Facebook Introduces Direct Topic Controls for User Recommendations

Meta announced that Facebook users can now shape their social feeds by writing plain text instructions directly to the platform's recommendation engine. According to Meta's product update, people can type out specific topics they want to see more or less of within a new management control tool or directly after rating individual posts. The system uses large language models to process these text requests and modify feed content during the active session.

For years, social media platforms inferred user preferences almost entirely by tracking passive habits like viewing time, likes, and comments. This indirect approach often trapped users in echo chambers or showed them unwanted content simply because they paused on a video. Allowing direct text controls gives everyday users a simple way to steer their feeds toward helpful hobbies or away from upsetting topics without creating a fresh account.

It remains to be seen how effectively Meta can prevent these direct inputs from being overbalanced by passive behavioral data over long periods. While immediate updates within a single session sound convenient, algorithms often revert to prioritizing highly engaging content over explicit user preferences once a session ends. How Meta balances short term user commands against its core business goal of maximizing screen time will determine whether this feature provides lasting control or just temporary relief.