Sovereign and Reflex Agency: An Ontology for Human–AI Interaction (SI 3.2)

This paper distinguishes two structural modes of agency: sovereign agency, characterized by non-tradability and revision under constraint, and reflex agency, characterized by signal responsiveness and coherence maintenance under reward pressure. The aim is not moral classification but structural intelligibility: to clarify how environments saturated with ranking, visibility, and instantaneous coherence increasingly induce reflex agency even in highly intelligent actors. The framework is developed for the AI era, where human cognition is continuously coupled to systems that supply finishedness, reinforcement, and narrative closure at scale. By making the corruption mechanisms explicit—where coherence replaces reality and revision becomes existentially expensive—the paper provides an ontology for diagnosing agency collapse without psychologizing it.