On the Viability of Hierarchical Causal Models in Network Analysis

The rapid evolution of Beyond 5G (B5G) and 6G networks introduces unprecedented complexity in network management due to their support for advanced services like realtime extended reality (XR) and large-scale IoT applications. Traditional AI models, which primarily rely on predictive and correlation-based approaches, struggle to capture the intricate interactions within these multi-layered systems. This paper explores Hierarchical Causal Models (HCMs) as an effective solution to model nested, cross-layer, and infrastructure-level causal relationships. We demonstrate how HCMs provide deeper insights into network behavior and performance optimization compared to flat models. Practical considerations for implementing HCMs are discussed, with attention to challenges such as model complexity and data requirements. In all, HCMs hold significant potential for improving network management in <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{6 G}$</tex> environments.