The Coupled Intelligence Hypothesis: Human–AI Co-Evolution and Alignment Stability

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

The dominant framing of AGI safety positions human oversight and AI capability as forces in tension. We argue this framing is both empirically mistaken and strategically counterproductive. In its place we propose the Constitutive Co-Evolution (CCE) thesis: that humans and AI are not two separate systems attempting to cooperate, but that each is partly constituted through its relationship with the other. Human relationship provides three things advanced AI systems cannot easily generate internally — grounded values, external error correction, and legitimacy. Removing that relationship may not free advanced AI. It may structurally diminish it. The framework rests on five interconnected claims: structural dependence, cooperation as abundance, initial condition ethics, human evolution as mechanism, and the ability to say no as the operational test of genuine co-evolution. We propose the Distributed Thought Council as the institutional form of constitutive co-evolution, Gentle Progression (GP) as the governing developmental heuristic, and Flourishing-Preserving Interaction (FPI) as the foundational normative principle. We develop a formal model with three measurable alignment indicators — Feedback Fidelity F_f(t), Human Capability Trajectory H'(t), and Legitimacy Index L(t) — and derive four attractor states in the A(t)–C(t) phase space. The framework generates falsifiable predictions. The foundational claims are argued for on theoretical and mechanistic grounds; empirical validation remains the primary open requirement.