Cognitive Offloading and AI Literacy in Higher Education: A Dual-Process Model of Generative AI Dependency and Student Meta cognitive Agency

Abstract

With the recent rise of Generative Artificial Intelligence (Gen-AI) in higher education, the nature of engagement, processing, and cognitive effort is changing. While Gen-AI tools offer opportunities for tailored scaffolding and rapid information synthesis, concerns about overreliance and atrophy of natural process are growing. Building on Dual-Process and Cognitive Load Theories, this paper proposes and empirically evaluates a conceptual framework investigating how AI Literacy and Perceived Task Complexity influence Cognitive Offloading Behaviors, and how these behaviors in turn impact Meta cognitive Agency and Academic Self-Efficacy. Using a quantitative SEM approach based on primary survey data (N = 684 across multi-disciplinary undergraduate programs), this study evaluates six hypotheses. The paper finds that high AI Literacy leads to reduced uncritical, passive dependency and more strategic cognitive offloading, while high Perceived Task Complexity increases dependency when AI Literacy is low, resulting in reduced meta cognitive agency and self-efficacy. Modern analysis further finds a moderating role of institutional AI policy clarity. This paper's theoretical contributions refine cognitive theories in the context of AI-assisted learning, while practical implications give guidance to higher education leaders, curriculum designers, and faculty in developing robust, AI-integrated assessment frameworks. Keywords: Generative Artificial Intelligence, Cognitive Offloading, AI Literacy, Meta cognitive Agency, Dual-Process Theory, Higher Education, Academic Self-Efficacy.

https://doi.org/10.5281/zenodo.22873728