TEFL: Turbo Explainable Federated Learning for 6G Trustworthy Zero-Touch Network Slicing

Explainability is essential for AI trustworthiness in critical infrastructure: The paper demonstrates that AI systems managing 6G networks—which support essential services across multiple sectors—require built-in explainability tools to create transparency between network operators, infrastructure providers, and service users, rather than relying on opaque "black-box" AI decisions.

Federated learning enables AI without centralizing sensitive network data: By using federated learning, the approach allows AI models to be trained across distributed network locations while keeping sensitive operational data decentralized, addressing privacy concerns that arise when multiple organizations share network infrastructure.

Explanations can be validated for reliability: The paper introduces a "confidence metric" based on attribution methods (Integrated Gradient, SHAP) to measure whether AI explanations are actually faithful and trustworthy, rather than simply plausible-sounding—a critical safeguard for high-stakes network management decisions.

Responsible AI requires iterative feedback loops: The "turbo" approach uses closed-loop communication between the AI model and an explainer component to continuously improve both performance and transparency, suggesting that trustworthy AI systems benefit from ongoing dialogue between decision-making and explanation mechanisms rather than one-time audits.

Technical constraints can enforce responsible behavior: By embedding SLA (service level agreement) compliance and explanation confidence directly into the AI optimization process, the system ensures responsible outcomes are built into the model's objectives from the start, not added as an afterthought.