This paper introduces Regulatory Intelligence (RI), a paradigm for artificial intelligence that treats cognition as a regulated dynamical process rather than a purely performance-optimized predictor. The work argues that many failure modes in modern AI—hallucination, brittleness, and runaway inference—arise from the absence of internal regulatory mechanisms analogous to homeostasis in biological systems. RI reframes intelligence in terms of viability, stability, and self-regulation, proposing architectural constraints and control principles that prioritize internal coherence over raw task accuracy. The paper integrates ideas from cognitive architectures, control theory, and dynamical systems to outline how regulated cognition can mitigate instability under stress. This work is intended for researchers in artificial intelligence, cognitive systems, AI safety, and complex systems who are interested in alternatives to scale-driven optimization and in principled approaches to stable, governable intelligent systems.
John Cragin (Sun,) studied this question.