The Recursive Differentiation Hypothesis: Cognitive Externalization, Functional Unbundling, Reintegration, and Informational Inheritance

Abstract

The Recursive Differentiation Hypothesis (RDH) proposes that capacities bundled in biological organisms—access control, intelligence, self-modeling, and agency—can be functionally differentiated by cumulative culture and artificial intelligence, then recombined in new architectures. Differentiation is recursive because externalized cognitive products become inherited structure for later systems that search, learn, and externalize again. RDH's empirical core is functional rather than phenomenal. It predicts that novel tasks with unreliable or costly verification recruit globally accessible attention, working memory, and flexible control beyond what task difficulty alone explains, whereas reliable verification permits automatization. It further distinguishes decoder cost, which constrains recovery of inherited information, from verifier cost, which constrains evaluation of new candidates. Informational inheritance is target-relative: inherited structure may reduce search costs along some trajectories while canalizing or foreclosing others. RDH therefore organizes two competing long-run hypotheses: Progressive Decoupling, under which advanced cognition becomes less dependent on consciousness-related and self-modeling machinery, and Self-Model Convergence, under which autonomy makes recursive self-modeling increasingly necessary. The framework does not infer phenomenal consciousness from competence, attention, or self-report and requires no cosmological, simulative, or reincarnative assumptions.