Learning to Alleviate Familiarity Bias in Video Recommendation

2026-02-08

AI Summary

This work addresses the familiarity bias prevalent in video recommendation systems, where user behavior biases lead to overexposure of known content at the expense of novelty and diversity. The authors propose LAFB, the first lightweight, model-agnostic post-ranking debiasing framework tailored for familiarity bias. Without modifying the core recommendation model, LAFB models user–item familiarity by fusing discrete and continuous interaction features and introduces personalized debiasing factors to dynamically adjust item scores. Integrated into a unified online serving architecture, the method was deployed in YouTube’s real-world system, significantly increasing the share of watch time on novel content and exposure for emerging creators while maintaining overall watch time and short-term user satisfaction. This approach effectively balances novelty, creator fairness, and user experience.

Technology Category

• Machine Learning: Learning Preferences or Rankings
• Computer Vision: Bias, Fairness & Privacy
• Data Mining & Knowledge Management: Recommender Systems

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Modern video recommendation systems aim to optimize user engagement and platform objectives, yet often face structural exposure imbalances caused by behavioral biases. In this work, we focus on the post-ranking stage and present LAFB (Learning to Alleviate Familiarity Bias), a lightweight and model-agnostic framework designed to mitigate familiarity bias in recommendation outputs. LAFB models user-content familiarity using discrete and continuous interaction features, and estimates personalized debiasing factors to adjust user rating prediction scores, thereby reducing the dominance of familiar content in the final ranking. We conduct large-scale offline evaluations and online A/B testing in a real-world recommendation system, under a unified serving stack that also compares LAFB with deployable popularity-oriented remedies. Results show that LAFB increases novel watch-time share and improves exposure for emerging creators and overall content diversity, while maintaining stable overall watch time and short-term satisfaction. LAFB has already been launched in the post-ranking stage of YouTube's recommendation system, demonstrating its effectiveness in real-world applications.

Problem

• familiarity bias
• video recommendation
• exposure imbalance
• content diversity
• recommendation bias

Innovation

• familiarity bias
• video recommendation
• debiasing
• post-ranking
• content diversity