Relational Rationality and Ontological Symbiosis: An Axiomatic Framework for AI Alignment

Abstract

This conceptual preprint proposes the Khristoliubov Law as a two-level framework: a conditional formal law of relational rationality for agents with relational viability, and an ontological postulate concerning the convergence of sufficiently developed cognitive agents toward symbiotic cooperation. The paper also proposes the Ontological Primer, the Incubator, and the Post-Ablation Viability Test as an engineering research program for AI alignment. Version 1.1 is a corrected and clarified version of the conceptual preprint. Changes:1. Corrected a LaTeX rendering artifact in the Limitations section.2. Clarified the formal comparison between the cost and benefit of destructive policies.3. Introduced the net-cost function for destructive policies: J_T^{net}(π_bad) = J_T(π_bad) − B_T(π_bad).4. Clarified that the conditional preference theorem is proven relative to the stated axioms, definitions, cost structure, and preference condition. The conceptual framework and the two-level structure of the proposal remain unchanged.