Causal embeddings for recommendation

By Stephen Bonner, Flavian Vasile. ACM Conference on Recommender Systems, 2017

317 citations recorded by the source corpus.

Many current applications use recommendations in order to modify the natural user behavior, such as to increase the number of sales or the time spent on a website. This results in a gap between the final recommendation objective and the classical setup where recommendation candidates are evaluated by their coherence with past user behavior, by predicting either the missing entries in the user-item matrix, or the most likely next event. To bridge this gap, we optimize a recommendation policy for the task of increasing the desired outcome versus the organic user behavior. We show this is equivalent to learning to predict recommendation outcomes under a fully random recommendation policy. To this end, we propose a new domain adaptation algorithm that learns from logged data containing outcomes from a biased recommendation policy and predicts recommendation outcomes according to random exposure. We compare our method against state-of-the-art factorization methods, in addition to new approaches of causal recommendation and show significant improvements.

Subfields: Artificial Intelligence, Information Systems, Management Science and Operations Research

View Paper: https://www.semanticscholar.org/paper/c8211bc846a42de016fc6b31e17bda2d33016ef2

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Xiangnan He
2. Xiang Wang
3. Yong Li
4. Fuli Feng
5. Chen Gao
6. Meng Wang
7. Depeng Jin
8. Tat-seng Chua

In Information Systems

1. Xiangnan He
2. Xiang Wang
3. Fuli Feng
4. Yong Li
5. Chen Gao
6. Meng Wang
7. Depeng Jin
8. Tat-seng Chua

In Management Science and Operations Research

1. Xiangnan He
2. Xiang Wang
3. Meng Wang
4. Fuli Feng
5. Jing Gao
6. Jiawei Chen
7. Aidong Zhang
8. Zhixuan Chu

Citations in the corpus

• A Survey on Causal Inference (2020, methodology): “rrect selection bias. In this case, the dataset contains a large set of logged feedback records under control policy and a small set of records under the randomized recommendation. CausalEmbed(CausE) [19] is a representative method in this direction, which proposed a new matrix factorization algorit”
• Unbiased Recommender Learning from Missing-Not-At-Random Implicit Feedback (2019, background): “Recommender systems are widely used in industries such as personalized advertising and video streaming [3, 9, 10].”
• Causal Intervention for Leveraging Popularity Bias in Recommendation (2021, methodology): “Effective usage of small uniform data remains an open problem in recommendation research [8, 27].”
• Bias and Debias in Recommender System: A Survey and Future Directions (2020, methodology): “[142] employ two recommendation models for the biased data and uniform data, and further use a regularization term to transfer the knowledge between the models; Liu et al.”
• Disentangling User Interest and Conformity for Recommendation with Causal Embedding (2021, methodology): “Data Preprocessing In order tomeasure the performance of causal learning under non-IID circumstances, intervened test sets are needed, and thus all datasets are transformed following the standard protocol introduced in related literatures [9, 30].”