用多模态大模型审计TikTok青少年内容风险，发现搜索比刷屏更易暴露于有害内容。
Auditing Exposure to Harmful Content on TikTok using Multimodal Language Models: A Cross-National, Age-Stratified Study
- 用多模态大模型自动标注视频，成本仅为人工一半
- 关键词搜索使有害内容占比达35%-56%，是刷屏的1.5-7.5倍
- 适合关注跨国家、跨年龄平台安全的政策研究者
在线视频平台可能使年轻用户暴露于有害内容，但独立审计困难，因视频标注成本高且不同语言间审核标准不一。本研究在法国、意大利和瑞典使用代表四个年龄层（13、16、19、40岁）的傀儡账号，通过被动推荐页浏览和主动搜索有害关键词再浏览，共收集36,971条视频。为实现规模化标注，我们在300条视频样本上验证四种多模态大模型，结果显示Gemini 2.5 Flash结合八帧图像与文本表现最佳（综合卡帕系数=0.42），单次调用成本仅为上传原视频的一半，最终在10%样本上总API支出约50美元。关键词搜索下有害内容占比达35%-56%，在十二组国家-年龄组合中十组较被动浏览提升1.5-7.5倍；该峰值为暂时性，削弱了法、瑞两国的年龄差异。被动浏览时意大利危害率最高，19岁群体达48.6%。总体而言，多模态大模型审计可实现跨国家青少年安全评估，而平台安全过滤器（拒绝率1.1%）对最明显有害内容存在低估。
原文摘要 · Abstract (English)
Online video platforms can expose young users to harmful content, but independent audits remain difficult because video annotation is costly and moderation judgments vary across languages. We audit TikTok in France, Italy, and Sweden with sockpuppet accounts representing four age personas (13, 16, 19, 40), collecting 36,971 videos from passive For-You-page scrolling and active sessions that scroll, search for harm keywords, and scroll again. To scale annotation, we validate four multimodal LLMs against native-speaker labels on a 300-video reference set. Gemini 2.5 Flash with eight sampled frames plus text performs best (aggregate kappa = 0.42), at half the per-call cost of native-video upload, and we apply it to a 10% sample for approximately \$50 in total API spend across both modalities. Keyword search returns 35-56% harmful content, a 1.5-7.5x increase over the scrolling baseline in ten of twelve country-age combinations; the spike is temporary and flattens the age differences observed in France and Sweden. Under passive scrolling, Italy has the highest harm rate at every age, with Italian age-19 reaching 48.6%. Overall, MLLM-based auditing offers a scalable approach for cross-national youth-safety audits, while provider safety filters (1.1% refusal rate) under-count the most explicit harms.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。