Collective-Memory-Driven Adaptation and Robustness in Hierarchical Network Systems

Understanding how multi-level systems adapt to changing environments is central to both natural and social sciences. While individuals learn and respond locally, their interactions can generate collective memory: a system-level property that shapes future responses and enhances robustness beyond the sum of individual behaviors. In this work, we investigate how collective memory emerges and functions across networked systems with different hierarchical structures. Using a hybrid framework that integrates differential equation modeling with agent-based network simulations, we model agents whose strategy probabilities evolve through local learning from neighbors and personal history. The probability updates are non-Markovian and history-dependent, enabling the system to encode information from past perturbations. We quantify emergent memory through a global fitness metric and analyze how robustness depends on network topology and the frequency of external shocks. Our results suggest that hierarchical networks exhibit enhanced memory retention and resilience, particularly under low-frequency perturbations, highlighting that memory encoding is inherently scale-dependent and geometrically constrained. These findings reveal collective memory as a unifying mechanism linking adaptation, resilience, and structure in complex systems, with implications for designing bio- and socially inspired decentralized architectures.