基于视图解耦与反事实增强的公平图学习

doi:10.16088/j.issn.1001-6600.2025122801

摘要/Abstract

摘要： 图神经网络(GNN)因其对图结构数据的强大建模能力,已在诸多真实场景中得到广泛应用。近年来的研究表明,GNNs可能会继承并放大训练数据中固有的偏差,从而使得由敏感属性定义的特定群体受到不公平对待。为缓解图数据中的偏差问题,本文提出一种基于视图解耦与反事实增强的公平图学习框架FairDC,分别针对特征偏差和结构偏差进行去偏。该框架首先将原始图数据解耦为特征视图和结构视图,然后在此基础上引入反事实视图,最后采用多视图融合策略学习公平节点表示。在多个基准数据集上的实验结果表明,FairDC在保持预测性能基本稳定的同时,相比于最新基线方法DAB,公平性指标人口平等(DP)和机会均等(EO)分别下降32%和36%,验证了其在效用与公平性之间的有效权衡能力。

关键词: 图神经网络, 公平性, 解纠缠, 反事实增强, 多视图融合

Abstract: Graph Neural Network (GNN) have been widely applied in numerous real-world scenarios due to their powerful modeling capabilities for graph-structured data. Recent studies indicate that GNN may inherit and amplify biases inherent in training data, leading to unfair treatment of specific groups defined by sensitive attributes. To mitigate bias in graph data, this paper proposes FairDC, a fair graph learning framework based on view decoupling and counterfactual augmentation, which addresses feature bias and structural bias separately. The framework first decouples raw graph data into feature views and structural views, then introduces counterfactual views. Finally, a multi-view fusion strategy is employed to learn fair node representations. Experimental results on multiple benchmark datasets demonstrate that FairDC maintains stable prediction performance while reducing fairness metrics DP and EO by 32% and 36%, respectively, compared with the state-of-the-art baseline DAB. This validates the proposed method’s effective trade-off between utility and fairness.

Key words: graph neural network, fairness, disentanglement, counterfactual augmentation, multi-view fusion

中图分类号: TP18

韦吴杰, 陈庆锋. 基于视图解耦与反事实增强的公平图学习[J]. 广西师范大学学报（自然科学版）, 2026, 44(4): 96-106.

Wei Wujie, Chen Qingfeng. Fair graph learning via view disentanglement and counterfactual augmentation[J]. Journal of Guangxi Normal University(Natural Science Edition), 2026, 44(4): 96-106.

