Decision Engineering Science From Prediction to Decision The Missing Layer Between Neuroscience and Artificial Intelligence A foundational framework for engineering decision systems beyond predictive models.

This working paper introduces Decision Engineering Science (DES) as a foundational framework for the design, analysis, and evaluation of decision systems across biological, artificial, and organizational domains. While contemporary research in neuroscience and artificial intelligence has significantly advanced the understanding of perception, learning, and prediction, the structural foundations of decision-making remain insufficiently formalized.

This work addresses a critical gap: the absence of a unified architecture that treats decision-making as an engineered system rather than an emergent property of predictive or adaptive processes.

Drawing on insights from Predictive Coding, Reinforcement Learning, and Signal Detection Theory, the paper demonstrates that both biological and artificial systems operate primarily as predictive systems, lacking explicit mechanisms for structured decision-making. These systems optimize representation accuracy or reward signals, but do not guarantee decision quality, alignment, or stability over time.

To address these limitations, the paper formalizes decision systems as structured entities defined by state spaces, action sets, constraints, transition dynamics, and governance mechanisms. It introduces the Architectural Primacy Principle (APP), which establishes that normative structures—objectives, constraints, and governance—must precede and constrain predictive and optimization layers within any coherent decision system.

A central contribution of this work is the introduction of the Decision Quality Index (DQI), a composite metric that captures the interaction between information quality, alignment, transparency, and risk. In addition, the paper defines the concept of decision drift, describing the temporal divergence between explicit system objectives and implicit adaptive dynamics.

By mapping neural processes, artificial intelligence systems, and engineered decision architectures onto a unified framework, this work identifies structural gaps that limit current approaches and proposes DES as a control layer that enables the systematic design of aligned, auditable, and robust decision systems.

The implications of this framework extend to artificial intelligence, autonomous systems, organizational decision-making, and regulatory contexts, including alignment with emerging standards such as the EU AI Act. At a broader level, the paper contributes to the conceptual foundations of cognitive infrastructure and the emerging cognitive economy, where value is increasingly determined by the quality of decisions rather than the quantity of information.

Decision Engineering Science is positioned not as an extension of existing paradigms, but as a new discipline that transforms decision-making from an emergent phenomenon into a designed and measurable capability.