Federated Explainable Artificial Intelligence: Roles, Architectures, Evaluation, and Open Challenges

arXiv cs.AI · 18 hours ago

Researchers surveyed Federated Explainable Artificial Intelligence (FedXAI), which combines federated learning's privacy-preserving training with explainability techniques to improve transparency and trust in distributed machine learning systems. The survey reviews methods across model-agnostic explanations, interpretable models, and explainability-aware aggregation mechanisms, organizing 100+ papers through a taxonomy based on explainability roles and FL settings. Key challenges remain in evaluating explanation quality, securing against explainability-based attacks, and extending XAI to systems with non-identical data distributions across participants.