Decision Engineering: The Architecture of Decision Systems

This paper introduces Decision Engineering as an architectural approach to the design of decision systems in artificial intelligence and complex environments. Current AI systems are predominantly structured around predictive capabilities, focusing on the estimation of future states. While prediction enables modeling and forecasting, it does not, on its own, provide a formal structure for decision-making. Real-world systems must operate under uncertainty, constraints, and competing objectives, requiring explicit mechanisms for action selection. This work identifies a fundamental gap between prediction and decision-making and proposes Decision Engineering as a framework for addressing this gap. The paper formalizes the concept of a decision layer as a distinct and irreducible component of intelligent systems, responsible for transforming representations and predictions into actionable decisions. A simplified layered architecture of decision systems is presented, consisting of representation, prediction, decision logic, execution, and feedback. Within this structure, predictive and simulation components are embedded within a broader decision architecture that defines objectives, constraints, and accountability mechanisms. The paper also introduces the concept of decision quality as a central metric for evaluating system performance. Unlike traditional approaches that emphasize predictive accuracy, decision quality captures the structural adequacy of decisions in relation to information, alignment, transparency, and risk. By reframing artificial intelligence systems as decision systems, Decision Engineering provides a foundation for more robust, accountable, and governable AI. The proposed architecture has implications for system design, evaluation, and regulation, as well as for the integration of human and artificial decision-making processes.