参考文献

[1] 鲁晓颖, 郑国勋. 基于图神经网络的社交网络数据挖掘算法[J]. 兰州文理学院学报(自然科学版), 2025, 39(6): 60-65. DOI: 10.13804/j.cnki.2095-6991.2025.06.003.
[2] Chen Q F, Wei B Q, Cheng D B, et al. A novel shadow variable catcher for addressing selection bias in recommendation systems[C]//2024 IEEE International Conference on Data Mining (ICDM). Los Alamitos, CA: IEEE Computer Society, 2024: 71-80. DOI: 10.1109/ICDM59182.2024.00014.
[3] 贾子硕, 张俭鸽, 贺浩峰, 等. 大模型与知识图谱互增技术与应用综述[J/OL]. 计算机科学[2025-12-28]. https://link.cnki.net/urlid/50.1075.TP.20251201.1003.004.
[4] Zhu Y H, Feng L, Deng Z Y, et al. Robust node classification on graph data with graph and label noise[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2024, 38(15): 17220-17227. DOI: 10.1609/aaai.v38i15.29668.
[5] Li S Y, Liu Y X, Chen Q F, et al. Noise-resilient unsupervised graph representation learning via multi-hop feature quality estimation[C]//CIKM’24: Proceedings of the 33rd ACM International Conference on Information and Knowledge Management. New York, NY: Association for Computing Machinery, 2024: 1255-1265. DOI: 10.1145/3627673.3679758.
[6] Liu Y X, Li S Y, Zheng Y, et al. ARC: a generalist graph anomaly detector with in-context learning[C]//Advances in Neural Information Processing Systems 37 (NeurIPS 2024). Red Hook, NY: Curran Associates Inc., 2024: 50772-50804. DOI: 10.52202/079017-1606.
[7] Zhu J, Li G T, Yang Y A, et al. On the impact of feature heterophily on link prediction with graph neural networks[C]//Advances in Neural Information Processing Systems 37 (NeurIPS 2024). Red Hook, NY: Curran Associates Inc., 2024: 65823-65851. DOI: 10.52202/079017-2104.
[8] Kipf T N, Welling M. Semi-supervised classification with graph convolutional networks[PP/OL]. V4.arXiv(2017-02-22)[2025-12-28]. https://doi.org/10.48550/arXiv.1609.02907.
[9] Veličković P, Cucurull G, Casanova A, et al. Graph attention networks[PP/OL]. V3.arXiv(2018-02-04)[2025-12-28]. https://doi.org/10.48550/arXiv.1710.10903.
[10] Xu K, Hu W H, Leskovec J, et al. How powerful are graph neural networks[PP/OL]. V3.arXiv(2019-02-22)[2025-12-28]. https://doi.org/10.48550/arXiv.1810.00826.
[11] Hamilton W L, Ying R, Leskovec J. Inductive representation learning on large graphs[C]//Advances in Neural Information Processing Systems 30 (NIPS 2017). Red Hook, NY: Curran Associates Inc., 2017: 1025-1035.
[12] Dong Y S, Ma J, Wang S, et al. Fairness in graph mining: a survey[J]. IEEE Transactions on Knowledge and Data Engineering, 2023, 35(10): 10583-10602. DOI: 10.1109/TKDE.2023.3265598.
[13] 周璟昊, 石磊, 石拓, 等. 融合多维特征的电诈犯罪时空预测研究[J]. 智能系统学报, 2025, 20(5): 1112-1122. DOI: 10.11992/tis.202412025.
[14] 曹伟萍, 张劲松. 基于不平衡数据处理与加权软投票异质集成的农户贷款违约风险预测[J]. 计算机应用与软件, 2025, 42(8): 71-79. DOI: 10.3969/j.issn.1000-386x.2025.08.010.
[15] Dai E Y, Wang S H. Say no to the discrimination: learning fair graph neural networks with limited sensitive attribute information[C]//WSDM’21: Proceedings of the 14th ACM International Conference on Web Search and Data Mining. New York, NY: Association for Computing Machinery, 2021: 680-688. DOI: 10.1145/3437963.3441752.
[16] Wang Y, Zhao Y Y, Dong Y S, et al. Improving fairness in graph neural networks via mitigating sensitive attribute leakage[C]//KDD’22: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Minin. New York: Association for Computing Machinery, 2022: 1938-1948. DOI: 10.1145/3534678.3539404.
[17] Agarwal C, Lakkaraju H, Zitnik M. Towards a unified framework for fair and stable graph representation learning[C]//Proceedings of the Thirty-Seventh Conference on Uncertainty in Artificial Intelligence: PMLR 161. Cambridge, MA: JMLR, 2021: 2114-2124.
[18] Ma J, Guo R C, Wan M T, et al. Learning fair node representations with graph counterfactual fairness[C]//WSDM’22: Proceedings of the Fifteenth ACM International Conference on Web Search and Data Mining. New York, NY: Association for Computing Machinery, 2022: 695-703. DOI: 10.1145/3488560.3498391.
[19] Kipf T N, Welling M. Variational graph auto-encoders[PP/OL]. V1.arXiv(2016-11-21)[2025-12-28]. https://doi.org/10.48550/arXiv.1611.07308.
[20] Guo Z M, Li J L, Xiao T, et al. Towards fair graph neural networks via graph counterfactual[C]//CIKM’23: Proceedings of the 32nd ACM International Conference on Information and Knowledge Management. New York, NY: Association for Computing Machinery, 2023: 669-678. DOI: 10.1145/3583780.3615092.
[21] Spinelli I, Scardapane S, Hussain A, et al. FairDrop: biased edge dropout for enhancing fairness in graph representation learning[J]. IEEE Transactions on Artificial Intelligence, 2022, 3(3): 344-354. DOI: 10.1109/TAI.2021.3133818.
[22] Dong Y S, Liu N H, Jalaian B, et al. EDITS: modeling and mitigating data bias for graph neural networks[C]//WWW’22: Proceedings of the ACM Web Conference 2022. New York, NY: Association for Computing Machinery, 2022: 1259-1269. DOI: 10.1145/3485447.3512173.
[23] Liu J, Sun L L, Nie W Z, et al. Graph disentangled contrastive learning with personalized transfer for cross-domain recommendation[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2024, 38(8): 8769-8777. DOI: 10.1609/aaai.v38i8.28723.
[24] Ye G Z, Wu W, Wang G Q, et al. Disentangled modeling of preferences and social influence for group recommendation[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2025, 39(12): 13052-13060. DOI: 10.1609/aaai.v39i12.33424.
[25] Zhu Y C, Li J T, Zheng Z B, et al. Fair graph representation learning via sensitive attribute disentanglement[C]//WWW’24: Proceedings of the ACM Web Conference 2024. New York, NY: Association for Computing Machinery, 2024: 1182-1192. DOI: 10.1145/3589334.3645532.
[26] Lee Y C, Shin H, Kim S W. Disentangling, amplifying, and debiasing: learning disentangled representations for fair graph neural networks[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2025, 39(11): 12013-12021. DOI: 10.1609/aaai.v39i11.33308.
[27] Li Z X, Dong Y S, Liu Q, et al. Rethinking fair graph neural networks from re-balancing[C]//KDD’24: Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. New York, NY: Association for Computing Machinery, 2024: 1736-1745. DOI: 10.1145/3637528.3671826.
[28] Yang C, Liu J X, Yan Y H, et al. FairSIN: achieving fairness in graph neural networks through sensitive information neutralization[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2024, 38(8): 9241-9249. DOI:10.1609/aaai.v38i8.28776.
[29] Wang X M, Gu T L, Bao X G, et al. Individual fairness for local private graph neural network[J]. Knowledge-Based Systems, 2023, 268: 110490. DOI: 10.1016/j.knosys.2023.110490.
[30] Lin Y J, Zhao C, Shao M L, et al. Towards counterfactual fairness-aware domain generalization in changing environments[C]//Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence (IJCAI-24). Jeju: IJCAI, 2024: 4560-4568. DOI: 10.24963/ijcai.2024/504.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed