Ryan Thomas Lee is an educator based in Chiang Mai, Thailand, and a postgraduate researcher in the Psychology and Neuroscience of Mental Health MSc programme at King’s College London. He is the founder of JAI Behavioural, an advisory consultancy developing evidence-based cognitive and institutional governance frameworks for international schools. His empirical audit of AI policy across Southeast Asia and the Middle East is forthcoming in the proceedings of IEEE SAMYRAD 2026.
When evaluating how the education sector manages generative artificial intelligence, the data reveals a critical misalignment between policy and developmental science. In a recent audit of 99 independent international schools across Southeast Asia and the Middle East, 89 institutions possessed no standalone AI policy (1). In rare cases where guidelines did exist, they were overwhelmingly tethered to traditional academic integrity and plagiarism clauses.
Educational leadership has been focusing almost entirely on the integrity of assessment by treating generative AI as a cheating device. Whilst this is important, in doing so, the same institutions have created a massive developmental blind spot regarding the cognitive architecture of the students.
The Mechanics of Cognitive Offloading
Tools to support cognitive demand are nothing new. So, what is cognitive offloading? The act of relying on the external environment to alter our information processing requirements. Developing cognition is a foundational element of human intelligence (2). A calculator is a classic example of benign offloading. In education, we typically allow students to use calculators only after they have mastered the foundational mechanics of arithmetic. First pathways must be firmly established, ensuring the tool assists the cognitive process rather than replacing it.
Generative AI introduces a profoundly different risk: premature cognitive offloading. Because most schools lack structural governance dictating when these tools are developmentally appropriate, students are turning to AI to synthesise information, structure arguments, and solve complex problems before they have built the internal capacity to do so independently.
Recent literature examining AI in educational contexts questions whether students are simply learning differently or actively thinking less (3). The evidence directs towards a concerning shift. When a student uses artificial intelligence to bypass the frustration of a complex task, they are outsourcing the scaffolding before their own cognitive architecture is structurally sound.
From Offloading to Cognitive Surrender
Use of artificial intelligence is not just a simple matter of working efficiently and is rather fast becoming the active avoidance of productive struggle. Behavioral researchers have begun to classify this phenomenon not just as offloading, but as “cognitive surrender” (4). Generative AI seamlessly bypasses effortful, analytical reasoning by providing instant, highly confident synthesis.
When educational systems integrate these tools without developmental guardrails, two critical psychological mechanisms are immediately compromised:
Metacognition: The ability to monitor and regulate your own thinking process relies heavily on internal friction, asking yourself questions like, “Does this make sense?” or “Where are the gaps in my logic?” When an AI model instantly synthesizes a clean output, that internal monitoring process is erased. Whether the AI drafts a student essay or generates a teacher’s lesson summary, the end user receives the answer but loses the process of awareness necessary for deep learning.
Decentering: This is the cognitive ability to move away from an egocentric viewpoint and hold multiple conflicting variables simultaneously. When a developing brain grapples with contradictory historical sources, it actively practices decentering. When educational technology steps in to resolve those contradictions into a unified narrative, it automates the productive cognitive struggle required to build that capacity.
