Seeking Help, Facing Harm: Auditing TikTok's Mental Health Recommendations

• URL: http://arxiv.org/abs/2604.14832v1
• Date: Thu, 16 Apr 2026 10:10:03 GMT
• Title: Seeking Help, Facing Harm: Auditing TikTok's Mental Health Recommendations
• Abstract summary: We conduct a 7-day audit of TikTok's "For You" page using 30 fresh accounts and LLM-guided agents.<n>Across 8,727 recommended videos, interaction behavior dominates exposure outcomes.<n>These findings suggest limited sensitivity to user intent signals in TikTok's recommendations.
• Score: 1.2744523252873352
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Recommender systems on social media increasingly mediate how users encounter mental health content, yet it remains unclear whether they distinguish help-seeking from distress expression. We conduct a controlled 7-day audit of TikTok's "For You" page using 30 fresh accounts and LLM-guided agents that vary initial search framing (distress- vs. help-initiated) and interaction strategy (engaged, avoidant, passive). Across 8,727 recommended videos, interaction behavior dominates exposure outcomes: engagement rapidly saturates feeds with mental health content (~45% of daily recommendations), while avoidance and passive viewing reduce but do not eliminate exposure (~11-20%). Search framing mainly shifts composition rather than volume--help-initiated searches yield more potentially supportive material, yet potentially harmful content persists at low but non-zero levels, including content in the Suicide/Self-Harm category. These findings suggest limited sensitivity to user intent signals in TikTok's recommendations and motivate context-aware safeguards for sensitive topics.

