Paras Sheth
Huan Liu
Abstract
Online user engagement is highly influenced by various machine learning models, such as recommender systems. These systems recommend new items to the user based on the user’s historical interactions. Implicit recommender systems reflect a binary setting showing whether a user interacted (e.g., clicked on) with an item or not. However, the observed clicks may be due to various causes such as user’s interest, item’s popularity, and social influence factors. Traditional recommender systems consider these causes under a unified representation, which may lead to the emergence and amplification of various biases in recommendations. However, recent work indicates that by disentangling the unified representations, one can mitigate bias (e.g., popularity bias) in recommender systems and help improve recommendation performance. Yet, prior work in causal disentanglement in recommendations does not consider a crucial factor, that is, social influence. Social theories such as homophily and social influence provide evidence that a user’s decision can be highly influenced by the user’s social relations. Thus, accounting for the social relations while disentangling leads to less biased recommendations. To this end, we identify three separate causes behind an effect (e.g., clicks): (a) user’s interest, (b) item’s popularity, and (c) user’s social influence. Our approach seeks to causally disentangle the user and item latent features to mitigate popularity bias in implicit feedback–based social recommender systems. To achieve this goal, we draw from causal inference theories and social network theories and propose a causality-aware disentanglement method that leverages both the user–item interaction network and auxiliary social network information. Experiments on real-world datasets against various state-of-the-art baselines validate the effectiveness of the proposed model for mitigating popularity bias and generating de-biased recommendations.
- [1] . 2017. Controlling popularity bias in learning-to-rank recommendation. In RecSys.Google Scholar
- [2] . 2018. Causal embeddings for recommendation. Proceedings of the 12th ACM Conference on Recommender Systems, Association for Computing Machinery, New York, NY, 104–112.Google Scholar
- [3] . 2019. An analysis of the softmax cross entropy loss for learning-to-rank with binary relevance. In SIGIR.Google Scholar
- [4] . 2014. Exploring social network effects on popularity biases in recommender systems. In RSWeb@ RecSys.Google Scholar
- [5] . 2017. A probabilistic reformulation of memory-based collaborative filtering: Implications on popularity biases. In SIGIR.Google Scholar
- [6] . 2018. How algorithmic confounding in recommendation systems increases homogeneity and decreases utility. In RecSys.Google Scholar
- [7] . 2020. Bias and debias in recommender system: A survey and future directions. arXiv:2010.03240 (2020).Google Scholar
- [8] . 2020. Semi-disentangled representation learning in recommendation system. arXiv:2010.13282 (2020).Google Scholar
- [9] . 2021. Disentangled item representation for recommender systems. ACM Transactions on Intelligent Systems and Technology 12, 2 (2021).Google ScholarDigital Library
- [10] . 2012. A kernel two-sample test. The Journal of Machine Learning Research 13, 1 (2012), 723–773.Google Scholar
- [11] . 2016. node2vec: Scalable feature learning for networks. In SIGKDD.Google Scholar
- [12] . 2019. Learning disentangled representations for counterfactual regression. In ICLR.Google Scholar
- [13] . 2020. LightGCM: Simplifying and powering graph convolution network for recommendation. In SIGIR.Google Scholar
- [14] . 2014. Deep modeling of group preferences for group-based recommendation. In AAAI, Vol. 28.Google Scholar
- [15] . 2020. Graph neural news recommendation with unsupervised preference disentanglement. In ACL-IJCNLP.Google Scholar
- [16] . 2010. A matrix factorization technique with trust propagation for recommendation in social networks. In RecSys.Google Scholar
- [17] . 2017. Unbiased learning-to-rank with biased feedback. In Proceedings of the 10th ACM International Conference on Web Search and Data Mining. 781–789.Google ScholarDigital Library
- [18] . 2016. Semi-supervised classification with graph convolutional networks. arXiv:1609.02907 (2016).Google Scholar
- [19] . 1938. A generalization of Fisher’s Z test. Biometrika 30, 1/2 (1938), 180–187.Google ScholarCross Ref
- [20] . 2021. Be causal: De-biasing social network confounding in recommendation. arXiv:2105.07775 (2021).Google Scholar
