Cognitive Offloading and Automation Bias in Generative AI-Assisted Learning Systems: A Systematic Review and Secondary Evidence Synthesis
The use of Generative Artificial Intelligence (AI) tools in learning systems is growing rapidly and is changing the way people learn, solve problems and make decisions. Cognitive offloading refers to the transfer of mental tasks such as remembering, reasoning and evaluating to external systems, while automation bias describes the tendency to rely on the output of automated systems in place of independent judgement. Both phenomena are more visible in generative AI settings, because these systems produce fluent, human-like responses that imitate the process of reasoning itself. Over-reliance on generative AI can therefore threaten intellectual independence and accountability for decisions: it tends to reduce verification behaviour, weaken skill development and increase the risk of reasoning errors, particularly in educational and workplace learning. This paper examines the role of cognitive offloading and automation bias in generative AI-supported learning systems through a systematic review and secondary evidence synthesis of recent empirical studies from educational, technical and professional fields. The results show that unstructured use of generative AI is frequently associated with lower cognitive engagement and greater reliance on automation, and that these effects are reduced by structured and guided interaction models. The paper argues for a human-centred strategy for integrating AI into learning and outlines directions for future research on long-term cognitive outcomes and responsible AI-based learning design.
Authors
- Deepak Kumar (ORCID: https://orcid.org/0000-0001-5280-843X)
- Yadav Rana (ORCID: https://orcid.org/0009-0000-9645-2500)
Institutions
- GNA University (IN)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24409038
- Primary Topic
- Artificial Intelligence in Education
- Type
- preprint