This preprint presents a theoretical and methodological framework for describing cognitive and regulatory dynamics, referred to as the metabolic model. The work focuses on clarifying the conceptual status of attentional pressure, defining it as a model-level state variable that reflects structural priority within an internal system model rather than perceptual attention or phenomenological salience. The framework explicitly separates multiple levels of description (ontological, model, phenomenological, and interpretative), adopts a constraint-based and dynamical interpretation of causality, and employs thermodynamic terminology as a formal analogy rather than a claim of physical instantiation. These methodological choices are intended to avoid category errors, causal ambiguity, and overextension of physical metaphors. The proposed model does not offer empirical validation, measurement protocols, or computational implementations. Its scope is limited to systems characterized by multiple internal states, partial observability, and regulatory mechanisms governing relevance and stability. Transferability across domains—such as human cognition, biological regulation, or artificial systems—is treated in terms of structural correspondence rather than direct analogy. This work is intended as a framework paper that provides a coherent conceptual vocabulary and methodological baseline for future theoretical development, comparative modeling, and potential empirical or computational operationalization.