Recommendation Debiasing

This website collects recent works and datasets on recommendation debiasing and their codes. We hope this website could help you do search on this topic.

Contents

• 1. Survey Papers

• 2. Datasets

• 3. Debiasing Strategies

  • 3.1 Multiply Biases
  • 3.2 Selection Bias
  • 3.3 Conformity Bias
  • 3.4 Exposure Bias
  • 3.5 Position Bias
  • 3.6 Popularity Bias
  • 3.7 Inductive Bias
  • 3.8 Unfairness
  • 3.9 Loop Effect
  • 3.10 Other Bias

• Tips

1. Survey Papers

1. A Survey on the Fairness of Recommender Systems. TOIS 2023. [pdf]

2. Bias and Debias in Recommender System: A Survey and Future Directions. TOIS 2023. [pdf]

3. Bias Issues and Solutions in Recommender System. WWW 2021,Recsys 2021. [pdf]

4. A survey on bias and fairness in machine learning. Arxiv 2019. [pdf]

2. Datasets

We collect some datasets which include unbiased data and are often used in the research of recommendation debiasing.

1. Yahoo!R3: Collaborative Prediction and Ranking with Non-Random Missing Data. Recsys 2009. [pdf][data]

2. Coat: Recommendations as Treatments: Debiasing Learning and Evaluation. ICML 2016. [pdf][data]

3. KuaiRec: A Fully-observed Dataset for Recommender Systems. CIKM 2022. [pdf][data]

4. KuaiRand: An Unbiased Sequential Recommendation Dataset with Randomly Exposed Videos. CIKM 2022.[pdf][data]

3. Debiasing Strategies

3.1 Multiply Biases

1. Bounding System-Induced Biases in Recommender Systems with a Randomized Dataset. TOIS 2023.[pdf] [code]

2. Balancing Unobserved Confounding with a Few Unbiased Ratings in Debiased Recommendations. WWW 2023.[pdf]

3. Transfer Learning in Collaborative Recommendation for Bias Reduction. Recsys 2021.[pdf] [code]

4. AutoDebias: Learning to Debias for Recommendation. SIGIR 2021.[pdf] [code]

5. A General Knowledge Distillation Framework for Counterfactual Recommendation via Uniform Data. SIGIR 2020.[pdf] [code]

6. Causal Embeddings for Recommendation. Recsys 2018.[pdf] [code]

3.2 Selection Bias

1. Reconsidering Learning Objectives in Unbiased Recommendation A Distribution Shift Perspective. KDD 2023.[pdf]

2. Propensity Matters Measuring and Enhancing Balancing for Recommendation. ICML 2023.[pdf]

3. A Generalized Propensity Learning Framework for Unbiased Post-Click Conversion Rate Estimation. CIKM 2023.[pdf] [code]

4. CDR: Conservative Doubly Robust Learning for Debiased Recommendation. CIKM 2023.[pdf] [code]

5. UKD: Debiasing Conversion Rate Estimation via Uncertainty-regularized Knowledge Distillation. WWW 2022.[pdf]

6. Practical Counterfactual Policy Learning for Top-𝐾 Recommendations. KDD 2022.[pdf]

7. Debiasing Neighbor Aggregation for Graph Neural Network in Recommender Systems. CIKM 2022.[pdf]

8. Representation Matters When Learning From Biased Feedback in Recommendation. CIKM 2022.[pdf]

9. Hard Negatives or False Negatives: Correcting Pooling Bias in Training Neural Ranking Models. CIKM 2022.[pdf]

10. Be Causal: De-biasing Social Network Confounding in Recommendation. TKDD 2022.[pdf]

11. Debiased recommendation with neural stratification. AI OPEN 2022.[pdf]

12. ESCM2: Entire Space Counterfactual Multi-Task Model for Post-Click Conversion Rate Estimation. SIGIR 2022.[pdf]

13. Generalized Doubly Robust Learning Framework for Debiasing Post-Click Conversion Rate Prediction. KDD 2022.[pdf]

14. Combating Selection Biases in Recommender Systems with a Few Unbiased Ratings. WSDM 2021.[pdf]

15. Doubly Robust Estimator for Ranking Metrics with Post‐Click Conversions. RecSys 2020.[pdf] [code]

16. Asymmetric tri-training for debiasing missing-not-at-random explicit feedback. SIGIR 2020.[pdf]

17. Recommendations as treatments: Debiasing learning and evaluation. ICML 2016.[pdf] [code]

18. Doubly robust joint learning for recommendation on data missing not at random. ICML 2019.[pdf]

19. The deconfounded recommender: A causal inference approach to recommendation. arXiv 2018.[pdf]

20. Social recommendation with missing not at random data. ICDM 2018.[pdf]

21. Recommendations as treatments: Debiasing learning and evaluation. [pdf]

22. Boosting Response Aware Model-Based Collaborative Filtering. TKDE 2015.[pdf]

23. Probabilistic matrix factorization with non-random missing data. PMLR 2014.[pdf] [code]

24. Bayesian Binomial Mixture Model for Collaborative Prediction With Non-Random Missing Data. RecSys 2014.[pdf]

25. Evaluation of recommendations: rating-prediction and ranking. RecSys 2013.[pdf]

26. Training and testing of recommender systems on data missing not at random. KDD 2010.[pdf]

27. Collaborative prediction and ranking with non-random missing data. RecSys 2009.[pdf]

28. Collaborative filtering and the missing at random assumption. UAI 2007.[pdf] [code]

3.3 Conformity Bias

1. Popularity Bias Is Not Always Evil: Disentangling Benign and Harmful Bias for Recommendation. TKDE 2022.[pdf]

2. Disentangling user interest and Conformity for recommendation with causal embedding. WWW 2021.[pdf] [code]

3. When Sheep Shop: Measuring Herding Effects in Product Ratings with Natural Experiments. WWW 2018.[pdf] [code]

4. Learning personalized preference of strong and weak ties for social recommendation. WWW 2017.[pdf]

5. Are you influenced by others when rating?: Improve rating prediction by conformity modeling. RecSys 2016.[pdf]

6. Xgboost: A scalable tree boosting system. KDD 2016.[pdf] [code]

7. A probabilistic model for using social networks in personalized item recommendation. RecSys 2015.[pdf] [code]

8. Why amazon’s ratings might mislead you: The story of herding effects. Big data 2014.[pdf]

9. A methodology for learning, analyzing, and mitigating social influence bias in recommender systems. RecSys 2014.[[pdf](https://amplab.cs.berkeley.edu/wp-content/uploads/20