This preprint proposes a unified framework in which affective tagging is treated as a geometric operator acting on a cognitive map. Within prevailing computational-neuroscience modelling conventions (e.g., reinforcement-learning, active-inference, predictive-coding and value-based frameworks), affective and bodily influences are typically introduced as explicit reward/utility, prior, cost or preference terms. We hypothesize that tag weights directly deform the internal metric of the map, thereby changing geodesic proximity between states and biasing trajectory selection, generalization, and error structure. The framework links tagging dynamics to bodily and homeostatic variables (including interoception and immunoception) and outlines how symbolic code can act as an operator over trajectories and the tag field. We provide a formal notation that supports testable predictions and potential mappings to neurophysiological and behavioral markers. We also discuss implications for neuro-inspired artificial-agent architectures, including continual learning, internal valuation, and long-term self-stabilization.
Translation note: this work was originally developed in Russian and translated into English by the author. If you notice issues in the translation that affect technical meaning or clarity, please let me know.Request for feedback: I welcome critical comments on (i) scope and category fit, (ii) formal notation and testable predictions, and (iii) neurophysiological plausibility and links to behavioral markers.