An Explainable AI Framework for Dynamic Resource Management in Vehicular Network Slicing

AI decisions in critical infrastructure can be made transparent: The framework demonstrates that deep reinforcement learning systems managing vehicular network resources can explain their decisions in real-time using Shapley values and attention mechanisms, moving beyond "black box" AI in safety-critical applications.

Explainability improves performance: By making the AI's reasoning visible and interpretable, the system achieved measurable improvements in service quality—URLLC (ultra-reliable, low-latency) service satisfaction increased from 78.0% to 80.13%, suggesting that transparency and performance are complementary rather than trade-offs.

Different service types require different priorities: The research addresses how AI must balance competing demands in vehicular networks (high-speed data versus ultra-reliable communications), highlighting that responsible AI systems need to manage multiple, sometimes conflicting objectives fairly and transparently.

Interpretability methods matter for accountability: The paper shows that feature-based explainability using Shapley values outperformed simpler attention mechanisms alone, indicating that the choice of explanation method significantly affects how well stakeholders can understand and trust AI resource allocation decisions.

Real-time transparency is feasible for dynamic systems: The framework operates with "near-real-time" explanations while managing constantly changing network conditions, demonstrating that explainability doesn't require sacrificing responsiveness—important for applications where both speed and accountability are critical.