Causal Disentanglement with Network Information for Debiased Recommendations

Recommender 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 the different biases act as confounding factors. Therefore, to infer debiased preferences and to capture the causal relationship between exposure and the observed ratings, it is essential to account for any hidden confounders. To this end, we propose a novel causal disentanglement framework that decomposes latent representations into three independent factors, responsible for (a) modeling the exposure of an item, (b) predicting ratings, and (c) controlling for hidden confounders. Experiments on real-world datasets validate the effectiveness of the proposed Causal Disentanglement for DeBiased Recommendations (D2Rec) model in debiasing recommendations.