Towards Fair Graph Learning without Demographic Information

Zichong Wang, Nhat Hoang, Xingyu Zhang, Kevin Bello, Xiangliang Zhang, Sundararaja Sitharama Iyengar, Wenbin Zhang
Proceedings of The 28th International Conference on Artificial Intelligence and Statistics, PMLR 258:2107-2115, 2025.

Abstract

Fair Graph Neural Networks (GNNs) have been extensively studied in graph-based applications. However, most approaches to fair GNNs assume the full availability of demographic information by default, which is often unrealistic due to legal restrictions or privacy concerns, leaving a noticeable gap in methods for addressing bias under such constraints. To this end, we propose a novel method for fair graph learning without demographic information. Our approach leverages a Bayesian variational autoencoder to infer missing demographic information and uses disentangled latent variables to separately capture demographics-related and label-related information, reducing interference when inferring demographic proxies. Additionally, we incorporate a fairness regularizer that enables measuring model fairness without demographics while optimizing the fairness objective. Extensive experiments on three real-world graph datasets demonstrate the proposed method’s effectiveness in improving both fairness and utility.

Cite this Paper


BibTeX
@InProceedings{pmlr-v258-wang25f, title = {Towards Fair Graph Learning without Demographic Information}, author = {Wang, Zichong and Hoang, Nhat and Zhang, Xingyu and Bello, Kevin and Zhang, Xiangliang and Iyengar, Sundararaja Sitharama and Zhang, Wenbin}, booktitle = {Proceedings of The 28th International Conference on Artificial Intelligence and Statistics}, pages = {2107--2115}, year = {2025}, editor = {Li, Yingzhen and Mandt, Stephan and Agrawal, Shipra and Khan, Emtiyaz}, volume = {258}, series = {Proceedings of Machine Learning Research}, month = {03--05 May}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v258/main/assets/wang25f/wang25f.pdf}, url = {https://proceedings.mlr.press/v258/wang25f.html}, abstract = {Fair Graph Neural Networks (GNNs) have been extensively studied in graph-based applications. However, most approaches to fair GNNs assume the full availability of demographic information by default, which is often unrealistic due to legal restrictions or privacy concerns, leaving a noticeable gap in methods for addressing bias under such constraints. To this end, we propose a novel method for fair graph learning without demographic information. Our approach leverages a Bayesian variational autoencoder to infer missing demographic information and uses disentangled latent variables to separately capture demographics-related and label-related information, reducing interference when inferring demographic proxies. Additionally, we incorporate a fairness regularizer that enables measuring model fairness without demographics while optimizing the fairness objective. Extensive experiments on three real-world graph datasets demonstrate the proposed method’s effectiveness in improving both fairness and utility.} }
Endnote
%0 Conference Paper %T Towards Fair Graph Learning without Demographic Information %A Zichong Wang %A Nhat Hoang %A Xingyu Zhang %A Kevin Bello %A Xiangliang Zhang %A Sundararaja Sitharama Iyengar %A Wenbin Zhang %B Proceedings of The 28th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2025 %E Yingzhen Li %E Stephan Mandt %E Shipra Agrawal %E Emtiyaz Khan %F pmlr-v258-wang25f %I PMLR %P 2107--2115 %U https://proceedings.mlr.press/v258/wang25f.html %V 258 %X Fair Graph Neural Networks (GNNs) have been extensively studied in graph-based applications. However, most approaches to fair GNNs assume the full availability of demographic information by default, which is often unrealistic due to legal restrictions or privacy concerns, leaving a noticeable gap in methods for addressing bias under such constraints. To this end, we propose a novel method for fair graph learning without demographic information. Our approach leverages a Bayesian variational autoencoder to infer missing demographic information and uses disentangled latent variables to separately capture demographics-related and label-related information, reducing interference when inferring demographic proxies. Additionally, we incorporate a fairness regularizer that enables measuring model fairness without demographics while optimizing the fairness objective. Extensive experiments on three real-world graph datasets demonstrate the proposed method’s effectiveness in improving both fairness and utility.
APA
Wang, Z., Hoang, N., Zhang, X., Bello, K., Zhang, X., Iyengar, S.S. & Zhang, W.. (2025). Towards Fair Graph Learning without Demographic Information. Proceedings of The 28th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 258:2107-2115 Available from https://proceedings.mlr.press/v258/wang25f.html.

Related Material