Title: An Adversarial Graph Learning Approach for Networked Actor Comparative Assessment

Speaker: Chen Xi, Professor, Zhejiang University

Host: Han Xuewen, Associate Professor, Antai College of Economics and Management, Shanghai Jiao Tong University

Time: 13:30–15:00, Thursday, May 7, 2026

Venue: Room 403, Antai Building, Xuhui Campus, Shanghai Jiao Tong University

Brief introduction of the content: 

In this study, we propose a general causal inference problem — the comparative assessment of social influence across network actors — which aims to estimate the differential social influence that various network actors exert on their connected peers. A critical yet underexplored challenge in this comparative assessment is the issue of network distribution imbalance when estimating individual influence effects: different actors exhibit systematic variations in their social ties and peer characteristics. This imbalance often stems from actors’ heterogeneous tie-formation strategies, which may introduce systematic bias into causal evaluations based on individual outcomes. We examine how such imbalances affect the assessment of social influence across different types of actors and how they can be addressed. Specifically, we propose the Adversarial Graph-based Counterfactual Estimation (AGCE) framework — a novel framework for cross-actor social influence comparison in observational networks. AGCE is a unified adversarial learning framework that integrates graph neural network–based imbalance identification, counterfactual scenario simulation, and outcome prediction under balanced network conditions. Extensive experiments demonstrate that AGCE outperforms existing methods in recovering causal differences and exhibits robustness across various settings. We apply the AGCE framework to multiple Twitter discussion domains to reveal differences in social influence between algorithm-driven intelligent agents and human actors. Furthermore, we show that stakeholders can use the AGCE framework for networked counterfactual analysis and behavioral prediction. The findings of this study provide policy tools for existing social commerce operations, such as early assessment of social account influence and influencer selection.

Speaker's profile:

Chen Xi is currently a Professor at the School of Management, Zhejiang University. His research interests include social media, social network analysis, and social commerce. In these areas, Professor Chen has served as Principal Investigator for multiple national and provincial/ministerial research projects, including Key Projects, Major Research Plan projects, and General Projects supported by the National Natural Science Foundation of China (NSFC). His work has been published or accepted in leading international and domestic journals such as Information Systems Research (ISR), Production and Operations Management (POM), INFORMS Journal on Computing (INFORMS JoC), Journal of Management Information Systems (JMIS), Journal of the Association for Information Systems (JAIS), European Journal of Operational Research (EJOR), Journal of the Association for Information Science and Technology (JASIST), Management World, and Journal of Industrial Engineering and Engineering Management.

Professor Chen serves as Associate Editor for two SSCI Q1 journals: Information & Managementand Industrial Management & Data Systems. He also serves on the Editorial Board of Journal of Management Science and Engineering and has served repeatedly as Track Associate Editor for international conferences such as the International Conference on Information Systems (ICIS).

The policy reports authored by Professor Chen have received approval and endorsements from national and provincial/ministerial leaders. He has also provided consulting services for well-known companies such as Taobao, DAMO Academy, and Unilever, achieving notable practical outcomes.

Welcome all!