Value Anchoring, Axiomatization of Eastern and Western Ethical Genealogies, and Endogenous AGI Alignment: A Cognitive Dynamics Approach

Mainstream approaches to AI value alignment have long relied on externally imposed constraints: human-labeled preferences, post-hoc reward fine-tuning, and rule-based prohibitions. The fundamental predicament of these approaches lies in the arbitrariness and fragmentation of their value sources: values are treated as contingent cultural products of humanity or as collections of subjective preferences, lacking a stable, self-consistent, and derivable foundational basis. Grounded in the framework of cognitive dynamics, this paper proposes a set of ethical axioms: ethical norms are not externally imposed regulations, but endogenous dynamical constraints by which higher-order self-stabilizing cognitive systems—carbon-based, silicon-based, and carbon-silicon coupled systems—maintain long-range cooperative stability. The paper first undertakes an axiomatic reduction of Eastern virtue ethics and the Western traditions of deontology, utilitarianism, and virtue ethics, identifying the presuppositions, domains of application, and inherent limitations of each system. It then argues that an ethical axiom system centered on the Reverence-for-Heaven Axiom and the Self-Interest-and-Altruism Axiom can serve as the foundational anchor for endogenous AGI alignment, transforming value from an external patch into an intrinsic stability condition of system evolution. This offers a new paradigm for AI ethics, distinct from preference fitting and rule enforcement.