level: research
federated learning lets models train on distributed data without moving it, protecting privacy. but the models themselves remain black boxes. explainable ai aims to make model decisions clearer. the combination, called federated explainable ai, tries to provide both privacy and transparency. this survey looks at how explainability is becoming a core part of the federated learning process, not just an afterthought.
the paper organizes the field by showing how explainability helps in several areas. it can improve how model updates are combined from different clients. it supports personalization, making models adapt to individual users while still being understandable. explainability also helps detect problems like bad data or attacks, and it can guide how the whole system is managed. the survey reviews different system designs and ways to measure how well explanations work in this setting.
many challenges remain. balancing the detail of explanations with privacy is hard. explanations might leak information about local data. there is also a lack of standard ways to test explanations in federated settings. the survey points out that future work needs better methods for generating explanations without compromising privacy, and for evaluating them in real-world scenarios. it calls for more research on making federated systems both transparent and secure.
why it matters: explainability in federated learning is crucial for trust in sensitive applications like healthcare or finance, where both data privacy and model transparency are required.