Counterfactual Inference for Eliminating Sentiment Bias in Recommender Systems

TL;DR

This paper introduces a novel counterfactual inference approach to mitigate sentiment bias in review-based recommender systems, improving fairness and recommendation accuracy for critical users and niche items.

Contribution

It is the first to apply counterfactual inference for sentiment bias mitigation in recommender systems, modeling causal effects to enhance fairness and accuracy.

Findings

Effective reduction of sentiment bias in recommendations

Maintains comparable rating prediction performance

Improves fairness for critical users and niche items

Abstract

Recommender Systems (RSs) aim to provide personalized recommendations for users. A newly discovered bias, known as sentiment bias, uncovers a common phenomenon within Review-based RSs (RRSs): the recommendation accuracy of users or items with negative reviews deteriorates compared with users or items with positive reviews. Critical users and niche items are disadvantaged by such unfair recommendations. We study this problem from the perspective of counterfactual inference with two stages. At the model training stage, we build a causal graph and model how sentiment influences the final rating score. During the inference stage, we decouple the direct and indirect effects to mitigate the impact of sentiment bias and remove the indirect effect using counterfactual inference. We have conducted extensive experiments, and the results validate that our model can achieve comparable performance on rating prediction for better recommendations and effective mitigation of sentiment bias. To the best of our knowledge, this is the first work to employ counterfactual inference on sentiment bias mitigation in RSs.