- [21] . 2016. Causal inference for recommendation. In Causation: Foundation to Application, Workshop at UAI. AUAI.Google Scholar
- [22] . 2020. Explainable recommender systems via resolving learning representations. In CIKM.Google Scholar
- [23] . 2009. Collaborative prediction and ranking with non-random missing data. In RecSys.Google Scholar
- [24] . 1993. Network studies of social influence. Sociological Methods & Research 22, 1 (1993), 127–151.Google ScholarCross Ref
- [25] . 2001. Birds of a feather: Homophily in social networks. Annual Review of Sociology 27, 1 (2001), 415–444.Google ScholarCross Ref
- [26] . 2008. Probabilistic matrix factorization. In Advances in Neural Information Processing Systems. 1257–1264.Google ScholarDigital Library
- [27] . 2020. Untangle: Critiquing disentangled recommendations. (2020).Google Scholar
- [28] . 2011. Transportability of causal and statistical relations: A formal approach. Proceedings of the AAAI Conference on Artificial Intelligence 25, 1 (2011), 247–254.Google Scholar
- [29] . 2012. BPR: Bayesian personalized ranking from implicit feedback. arXiv:1205.2618 (2012).Google Scholar
- [30] . 2014. Diffusion of innovations. Routledge.Google Scholar
- [31] . 2002. Overt bias in observational studies. In Observational Studies.Google ScholarCross Ref
- [32] . 1983. The central role of the propensity score in observational studies for causal effects. Biometrika 70, 1 (1983), 41–55.Google ScholarCross Ref
- [33] . 1961. Anxiety, fear, and social isolation. The Journal of Abnormal and Social Psychology 62, 2 (1961), 356.Google ScholarCross Ref
- [34] . 2021. Disentangling multi-facet social relations for recommendation. IEEE Transactions on Computational Social Systems (2021).Google Scholar
- [35] . 2022. Causal disentanglement with network information for debiased recommendations. In International Conference on Similarity Search and Applications. Springer, 265–273.Google ScholarDigital Library
- [36] . 2017. Social media marketing: The effect of information sharing, entertainment, emotional connection and peer pressure on the attitude and purchase intentions. GSTF Journal on Business Review (GBR) 5, 1 (2017).Google Scholar
- [37] . 2001. Comparing recommendations made by online systems and friends. DELOS 106 (2001).Google Scholar
- [38] . 2020. Disentangled graph collaborative filtering. In SIGIR.Google Scholar
- [39] . 2021. Model-agnostic counterfactual reasoning for eliminating popularity bias in recommender system. In SIGKDD.Google Scholar
- [40] . 2017. Returning is believing: Optimizing long-term user engagement in recommender systems. In CIKM.Google Scholar
- [41] . 2021. Investigating and counteracting popularity bias in group recommendations. Information Processing & Management 58, 5 (2021), 102608.Google ScholarDigital Library
- [42] . 2021. Causal intervention for leveraging popularity bias in recommendation. arXiv:2105.06067 (2021).Google Scholar
- [43] . 2021. Disentangling user interest and conformity for recommendation with causal embedding. In WWW.Google Scholar
Index Terms
- Causal Disentanglement for Implicit Recommendations with Network Information
Recommendations
Dual disentanglement of user–item interaction for recommendation with causal embedding
AbstractTo achieve personalized recommendations, the recommender system selects the items that users may like by learning the collected user–item interaction data. However, the acquisition and use of data usually form a feedback loop, which leads to ...
Highlights- We propose a novel dual disentanglement for recommendation with causal embedding.
- We consider the popularity bias from a double-end perspective in the recommendation.
- We consider the item attributes in the item’s popularity.
- We ...
Causal Disentanglement with Network Information for Debiased Recommendations
Similarity Search and ApplicationsAbstractRecommender systems suffer from biases that may misguide the system when learning user preferences. Under the causal lens, the user’s exposure to items can be seen as the treatment assignment, the ratings of the items are the observed outcome, and ...
Causal embedding of user interest and conformity for long-tail session-based recommendations
AbstractSession-based recommendation is misleading by popularity bias and always favors short-head items with more popularity. This paper studies a new causal-based framework CauTailReS to increase the diversity of session recommendations. We first ...
Comments