address the following research question: Do Reddit users who engage with extreme misogynistic material at higher posting frequencies demonstrate evidence of reciprocal radicalization, specifically contributing to and receiving distinctly negative emotional content?

Data Collection and Processing

Data Source: Pushshift archival dataset provided historical Reddit data, preserving deleted/moderated posts, although lacking upvote counts.
Data Sampling:
- User Selection: Initial pool of 12,000 users actively engaged with a quarantined misogynistic subreddit.
- Final Sample: Randomly selected 1,000 users with consistent activity (average of one post per week) over the past two years.
- Data Extraction: Reddit’s API was used to collect each user's full comment history, capturing over 1.8 million comments.
Variables Collected:
- Comment Features:
  - Text content of each comment
  - Timestamp of the comment
  - Score (difference between upvotes and downvotes)
  - Subreddit name
  - Engagement frequency (time interval since last post, normalized from -1 to 1 for interpretation)
  - Total comments made by each user in the extremist subreddit.
- Reply Features:
  - Text content of replies received on each comment
  - Count of replies per comment.
- Submission Features:
  - Submission text where the comment was posted
  - Submission timestamp
  - Total comments in the submission thread
  - Submission score.
Data Normalization:
- Engagement Frequency: Calculated as the time between consecutive comments and normalized from -1 to 1, where higher values indicate more frequent posting.
- Sentiment Scores: LIWC was applied to categorize words in comments and replies as positive (e.g., “happy”) or negative (e.g., “hate”), with scores representing the proportion of emotional words per text.

Methodology

Multiple Linear Regression (MR) Models: A series of MR models were used to examine how users’ Reddit engagement frequency was predicted by comment features, peer feedback, and submission features. This statistical approach was chosen to quantify relationships between the frequency of user engagement and emotional content expressed in comments, facilitating the identification of significant predictors while controlling for potential confounding variables.

Findings

User Engagement Patterns: Users with higher engagement rates not only contributed more comments within the extremist subreddit but also interacted with a wider range of subreddits, challenging previous theories that suggest radicalization leads to isolation within extremist circles.
Emotional Content: Higher engagement with extremist content was associated with increased expression of negative emotions and decreased expression of positive emotions in users' comments. Users also received predominantly negative feedback from peers, reinforcing the emotional tone of their contributions.
Reciprocal Radicalization Evidence: The study provides empirical support for the reciprocal radicalization theory, indicating emotional feedback loops where users reciprocally create and receive distinctly negative content.

The findings highlight the dynamic interplay between individual users and extremist communities, suggesting that radicalization encompasses not only ideological shifts but also the adoption of negative emotional expressions. The study calls for further research to understand the evolving nature of reciprocal radicalization processes and the potential implications for online extremism across various contexts.

Read paper

Bottom-up psychosocial interventions for interdependent privacy: Effectiveness based on individual and content differences

Although a great deal of research has examined interventions to help users protect their own information online, less work has examined methods for reducing interdependent privacy (IDP) violations on social media (i.e., sharing of other people's information). This study tested the effectiveness of concept-based (i.e., general information), fact-based (i.e., statistics), and narrative-based (i.e., stories) educational videos in altering IDP-relevant attitudes and multimedia sharing behaviors. Our study revealed concept and fact videos reduced sharing of social media content that portrayed people negatively. The narrative intervention backfired and increased sharing among participants who did not believe IDP violations to be especially serious; however, the narrative intervention decreased sharing for participants who rated IDP violations as more serious. Notably, our study found participants preferred narrative-based interventions with real world examples, despite other strategies more effectively reducing sharing. Implications for narrative transportation theory and advancing bottom-up (i.e., user-centered) psychosocial interventions are discussed.

Read paper

Publications

Google Scholar Page

Page updated

Google Sites

Report abuse