A computational approach to cognitive architecture with adaptive networks and entropy-driven dynamics

This study presents a computational approach to cognitive architecture that integrates adaptive networks with entropy-driven dynamics to model the evolution and stabilization of mental states. The framework combines temporal–causal network modeling with entropy-based mechanisms inspired by Maxwell–Boltzmann distributions, allowing the system to balance local neural adaptability with global cognitive stability. Using a simulated environment, the model captures interactions among self-related cognitive constructs, including self-esteem, self-efficacy, and self-concept, and is examined through a case study on workplace self-confidence. The simulations reveal distinct dynamic patterns across scenarios characterized by high self-esteem, high self-efficacy, and high self-concept, reproducing theoretically consistent behaviors such as convergence toward stable cognitive states, sensitivity to feedback conditions, and a separation between short-term fluctuations and long-term stabilization. Overall, these results suggest that integrating adaptive networks with entropy-driven dynamics offers a biologically inspired and computationally efficient framework for modeling complex cognitive processes, with potential applications in intelligent forecasting, diagnostics, and adaptive decision-support systems, as well as a scalable basis for developing human-aligned cognitive architectures and supporting AI–human collaboration.