Your Algorithm, Your Words: New Ways to Shape What You See on Facebook

Facebook is introducing new AI-driven features that let users customize their Feed and Reels by describing their interests in their own words. These tools allow real-time adjustments to recommendations based on direct user input, replacing passive reliance on inferred preferences. The updates include a dedicated space called 'Your algorithm' within the three-dot menu, where users can add or manage topics of interest. This feature is designed to make the recommendation system more transparent and user-controlled.

Meta is also expanding feedback options by introducing follow-up questions after users select 'Interested' or 'Not interested' on posts. This allows the system to refine recommendations based on specific likes or dislikes expressed in natural language. The goal is to create a more conversational and responsive interaction with the algorithm, enabling users to shape their experience dynamically.

The new system leverages large language models (LLMs) to interpret user intent and context behind natural language feedback, marking a shift in how recommendations are generated. Unlike traditional ranking systems that infer preferences from engagement signals, this approach allows users to explicitly state what they want to see more or less of. The integration of LLMs aims to make recommendations more nuanced and immediately responsive to user input.

Underlying these changes is an evolution in Facebook’s ranking systems, which now combine freeform text input with existing engagement signals to refine recommendations. The system adapts within the same session, ensuring that recommendations stay relevant to the user’s current interests. This marks a significant step toward making the algorithm more interactive and user-driven, rather than relying solely on passive data analysis.