AI2030 Responsible AI Tools Marketplace

Key Takeaways

GNNs lack transparency in decision-making: Graph Neural Networks are increasingly used in important applications, but their complexity makes it difficult to understand how they reach their conclusions—a critical gap for responsible AI deployment.
Current explanation methods have significant trade-offs: Post-hoc explanation approaches (added after model training) require extra computational resources and may be unreliable, while self-interpretable models that provide immediate explanations struggle to work reliably across different scenarios.
Graph structure matters more than individual features: The paper emphasizes that understanding how the overall structure of data relationships influences predictions is more important than analyzing individual data points in isolation, suggesting that effective AI explanations for graph-based systems need to capture these complex interactions.
A new framework is needed for graph-based AI: The research proposes developing an explainability framework specifically designed for graph neural networks that balances computational efficiency with reliability—addressing a gap in current responsible AI practices for this increasingly common type of model.

Abstract

Graph Neural Networks (GNNs) have become a powerful tool for modeling and analyzing data with graph structures. The wide adoption in numerous applications underscores the value of these models. However, the complexity of these methods often impedes understanding their decision-making processes. Current Explainable AI (XAI) methods struggle to untangle the intricate relationships and interactions within graphs. Several methods have tried to bridge this gap via a post-hoc approach or self-interpretable design. Most of them focus on graph structure analysis to determine essential patterns that correlate with prediction outcomes. While post-hoc explanation methods are adaptable, they require extra computational resources and may be less reliable due to limited access to the model's internal workings. Conversely, Interpretable models can provide immediate explanations, but their generalizability to different scenarios remains a major concern. To address these shortcomings, this thesis seeks to develop a novel XAI framework tailored for graph-based machine learning. The proposed framework aims to offer adaptable, computationally efficient explanations for GNNs, moving beyond individual feature analysis to capture how graph structure influences predictions.

157 pages, Doctoral dissertation at Seoul National University (submitted in 2024.08 to SNU library, slightly updated in 2025.11 for open digital version)

Key Takeaways

GNNs lack transparency in decision-making: Graph Neural Networks are increasingly used in important applications, but their complexity makes it difficult to understand how they reach their conclusions—a critical gap for responsible AI deployment.
Current explanation methods have significant trade-offs: Post-hoc explanation approaches (added after model training) require extra computational resources and may be unreliable, while self-interpretable models that provide immediate explanations struggle to work reliably across different scenarios.
Graph structure matters more than individual features: The paper emphasizes that understanding how the overall structure of data relationships influences predictions is more important than analyzing individual data points in isolation, suggesting that effective AI explanations for graph-based systems need to capture these complex interactions.
A new framework is needed for graph-based AI: The research proposes developing an explainability framework specifically designed for graph neural networks that balances computational efficiency with reliability—addressing a gap in current responsible AI practices for this increasingly common type of model.

Abstract

157 pages, Doctoral dissertation at Seoul National University (submitted in 2024.08 to SNU library, slightly updated in 2025.11 for open digital version)

Enhancing Explainability of Graph Neural Networks Through Conceptual and Structural Analyses and Their Extensions

Key Takeaways

Abstract

Paper Info

Responsible AI Topics

Links

Enhancing Explainability of Graph Neural Networks Through Conceptual and Structural Analyses and Their Extensions

Key Takeaways

Abstract

Paper Info

Responsible AI Topics

Links