Separating and Learning Latent Confounders to Enhancing User Preferences Modeling

TL;DR

This paper introduces SLFR, a novel framework that disentangles user preferences from unmeasured confounders, including the influence of previous recommender systems, to improve recommendation accuracy.

Contribution

The paper proposes a new method to identify and separate unmeasured confounders, including the effect of prior recommenders, enhancing the modeling of true user preferences.

Findings

SLFR outperforms existing models on five real-world datasets.

Disentangling confounders improves recommendation accuracy.

Incorporating prior recommender effects as confounders enhances model robustness.

Abstract

Recommender models aim to capture user preferences from historical feedback and then predict user-specific feedback on candidate items. However, the presence of various unmeasured confounders causes deviations between the user preferences in the historical feedback and the true preferences, resulting in models not meeting their expected performance. Existing debias models either (1) specific to solving one particular bias or (2) directly obtain auxiliary information from user historical feedback, which cannot identify whether the learned preferences are true user preferences or mixed with unmeasured confounders. Moreover, we find that the former recommender system is not only a successor to unmeasured confounders but also acts as an unmeasured confounder affecting user preference modeling, which has always been neglected in previous studies. To this end, we incorporate the effect of the former recommender system and treat it as a proxy for all unmeasured confounders. We propose a novel framework, Separating and Learning Latent Confounders For Recommendation (SLFR), which obtains the representation of unmeasured confounders to identify the counterfactual feedback by disentangling user preferences and unmeasured confounders, then guides the target model to capture the true preferences of users. Extensive experiments in five real-world datasets validate the advantages of our method.