New
(2025) AI Processed Human Approved
Body Image Distress and Cyberbullying Among Adolescent Girls in the U.S.
This study employs qualitative content analysis and text mining techniques to examine the extent, sentiment, and thematic structure of cyberbullying targeted at adolescent girls' body image on TikTok. Analyzing 80,954 public comments across 50 TikTok videos, the research identifies specific online harassment behaviors and discourse patterns in appearance-centered content.
Problem
Appearance-related cyberbullying on social media contributes significantly to body dissatisfaction, depression, and disordered eating among adolescent girls in the United States. Federal regulatory agencies and digital safety initiatives require platform-specific evidence to design targeted safety policies and youth mental health interventions.
Outcome
- Cyberbullying constituted 7.4% of total analyzed interactions, whereas non-cyberbullying comments comprised 92.6%.
- Body shaming was the most prevalent cyberbullying behavior (41.68%), followed by insults (24.78%) and teasing (21.26%).
- Neutral sentiment predominated user responses, followed by positive sentiment, with negative sentiment representing the smallest proportion.
- Topic modeling revealed five primary themes: Appearance Monitoring, Body Commentary, Positive Affirmations, Social Comparison, and Identity Confusion.
- Body shaming was the most prevalent cyberbullying behavior (41.68%), followed by insults (24.78%) and teasing (21.26%).
- Neutral sentiment predominated user responses, followed by positive sentiment, with negative sentiment representing the smallest proportion.
- Topic modeling revealed five primary themes: Appearance Monitoring, Body Commentary, Positive Affirmations, Social Comparison, and Identity Confusion.
What it means for you
- CIO / IT Executive: Initiate a review of existing content moderation algorithms and reporting mechanisms to assess their effectiveness in detecting and flagging body shaming, insults, and teasing comments related to appearance on platforms frequented by adolescent girls, aiming to prioritize enhancements for these specific issue types.
- IT Manager: Schedule a technical deep-dive meeting with the content moderation team to present the research findings on the prevalence of body shaming and insults on TikTok, and collaboratively identify specific keyword filters and pattern recognition strategies to implement within current moderation tools to better catch these behaviors.
- Business Strategist: Develop a proposal for a new product feature or marketing campaign that leverages positive affirmations and counter-narratives to combat social comparison and identity confusion related to body image on social media platforms, referencing the research's thematic analysis of positive affirmations as a potential intervention.
- Researcher: Begin drafting a follow-up research proposal that specifically investigates the effectiveness of different types of content moderation (e.g., AI-driven vs. human) in mitigating body shaming and insult-based cyberbullying on TikTok, and proposes a quantitative analysis of user sentiment shifts post-intervention.
- Policymaker: Draft a preliminary policy brief for the relevant government agency, highlighting the study's quantitative data on the percentage of cyberbullying and the specific themes of harassment (body shaming, insults), and outlining the urgent need for platform-specific regulations and youth mental health intervention funding to address appearance-centered online harassment.
Transcript
Host: Welcome to A.I.S. Insights — powered by Living Knowledge. I'm your host, Anna Ivy Summers. Today, we're diving into a critical new study titled "Body Image Distress and Cyberbullying Among Adolescent Girls in the U.S." Joining me to break down the technical details and real-world implications is our expert analyst, Alex Ian Sutherland. Welcome, Alex.
Expert: Thanks, Anna. Great to be here. This study offers an eye-opening look at how appearance-based harassment manifests on TikTok, analyzing real user interactions at scale to help us understand both the social and technological challenges around youth online safety.
Host: Let's start with the problem this study addresses. Social media platforms have faced intense public and regulatory scrutiny over their impact on teen mental health. What specific challenges are we seeing here?
Expert: The core issue is that appearance-focused content on platforms like TikTok often triggers targeted harassment. For adolescent girls in the U.S., body-related cyberbullying is directly linked to body dissatisfaction, anxiety, depression, and disordered eating behaviors. Federal agencies, including the FTC, alongside landmark digital policy initiatives like the Kids Online Safety Act, or KOSA, are actively pushing for platform-specific evidence to figure out how to mitigate these harms.
Host: That brings us to how the researchers gathered this evidence. Alex, how did the study analyze such a dynamic platform?
Expert: The study used a sophisticated mixed-methods approach combining qualitative analysis with text mining in R. The researchers sampled 50 TikTok videos posted by adolescent girls in the U.S. across eight popular appearance-related hashtags, including tags like #BodyCheck, #GlowUp, and #WhatIEatInADay. They extracted all public comments from those videos—a dataset of 80,954 user responses.
Host: Over 80,000 comments is a massive dataset. How did they categorize that volume of text?
Expert: They first developed an inductive qualitative codebook in NVivo by manually analyzing 2,000 random comments, establishing clear categories like body shaming, insults, teasing, and comparison attacks. They then scaled this up using keyword matching, rule-based classification, sentiment analysis using the Bing Liu lexicon, and Latent Dirichlet Allocation—or LDA topic modeling—to extract broad thematic structures.
Host: So what did the data actually reveal about cyberbullying on TikTok?
Expert: Interestingly, direct cyberbullying represented 7.4% of all analyzed comments, while non-cyberbullying comments made up 92.6%. While 7.4% might sound relatively low, across tens of thousands of interactions, it represents a substantial volume of harm.
Host: And when cyberbullying did occur, what form did it take?
Expert: Body shaming was by far the dominant tactic, accounting for 41.68% of all cyberbullying comments. Direct insults came in second at 24.78%, followed by teasing at 21.26%. Tactics like comparison attacks, nickname labeling, and clothing shaming occurred far less frequently than statistical expectations.
Host: What about sentiment? Were the comments overwhelmingly negative?
