Explainability as an Emergent Property in AI-Enabled System-of-Systems: A Case Study of Autonomous Flight Plan Approval in U-Space

Explainability has become an essential requirement for trustworthy artificial intelligence (AI) in safety-critical operational environments. However, most explainable AI research focuses on interpreting the internal behaviour of machine learning models, while the influence of the broader operational environment is often overlooked. In complex operational ecosystems such as U-Space for drone traffic management, AI components operate within a distributed System-of-Systems (SoS) composed of multiple interacting services, including surveillance systems, airspace management services, environmental information providers, and operator inputs. This paper investigates explainability as an emergent property from a SoS perspective using the AI4HyDrop autonomous flight plan approval service as a case study. The system employs hierarchical reinforcement learning (HRL) to resolve trajectory conflicts in a simulated U-Space environment. Explainability is provided using SHapley Additive exPlanations (SHAP) to analyze how contextual features influence the selection of conflict resolution strategies. A representative conflict scenario involving an area-based and a linear trajectory is analyzed to reconstruct the reasoning behind the system's decision. The results show that the dominant explanatory factors originate from distributed operational services, including conflict geometry, temporal interaction between trajectories, and operator preferences. These findings indicate that the explanation of the decision emerges from the interaction of multiple constituent systems rather than from the AI model alone. The study therefore extends the notion of explainability as an emergent property beyond the internal architecture of AI systems to the broader SoS context in which AI-enabled services operate. This perspective provides important insights for designing transparent and trustworthy autonomous decision-making services in complex systems.