Cognitive Atrophy versus Epistemic Augmentation: The Bifurcation of Human Agency and the Structural Imperative of Human-in-Command in Collaborative Intelligence

【Abstract】This strategic systems and epistemological monograph addresses the defining civilizational question of whether artificial intelligence induces cognitive atrophy or enables unprecedented intellectual augmentation. Rejecting simplistic technological determinism, the paper establishes that cognitive outcomes are governed by a fundamental bifurcation of human agency. Under the "Dependent Mode," users engage in uncritical cognitive offloading—surrendering dialectical inquiry, logical verification, and normative judgment to automated black boxes—resulting in severe deskilling, automation bias, and the atrophy of critical faculties. Conversely, under the "Sovereign Mode" (Human-in-Command), artificial intelligence is operationalized not as an authoritative oracle, but as an externalized dialectical catalyst within Clark and Chalmers' "Extended Mind" framework. In this collaborative intelligence topology, the human retains sovereign intentionality, posing structural hypotheses, testing boundary conditions, and evaluating geopolitical and historical coherence, while the machine executes ultra-high-speed combinatorial search and simulation. The monograph proves that genuine intellectual sovereignty requires maintaining rigorous physical-layer epistemic grounding and ethical accountability, transforming generative models from instruments of passive cognitive erosion into powerful scaffolds for expanding human insight.