Tue 8 Sept 2026

A three-wave longitudinal study (N=589, Zhu et al., Frontiers in Psychology, July 16, 2026) distinguishes dependent cognitive offloading — delegating reasoning, synthesis, and evaluation to generative AI — from autonomous offloading, where AI scaffolds thinking without replacing agency. Both modes produced comparable immediate task benefits, but dependent offloading correlated with cognitive agency transfer and reduced intrinsic motivation, predicting declines in perceived creativity, deep processing, and independent judgment over time. Metacognitive monitoring attenuated agency transfer but did not buffer the motivational decline. The study's structurally significant finding: both modes felt equally productive in the moment, meaning users cannot distinguish adaptive from maladaptive AI engagement through introspection alone — detecting cognitive dependency requires an external longitudinal record of one's own usage patterns, the same class of verifiable evidence used for documents and legal proof.

Not submitted.