Taming Recommendation Bias with Causal Intervention on Evolving Personal Popularity

2025/05/20 by Shiyin Tan, Tan, Shiyin, Dongyuan Li +9
Computer Science · Physics and Astronomy · #Complex Network Analysis Techniques #Expert finding and Q&A systems #FOS: Computer and information sciences #Information Retrieval (cs.IR) #Machine Learning (cs.LG) #Recommender Systems and Techniques

paper · pdf · doi:10.48550/arxiv.2505.14310

openalex publication_date 2025/05/20 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Popularity bias occurs when popular items are recommended far more frequently than they should be, negatively impacting both user experience and recommendation accuracy. Existing debiasing methods mitigate popularity bias often uniformly across all users and only partially consider the time evolution of users or items. However, users have different levels of preference for item popularity, and this preference is evolving over time. To address these issues, we propose a novel method called CausalEPP (Causal Intervention on Evolving Personal Popularity) for taming recommendation bias, which accounts for the evolving personal popularity of users. Specifically, we first introduce a metric called Evolving Personal Popularity to quantify each user's preference for popular items. Then, we design a causal graph that integrates evolving personal popularity into the conformity effect, and apply deconfounded training to mitigate the popularity bias of the causal graph. During inference, we consider the evolution consistency between users and items to achieve a better recommendation. Empirical studies demonstrate that CausalEPP outperforms baseline methods in reducing popularity bias while improving recommendation accuracy.

Citations

• Popularity-Aware Alignment and Contrast for Mitigating Popularity Bias
• Fair Recommendations with Limited Sensitive Attributes: A Distributionally Robust Optimization Approach
• Towards Efficient and Effective Unlearning of Large Language Models for Recommendation
• Generative News Recommendation
• PromptMM: Multi-Modal Knowledge Distillation for Recommendation with Prompt-Tuning
• Challenging Low Homophily in Social Recommendation
• Robust Collaborative Filtering to Popularity Distribution Shift
• A Survey on Large Language Models for Recommendation
• How Graph Convolutions Amplify Popularity Bias for Recommendation?
• Causal Inference in Recommender Systems: A Survey and Future Directions
• Causal Inference in Recommender Systems: A Survey and Future Directions
• Mitigating Hidden Confounding Effects for Causal Recommendation
• Recommendation as Language Processing (RLP): A Unified Pretrain, Personalized Prompt & Predict Paradigm (P5)
• Causal Disentanglement for Semantics-Aware Intent Learning in Recommendation
• Popularity Bias Is Not Always Evil: Disentangling Benign and Harmful Bias for Recommendation
• Bias and Debias in Recommender System: A Survey and Future Directions
• Bias and Debias in Recommender System: A Survey and Future Directions
• The Connection Between Popularity Bias, Calibration, and Fairness in\n Recommendation
• Popularity Bias in Recommendation: A Multi-stakeholder Perspective
• LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation
• LightGCN
• Unbiased Recommender Learning from Missing-Not-At-Random Implicit Feedback
• BERT4Rec: Sequential Recommendation with Bidirectional Encoder Representations from Transformer
• The Deconfounded Recommender: A Causal Inference Approach to Recommendation
• Billion-scale Commodity Embedding for E-commerce Recommendation in Alibaba
• Unbiased Learning-to-Rank with Biased Feedback
• Recommendations as Treatments: Debiasing Learning and Evaluation
• Counterfactual Reasoning and Learning Systems
• Cumulated gain-based evaluation of IR techniques

Related

• Evolution of Popularity Bias: Empirical Study and Debiasing
• Popularity Bias Is Not Always Evil: Disentangling Benign and Harmful Bias for Recommendation
• Aligning Recommendations with User Popularity Preferences
• Popularity Bias in Recommendation: A Multi-stakeholder Perspective
• Reducing Popularity Bias in Recommendation Over Time