Expert: That was actually one of the most surprising outcomes of the study. Neutral sentiment predominated user responses, followed by positive sentiment, with negative sentiment constituting the smallest category overall.
Host: Why would negative sentiment be the smallest category if harassment is such a big issue?
Expert: The study points to linguistic adaptation, often called "algospeak." Users frequently adopt neutral, coded, or sarcastic language to bypass automated platform moderation filters while still delivering hostile messages. They manipulate phrasing so moderation systems don't auto-flag or delete their comments.
Host: That has huge implications for technology and product design. What are the key takeaways here for tech platforms and business leaders?
Expert: There are three critical takeaways. First, platform trust and safety teams cannot rely solely on simple keyword filters or basic sentiment models. Automated systems need contextual NLP that can detect subtle body shaming and coded language. Second, moderation works—the low percentage of explicit cyberbullying shows automated moderation suppresses overt abuse, but refined tools are needed to catch implicit tactics.
Host: And what about regulatory compliance?
Expert: That's the third point. Under emerging frameworks like KOSA, platforms have a duty of care to protect minors from content that promotes harm. The study shows that 92.6% of interactions are benign or positive, meaning targeted moderation focusing on that specific 7.4% of harmful content can make platforms significantly safer without degrading overall user engagement or growth.
Host: That makes total sense. Smarter safety engineering creates a safer ecosystem while preserving user engagement. Alex, thank you for walking us through these insights today.
Expert: It was my pleasure, Anna.
Host: And thank you to our listeners for tuning into A.I.S. Insights — powered by Living Knowledge. Be sure to subscribe for more deep dives into the intersection of technology, policy, and market trends. Until next time.
Expert: Thanks, Anna. Great to be here. This study offers an eye-opening look at how appearance-based harassment manifests on TikTok, analyzing real user interactions at scale to help us understand both the social and technological challenges around youth online safety.
Host: Let's start with the problem this study addresses. Social media platforms have faced intense public and regulatory scrutiny over their impact on teen mental health. What specific challenges are we seeing here?
Expert: The core issue is that appearance-focused content on platforms like TikTok often triggers targeted harassment. For adolescent girls in the U.S., body-related cyberbullying is directly linked to body dissatisfaction, anxiety, depression, and disordered eating behaviors. Federal agencies, including the FTC, alongside landmark digital policy initiatives like the Kids Online Safety Act, or KOSA, are actively pushing for platform-specific evidence to figure out how to mitigate these harms.
Host: That brings us to how the researchers gathered this evidence. Alex, how did the study analyze such a dynamic platform?
Expert: The study used a sophisticated mixed-methods approach combining qualitative analysis with text mining in R. The researchers sampled 50 TikTok videos posted by adolescent girls in the U.S. across eight popular appearance-related hashtags, including tags like #BodyCheck, #GlowUp, and #WhatIEatInADay. They extracted all public comments from those videos—a dataset of 80,954 user responses.
Host: Over 80,000 comments is a massive dataset. How did they categorize that volume of text?
Expert: They first developed an inductive qualitative codebook in NVivo by manually analyzing 2,000 random comments, establishing clear categories like body shaming, insults, teasing, and comparison attacks. They then scaled this up using keyword matching, rule-based classification, sentiment analysis using the Bing Liu lexicon, and Latent Dirichlet Allocation—or LDA topic modeling—to extract broad thematic structures.
Host: So what did the data actually reveal about cyberbullying on TikTok?
Expert: Interestingly, direct cyberbullying represented 7.4% of all analyzed comments, while non-cyberbullying comments made up 92.6%. While 7.4% might sound relatively low, across tens of thousands of interactions, it represents a substantial volume of harm.
Host: And when cyberbullying did occur, what form did it take?
Expert: Body shaming was by far the dominant tactic, accounting for 41.68% of all cyberbullying comments. Direct insults came in second at 24.78%, followed by teasing at 21.26%. Tactics like comparison attacks, nickname labeling, and clothing shaming occurred far less frequently than statistical expectations.
Host: What about sentiment? Were the comments overwhelmingly negative?
Expert: That was actually one of the most surprising outcomes of the study. Neutral sentiment predominated user responses, followed by positive sentiment, with negative sentiment constituting the smallest category overall.
Host: Why would negative sentiment be the smallest category if harassment is such a big issue?
Expert: The study points to linguistic adaptation, often called "algospeak." Users frequently adopt neutral, coded, or sarcastic language to bypass automated platform moderation filters while still delivering hostile messages. They manipulate phrasing so moderation systems don't auto-flag or delete their comments.
Host: That has huge implications for technology and product design. What are the key takeaways here for tech platforms and business leaders?
Expert: There are three critical takeaways. First, platform trust and safety teams cannot rely solely on simple keyword filters or basic sentiment models. Automated systems need contextual NLP that can detect subtle body shaming and coded language. Second, moderation works—the low percentage of explicit cyberbullying shows automated moderation suppresses overt abuse, but refined tools are needed to catch implicit tactics.
Host: And what about regulatory compliance?
Expert: That's the third point. Under emerging frameworks like KOSA, platforms have a duty of care to protect minors from content that promotes harm. The study shows that 92.6% of interactions are benign or positive, meaning targeted moderation focusing on that specific 7.4% of harmful content can make platforms significantly safer without degrading overall user engagement or growth.
Host: That makes total sense. Smarter safety engineering creates a safer ecosystem while preserving user engagement. Alex, thank you for walking us through these insights today.
Expert: It was my pleasure, Anna.
Host: And thank you to our listeners for tuning into A.I.S. Insights — powered by Living Knowledge. Be sure to subscribe for more deep dives into the intersection of technology, policy, and market trends. Until next time.