System-Forming Factor for Artificial Intelligence Systems: A Security-Oriented Synthesis Methodology

The paper addresses a critical gap in artificial intelligence (AI) design: the absence of a formal condition that guarantees the achievement of the intended goal of an AI system. Because AI is viewed as the automation of rational human behavior, the foundation of such behavior is a decision-making process. The decision maker (DM) operates with the categories of system, model, and purpose, which demands a systemforming factor that ensures the correct construction and functioning of the system. We propose a methodology grounded in the natural-scientific approach and based on the Law of Conservation of Object Integrity (LCOI). This law serves as the system-forming factor and yields an analytical condition for the existence of the AI system’s operation. The model integrates four processes: the target process, problem emergence, problem identification, and problem neutralization. The performance indicator is the probability that every problem is both detected and eliminated. The synthesis procedure is formulated, enabling the engineering of processes with specified properties and thus providing a guarantee of goal achievement. Numerical experiments confirm the main regularities of the system functioning.