Metabolic Attention Layer (MAD-ATT)

MAD-ATT (Metabolic Attention Layer) is an operational extension of Metabolic Adaptive Dynamics (MAD v1.2) introducing a macroscopic model of attention as a metabolic dynamical system. The framework models reasoning dynamics using three macroscopic variables: • Attention Pressure (P) — conceptual consolidation• Attention Volume (V) — exploration breadth• Cognitive Temperature (T) — fluctuation amplitude These variables define a cognitive state relation P · V = kT and evolve through coupled dynamical equations governing accumulation, instability, and release processes. Within this formulation, reasoning is interpreted as movement through a conceptual attractor landscape characterized by cyclic transitions between: exploration → pressure accumulation → instability → release → stabilization. MAD-ATT introduces a computational method for detecting these metabolic phases in textual or dialogical streams using proxy metrics derived from linguistic structure. The framework connects adaptive cognitive dynamics with concepts from: • dynamical systems theory• complex adaptive systems• thermodynamic analogies of regulation• AI reasoning monitoring A minimal computational prototype (MAD-ATT Dashboard) is provided alongside this work as an open instrumentation tool for detecting phase transitions in reasoning processes. Potential applications include: • analysis of scientific idea evolution• monitoring of dialogue dynamics• AI reasoning stability and hallucination detection• human-AI collaborative reasoning systems MAD-ATT operationalizes MAD v1.2 by transforming an abstract adaptive cognition model into a candidate cognitive instrumentation framework for observing and regulating reasoning dynamics in complex adaptive systems. First public release of the MAD-ATT framework. Keywords: cognitive dynamics, dynamical systems cognition, reasoning phase transitions, attractor dynamics, attention dynamics, complex adaptive systems, AI reasoning monitoring