Beyond Conventional Optimization: The Theory of Non-Knowledge, Nullification, and Epistemological Optimality in Computational Systems

Conventional computational optimization operates through predefined objective functions, constraints, and evaluation criteria. Although such systems can identify optimal solutions relative to their formal specifications, they cannot thereby establish that the selected objectives represent epistemologically optimal outcomes. This limitation becomes particularly relevant when computational systems address complex linguistic, social, and human-centered concepts, such as success, well-being, efficiency, and value. This paper introduces an alternative conception of optimality based on the Theory of Non-Knowledge (TNK). Grounded in the epistemological problem represented by Agrippa's trilemma, TNK proposes nullification as an operation that assigns a non-contradictory property to a linguistic unit, transforming it into an X(NS) unit. This assignment does not eliminate the contradictions inherent in the concept or provide an ultimate justification for the operation itself. Rather, X(NS) represents the least contradictory possible epistemological formalization within the theory, including its practical implementations. The paper distinguishes this conception from conventional optimization and proposes a computational architecture for implementing nullification recursively across conceptual representations, objective functions, and decision-making procedures. The Traveling Salesperson Problem (TSP) is used to illustrate the distinction between mathematical optimality and epistemological optimality. The proposed architecture does not claim to compute an absolute optimum or eliminate infinitely many contradictions. Instead, it establishes a research framework for finite, iterative, and reflexive computational approximations of nullification. The contribution is a preliminary computational model in which optimality is not defined by maximizing a selected property but by the formal transformation of epistemologically problematic linguistic units into X(NS).