The Dependent Origination Engine and I-Gi Controller: A Control-Theoretic Framework for AI Alignment via Buddhist-Confucian Philosophy

This paper introduces the Dynamic Moral Attention System (DMAS), a conceptual AI alignment framework that bridges East Asian metaphysics and modern control theory. Rather than treating alignment as static rule enforcement (e.g., RLHF or Constitutional AI), the framework reframes alignment as dynamic restoration: generated outputs (Gi, 氣) continuously converging toward an external principle (I, 理). Drawing on Buddhist dependent origination and Neo-Confucian I–Gi metaphysics, the system integrates four components: a nonlinear generative model, a Sadan-based PD controller, a Gyeong adaptive gain mechanism, and a Gyeokmul Chiji observer for estimating the underlying principle. A proof-of-concept simulation on a double integrator plant demonstrates that the Gyeong adaptive gain stabilizes system behavior under noise, enabling flexible control in calm conditions and principled correction under disturbance. The result is a philosophical–engineering synthesis that reimagines alignment as a continuous process of convergence between principle and manifestation.