AI Agent Plasticity: A System Dynamics Model of Developmental Artificial Agency

This paper introduces AI Agent Plasticity (AIA Plasticity) as a core developmental variable in artificial agency. Rather than treating the distinction between large language models and AI agents as categorical, the study reframes artificial agency as a dynamic systems process. AIA Plasticity is defined as the capacity of an artificial agent to reorganize its enacted worldview architecture in response to environmental interaction across time. Using a system dynamics framework, the paper develops a three-layer causal model. The first layer describes a reinforcing plasticity expansion loop: environmental novelty generates worldview–environment mismatch, increasing adaptation pressure and investment in plasticity, thereby improving adaptive fit and expanding operational scope. The second layer models the stability–plasticity tradeoff. Elevated plasticity accelerates worldview change, which may reduce coherence and behavioral reliability, triggering stabilization mechanisms. During high-plasticity phases, overcorrection dynamics can produce oscillatory restructuring before convergence toward a new equilibrium. The third layer situates artificial agents within a co-evolutionary environment. Increasing deployment reshapes environmental complexity, intensifies selection pressure for adaptivity, and stimulates further plasticity demand, while governance pressure constrains permissible action space. The integrated model predicts developmental waves characterized by adaptive expansion, temporary instability, and re-stabilization at higher levels of structural integration. Developmental direction emerges from adaptive coupling between agents and dynamic environments rather than from teleological progression. This framework provides a foundation for a developmental systems theory of artificial agency and outlines pathways for subsequent formal modeling and empirical testing.