Related papers

• The Tone of Awareness: Topic, Sentiment, and Toxicity Maps During Mental Health Month on TikTok [1.0210993177391716]
  We collect the content of 28,341 TikTok videos and 80,130 comments from Mental Health Awareness Month (May) in 2023 and 2024 via the TikTok Research API.<n>We characterize "tone" as the emotional and interpersonal framing of mental health discourse.<n>We find a stable set of recurring themes across years, spanning clinical conditions, emotional disclosure, self-care, and campaign-oriented content.
  arXiv  Detail & Related papers  (2026-06-11T17:04:11Z)
• Transparent and Controllable Recommendation Filtering via Multimodal Multi-Agent Collaboration [48.028080782524064]
  We introduce a novel framework that integrates end-to-cloud collaboration, multimodal perception, and multi-agent orchestration.<n>Our system employs a fact-grounded adjudication pipeline to eliminate inferential hallucinations.<n>It constructs a dynamic, two-tier preference graph that allows for explicit, human-in-the-loop modifications.
  arXiv  Detail & Related papers  (2026-04-19T14:19:28Z)
• Attention at Rest Stays at Rest: Breaking Visual Inertia for Cognitive Hallucination Mitigation [50.51650162235191]
  We propose a training-free Inertia-aware Visual Excitation (IVE) method that breaks this inertial pattern by modeling cognitive inference as the dynamic responsiveness of visual attention.<n>IVE is effective across various base MLLMs and multiple hallucination benchmarks, particularly for cognitive hallucinations.
  arXiv  Detail & Related papers  (2026-04-02T12:51:07Z)
• Disentangling Prompt Element Level Risk Factors for Hallucinations and Omissions in Mental Health LLM Responses [11.099638001699276]
  Mental health concerns are often expressed outside clinical settings, including in high-distress help seeking.<n>We introduce UTCO, a prompt construction framework that represents an inquiry as four controllable elements for systematic stress testing.<n>Hallucinations occurred in 6.5% of responses and omissions in 13.2%, with omissions concentrated in crisis and suicidal ideation prompts.
  arXiv  Detail & Related papers  (2026-03-10T17:40:24Z)
• The Alignment Waltz: Jointly Training Agents to Collaborate for Safety [80.16102254128518]
  WaltzRL is a novel multi-agent reinforcement learning framework that formulates safety alignment as a collaborative, positive-sum game.<n>At the core of WaltzRL is a Dynamic Improvement Reward (DIR) that evolves over time based on how well the conversation agent incorporates feedback.<n>Our experiments, conducted across five diverse datasets, demonstrate that WaltzRL significantly reduces both unsafe responses and overrefusals.
  arXiv  Detail & Related papers  (2025-10-09T14:03:05Z)
• Evaluating the Sensitivity of LLMs to Harmful Contents in Long Input [53.19281984086319]
  Large language models (LLMs) increasingly support applications that rely on extended context, from document processing to retrieval-augmented generation.<n>We evaluate LLMs' sensitivity to harmful content under extended context, varying type (explicit vs. implicit), position (beginning, middle, end), prevalence (0.01-0.50 of the prompt), and context length (600-6000 tokens).<n>We observe similar patterns: performance peaks at moderate harmful prevalence (0.25) but declines when content is very sparse or dominant; recall decreases with increasing context length; harmful sentences at the beginning are generally detected more reliably; and explicit content is more consistently recognized than implicit
  arXiv  Detail & Related papers  (2025-10-07T12:33:21Z)
• Attention Hijackers: Detect and Disentangle Attention Hijacking in LVLMs for Hallucination Mitigation [123.54980913741828]
  Large Vision-Language Models (LVLMs) remain vulnerable to hallucinations.<n>We propose a novel, training-free strategy namely Attention HIjackers Detection and Disentanglement (AID)<n>AID identifies Attention Hijackers by calculating instruction-driven visual salience.<n>Next, Attention Disentanglement mechanism is proposed to mask the visual attention of these identified Hijackers.<n>Re-Disentanglement recalculates the balance between instruction-driven and image-driven visual salience to avoid over-masking effects.
  arXiv  Detail & Related papers  (2025-03-11T09:35:55Z)
• Hashtag Re-Appropriation for Audience Control on Recommendation-Driven Social Media Xiaohongshu (rednote) [17.873872681980437]
  Women on Xiaohongshu (rednote) proactively re-appropriate hashtags by using them in posts unrelated to their literal meaning.<n>We analyzed the practice of hashtag re-appropriation based on 5,800 collected posts and interviewed 24 active users from diverse backgrounds.<n>This practice highlights how users can reclaim agency over content distribution on recommendation-driven platforms.
  arXiv  Detail & Related papers  (2025-01-30T08:55:32Z)
• YouTube Recommendations Reinforce Negative Emotions: Auditing Algorithmic Bias with Emotionally-Agentic Sock Puppets [0.624709220163167]
  This study examines YouTube's capacity to recognize and reinforce emotional preferences.<n>Our findings reveal that YouTube amplifies negative emotions, such as anger and grievance, by increasing their prevalence and prominence in recommendations.<n>Findings suggest the algorithm amplifies user biases, contributing to emotional filter bubbles and raising concerns about user well-being and societal impacts.
  arXiv  Detail & Related papers  (2025-01-25T03:04:53Z)
• Online posting effects: Unveiling the non-linear journeys of users in depression communities on Reddit [0.12564343689544843]
  We introduce a data-informed framework reconstructing online dynamics from 303k users interacting over two years.<n>Our analysis unveils online posting effects: a user can transition to another psychological state after online exposure to peers' emotional/semantic content.<n>Interpreted in light of psychological literature, our findings can provide evidence that the type and layout of online social interactions have an impact on users' "journeys" when posting about depression.
  arXiv  Detail & Related papers  (2023-11-29T14:45:11Z)
• Critical Behavioral Traits Foster Peer Engagement in Online Mental Health Communities [28.17719749654601]
  We introduce BeCOPE, a novel behavior encoded Peer counseling dataset comprising over 10,118 posts and 58,279 comments sourced from 21 mental health-specific subreddits. Our analysis indicates the prominence of self-criticism'' as the most prevalent form of criticism expressed by help-seekers, accounting for a significant 43% of interactions. We highlight the pivotal role of well-articulated problem descriptions, showing that superior readability effectively doubles the likelihood of receiving the sought-after support.
  arXiv  Detail & Related papers  (2023-09-04T14:00:12Z)
• Learning Language and Multimodal Privacy-Preserving Markers of Mood from Mobile Data [74.60507696087966]
  Mental health conditions remain underdiagnosed even in countries with common access to advanced medical care. One promising data source to help monitor human behavior is daily smartphone usage. We study behavioral markers of daily mood using a recent dataset of mobile behaviors from adolescent populations at high risk of suicidal behaviors.
  arXiv  Detail & Related papers  (2021-06-24T17:46:03Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.