Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
September 20, 2025

Causality-Inspired Disentanglement for Fair Graph Neural Networks

View Full Paper
Ask AI
Bookmark
Share

Authors

GZGuixian ZhangChina University of Mining and TechnologyDCDebo ChengGuangxi University of Finance and EconomicsGYGuan YuanChina University of Mining and Technology

Discussion

Loading...

Member takes

Overview

Proposed framework improves fair representation by disentangling sensitive attributes from causal factors.

Key Points

  • The framework uses counterfactual data generation to separate causal and sensitive factors in predictions.
  • CDFG outperforms traditional methods on three datasets, enhancing both fairness and utility.
  • Causal representation is extracted to ensure fairness in graph-based predictions, addressing optimisation challenges.
  • The approach provides a means to achieve label independence from sensitive attributes, reducing biases.

Cite This Study

Zhang et al. (2025) studied this question.

synapsesocial.com/papers/68d46aa631b076d99fa67323https://doi.org/10.24963/ijcai.2025/72
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Towards Fair Graph Neural Networks via Counterfactual and Balance2025 · 1 citations
  2. 2One Fits All: Learning Fair Graph Neural Networks for Various Sensitive Attributes2024 · 9 citations
  3. 3Fair Graph Neural Network with Supervised Contrastive Regularization2024
  4. 4One Fits All: Learning Fair Graph Neural Networks for Various Sensitive Attributes2024
  5. 5Fair Graph Representation Learning via Sensitive Attribute Disentanglement2024