Independent Research

The Salience
Engine

A coordination architecture for precision-weighted mode dynamics

A formal framework proposing that cognitive coordination can be modeled as motion through a salience-structured state space — governed by a five-stage pipeline and precision-weighted transitions between metastable processing modes.

Pressure Parsing Salience Mode Action
Preprint 2026 v1.1
The Salience Engine: A Coordination Architecture for Salience-Weighted Mode Dynamics
Patrick St-Hilaire — Independent Researcher, Montréal
A coordination-level architecture proposing that cognitive coordination can be formalized as a five-stage pipeline — Pressure, Parsing, Salience, Mode, and Action — in which precision-weighted signals govern transitions between metastable cognitive configurations. Grounded in predictive processing and the Free Energy Principle. Includes compact formal equation set, five publication-grade simulation figures, Spearman correlation statistics, and four quantitative falsifiable predictions.
License
CC BY 4.0
arXiv
Forthcoming
Research Areas
Predictive Processing Active Inference Free Energy Principle Cognitive Architecture Salience Network Mode Dynamics Precision Weighting Metastable States Cognitive Coordination Dynamical Systems Computational Neuroscience Computational Psychiatry Executive Function Cognitive Flexibility Attention Regulation Cognitive Mode Switching Salience Dysregulation Neurodivergence ADHD Autism Trauma Attractor Dynamics Prediction Error Bayesian Brain Karl Friston
Patrick St-Hilaire
Independent Researcher
Montréal, Québec

My research concerns the problem of cognitive coordination — how a mind decides what matters, configures itself accordingly, and moves between states in a world that never stops changing.

The Salience Engine is a formal framework at the coordination level of cognitive organization. It proposes that relevance-weighted signals govern transitions between metastable processing modes through a five-stage pipeline grounded in predictive processing and the Free Energy Principle. The framework generates empirically testable predictions and has been computationally instantiated with parameterized simulation results.

The work spans cognitive science, dynamical systems theory, clinical psychology, and artificial intelligence — not as metaphor, but as convergent formalization of a shared coordination problem.

The framework speaks directly to questions that appear across disciplines under different names: why attention dysregulation in ADHD reflects shallow attractor dynamics rather than deficit; why trauma produces persistent threat-mode occupancy through precision inflation; why cognitive flexibility depends on baseline regulation before it depends on effort; why executive function is best understood as meta-access to one's own salience dynamics rather than as top-down control; and why the same coordination architecture that governs biological cognition appears in attention mechanisms in artificial neural networks.

Correspondence regarding the research,
endorsement, or collaboration
is welcome.