Meta now gives a closer look at upgrades to Facebook Reels recommendations. The company listens to users directly and uses their feedback to improve what they see. By running quick surveys, Meta learns what you enjoy, not just what you interact with. This approach goes beyond old measurements like likes or watch time.
Takeaways 🚀
✅Meta improves Reels recommendations.
✅User feedback refines content delivery.
✅Surveys enhance personalization and engagement.
✅UTIS model boosts video relevance.
✅Advanced AI ensures tailored experiences.
The Challenge of Personal Recommendations
Giving you videos you actually want to watch isn’t easy, especially on a huge platform. Old systems guess your preferences based on interactions, but these don’t always reveal your true interests. Meta changed its focus to “interest matching.” Now, it considers details such as sound, production style, mood, and motivation. This helps deliver content that feels more personal and draws you in.
How the UTIS Model Works
Meta built the User True Interest Survey (UTIS) model to meet this challenge. The app now asks some users to rate how much a video fits their interests, using a 1-to-5 scale. Meta collects these survey responses in real time. Then, their alignment model uses this data to work alongside the core recommendation engine.
The UTIS model gives each video a score. This score predicts how much satisfaction a user will get from that content. The ranking system uses the score to show better-matching videos higher in your feed, while pushing less relevant ones lower. This leads to a more engaging and tailored experience for every user.
Results and Benefits
Meta’s new system achieves real results. The UTIS model increased recommendation accuracy from 59.5% to 71.5%. Precision also improved, climbing from 48.3% to 63.2%.
When Meta tested the model on over 10 million users, high survey ratings climbed by 5.4%. Overall user engagement also went up by 5.2%. At the same time, low survey ratings and integrity issues dropped.
Thanks to these changes, users now see more high-quality, unique videos instead of only broad, popular content.
Next Steps
Meta is not stopping here. The team wants to address gaps like users with little data, survey bias, and delivering even more variety in recommendations. To do this, they are also testing advanced AI techniques, including large language models, to make your feed smarter and more relevant.
When Meta matches feedback with machine learning, users gain a better experience. This not only makes content more personal but also keeps you coming back for more.
