Recently, INFORMS Journal on Computing, a journal of the Institute for Operations Research and the Management Sciences (INFORMS), has published online the latest research conducted by Associate Professor and Doctoral Supervisor Geng Ruibin from the School of Management, Northwestern Polytechnical University, and his collaborators. The paper is titled "Eliminating Social Popularity Bias in Recommendation: Causal Inference-Based Social Graph Neural Networks", with Associate Professor Geng Ruibin serving as the corresponding author. This study addresses the issue of "social popularity bias" in social recommendation systems. Traditional recommendation algorithms tend to favour popular content. In social network environments, such bias is further amplified by "social influence", concentrating recommendation results on content followed by opinion leaders or high-impact users and impairing information diversity and fairness.

To tackle this critical problem, the research team has made innovative achievements as follows: it constructs a Causal Graph model to systematically reveal the interference mechanism of "item popularity" and "social influence" as confounding factors in user preference learning; proposes a social graph neural network model (CISGNN) based on Backdoor Adjustment and Counterfactual Reasoning to fundamentally eliminate social popularity bias; validates the model’s effectiveness using four real-world social datasets (Ciao, Epinions, Yelp-Philadelphia, Yelp-Tucson). The results demonstrate that CISGNN significantly outperforms existing state-of-the-art models in recommendation accuracy metrics (HR, NDCG) and diversity metrics (Novelty, NicheRate). The framework proposed in this research not only promotes the integration of causal inference and recommendation systems at the algorithmic level, but also provides important managerial implications for platform algorithm governance, fair recommendation and incentive mechanism design for long-tail content.

INFORMS Journal on Computing is one of the world’s top journals in management science and operations research, and is included in the UTD-24 list, a collection of 24 most influential core business journals selected by the University of Texas at Dallas. It enjoys an outstanding academic reputation at the intersection of artificial intelligence and managerial decision-making.

About the speaker:Geng Ruibin is an Associate Professor and Doctoral Supervisor at the School of Management, Northwestern Polytechnical University. His research interests include digital economy and digital governance, data sharing and information security, and AI-driven decision analysis. He has published more than 10 papers in leading international journals such as Information Systems Research (UTD-24), INFORMS Journal on Computing (UTD-24), European Journal of Operational Research, and Information & Management. His publications have received over 1047 SCI citations, with a maximum of 655 citations for a single paper, and one paper was selected as an ESI Highly Cited Paper in 2024. He has presided over more than 10 research projects including the Youth Project and General Project of the National Natural Science Foundation of China, and the Youth Fund for Humanities and Social Sciences of the Ministry of Education. He has received several honours, including the Third Prize of the Outstanding Achievements in Humanities and Social Sciences Research of Shaanxi Provincial Universities (2021), Emerald Outstanding Paper Award (2021), and Nomination for Zhejiang Provincial Excellent Doctoral Dissertation (2017).

(Written by:Geng Ruibin Reviewed by:Jia Ming)