E-Paper
AI makes learning faster, but is it making thinking harder?
This article is authored by Partha Chatterjee, Dean of Academics, Shiv Nadar University, Delhi-NCR.
When ChatGPT reached 100 million users within two months of its launch, faster than any application in history, educators did not need a research study to understand what was coming. The classroom had already changed. The question was only how much, and what to do about it.
For those of us in higher education, the first alarm sounded around writing. It was obvious, almost immediately, that students would reach for AI to generate essays, assignments, and theses rather than work through them and write them. This was not merely a concern about academic integrity. Writing is not a mere transcription of thought, it is the process for thought. The act of structuring an argument on paper, of choosing one word over another, of recognising midway through a paragraph that the logic does not hold — these are the cognitive processes through which learning actually happens. Outsource the writing, and you outsource the thinking that produces it.
We responded by redesigning curriculums and assessments to make AI-assisted shortcuts harder to conceal and genuine engagement harder to avoid. But before we could call that a success, a second problem announced itself: students had stopped reading. Not struggling with difficult texts — simply bypassing them. Why read a hundred pages when an AI can summarise the argument in three bullet points? That shortcut short-circuits another process of intellectual development. Reading builds connections across ideas in ways that a digest cannot replicate. An offhand line on a page lights a spark that no summary would have flagged. Those sparks are where genuine intellectual formation happens. We, in the education space, are working on this, though I cannot claim we have had much success yet.
But the larger danger - the one that deserves the most serious attention--is the third and deepest shift: Students are outsourcing their thinking itself. Research in this area is nascent, yet it is clear that frequent use of AI encourages increased cognitive offloading leading to reduced critical thinking. What makes the matter worse is that the reduced cognitive load leads to even more trust in what AI is producing, which in turn reduces critical thinking even further. A very recent study from Oregon State University put a name to the mechanism: The cognitive debt cycle. Routine use of AI tools progressively weakens students’ habits of reflection, questioning, and independent reasoning--which in turn drives greater dependence on AI, which weakens those habits further. Research also shows that the worst affected are the younger ones, those with less developed self-regulatory capacities, showing the highest dependency and the lowest scores. A 2025 MIT Media Lab study made the point viscerally: students who wrote essays using ChatGPT showed the lowest brain engagement across 32 neural regions, their writing becoming increasingly formulaic with each session. The lead researcher’s conclusion was stark--the task was being executed, but nothing was being integrated into the brain’s memory networks. This is not simply a pedagogical inconvenience. It is a question about the kind of minds we are cultivating.
And yet, this is not a lost cause. The same research that documents the risk also points to some remedy, when AI is used as a structured scaffold rather than a substitute, and when students bring developed self-regulated learning skills to the interaction, higher-order thinking is not eroded. It is enhanced. The distinction is critical: AI as a thinking aid versus AI as a thinking substitute. The first requires a student who already knows how to question, evaluate, and synthesise, and uses AI to extend that capacity. The second requires nothing of the student at all.
The implication for institutions is clear. We cannot simply ban AI use, that battle is already lost, and the technology is too genuinely useful to forfeit. Nor can we permit unrestricted use and hope students self-regulate. The answer lies in deliberate, graded integration.
Our approach at our university is to develop an institutional AI use policy that operationalises exactly this. Faculty may assign any of several levels of AI use to a course or assignment--from fully AI-free to AI-assisted - depending on the learning objective at stake. Each level carries a reporting requirement: students must document how AI was used and to what end. This is not paperwork. It is a forcing function for purposeful engagement, compelling students to ask why they are using AI, to achieve what, rather than simply reaching for it to complete a task.
The goal is not to keep AI out of education. It is to ensure that when students arrive at AI, they bring their minds with them and use AI to develop beautiful minds.
(The views expressed are personal)
This article is authored by Partha Chatterjee, Dean of Academics, Shiv Nadar University, Delhi-NCR.
