Meta 在 Facebook 上线 Your Algorithm,可用自然语言提示词调整 Feed 和 Reels
Facebook's Your Algorithm lets users tune Feed and Reels
Meta 于 10 月 13 日宣布在 Facebook 推出 Your Algorithm,用户可通过自然语言提示词指定想多看或少看的话题,并对单条帖子选择 Interested 或 Not interested。该功能入口在帖子三点菜单中,Meta 称排序系统会在同一次会话中结合提示词与帖子级反馈调整推荐,此前 Instagram 和 Threads 已有类似工具。MediaNama 实测印度账号发现该功能及类似功能目前均未在印度上线,并已就可用地区、提示词数据存储期限、是否用于广告画像或模型训练等问题向 Meta 提问。
推荐理由
AI 评分 62
Meta 把自然语言提示词接入 Facebook 推荐排序,读者可了解该功能的入口、可用范围与数据留存疑问。
主题
正文
当前显示原文,中文全文译文尚未就绪。
Meta announced on October 13 that it is rolling out Your Algorithm on Facebook. This AI-powered tool lets Facebook users customise recommendations in Feed and Reels using natural language prompts. The tool allows users to specify the topics they want to see more or less of, while also providing feedback on individual posts.
Meta says its ranking systems incorporate users’ prompts and post-level feedback during the same session to adjust recommendations. The announcement follows similar rollouts on Instagram and Threads earlier this year, where Meta introduced tools to give users more control over their recommended content.
How Your Algorithm works: Users can access Your Algorithm through the three-dot menu on Facebook posts, where they can add and manage topics they want to see in Feed and Reels.
Additionally, users can select ‘Interested’ or ‘Not interested’ on individual posts. Facebook may then ask follow-up questions to understand their preferences in greater detail. The system combines these natural language inputs with existing signals, such as likes, comments, and viewing activity, to rank content.
Feature not available in India: MediaNama checked an India-based account and found that neither Your Algorithm nor similar features are currently available on Facebook, Instagram or Threads in India. However, when MediaNama tested the Dear Algo feature on Threads in February using a UK-based account, it initially appeared to work as intended.
We published a post saying, “dear algo give me more luxury content”, after which the feed showed more UK-based English posts about luxury products. However, the feed subsequently began displaying several posts in Tagalog, a language spoken in the Philippines, despite the original prompt seeking luxury-related content.
LLMs in Meta’s recommender systems: Meta has been experimenting with LLMs in its recommendation systems since the start of the year. During its Q4 FY25 earnings call, CEO Mark Zuckerberg said the company planned to integrate LLMs into recommendations across Facebook, Instagram, Threads and advertising.
In its Q1 FY26 earnings call in April, CFO Susan Li said Meta had doubled the length of Instagram interaction sequences used for training and introduced more detailed descriptions of user activity. Meta said these changes helped increase the time users spent watching Instagram Reels by 10%.
By July, the company said it was processing every public Instagram Feed and Reels post through an LLM to analyse topics and tone, informing ranking, recommendations and content moderation. It also began using Muse models for video classification and summarisation.
Questions for Meta: MediaNama has sent the following questions to the company. This article will be updated when we receive a response:
Meta announced the rollout of Your algorithm on Facebook on October 13, 2026. Which countries and regions currently have access to the feature, and what is the timeline for its rollout in India across Facebook, Instagram, and Threads?; How does Meta store and process the natural-language prompts users submit through Your Algorithm and the ‘Interested’ and ‘Not interested’ feedback options? For how long does Meta retain this data, and can users view or delete their prompt history?; Does Meta use these prompts and feedback to build or update individual user profiles for targeted advertising, ad personalisation, or other commercial purposes? If so, how are these inputs used, and can users opt out?; Does Meta use users’ prompts and feedback to train or fine-tune its large language models or other AI systems?; Are users’ natural-language prompts processed by Meta’s AI systems subject to any additional privacy or data protection safeguards?
Also read:
Meta tells Bombay HC it disabled Blink Digital’s ad accounts over alleged scam and sexual services ads; Delhi Police spotted with Meta AI glasses during Oct 10 protest, compounding privacy concerns
For You
Related Stories
Australia to propose a law to turn off algorithmic recommendations on social media
X opens more of its algorithm, giving users a look at possible ‘shadowbans’
Here is what Pope Leo XIV said about artificial intelligence
Why banning children from Social Media avoids the real problem
Video: Why Are Gig Workers Asking For Algorithmic Transparency?
From the Archives
Post navigation
Centre pushes child rights body to fast-track AI tool to detect child sexual abuse material
NCPCR is developing an AI tool to detect child sexual abuse material online and preparing social media guidelines for technology companies.
Nandan Nilekani suggests last-minute AI-generated exam papers to prevent leaks
Nandan Nilekani has proposed generating examination papers at the last minute using AI to reduce incentives for leaks. But can technology address the deeper failures in India’s examination system?