Does AI Make Students Dumber, or Is Our Definition of Intelligence Obsolete?
 |
| Source: Wikipedia |
The conversation around AI and education often begins with a familiar accusation: students who use tools like ChatGPT are somehow cheating themselves, producing shallow work, or “learning less.” A recent MIT study has fueled the narrative, suggesting that AI usage diminishes cognitive engagement and recall. On the surface, the conclusion appears straightforward. But beneath the data lies a deeper issue: the debate itself is anchored in an outdated metaphysics of presence.
The underlying assumption of this discourse is that intelligence is something wholly present, a possession of the mind, measurable once the tools are removed. In other words, cognition is treated as a book: bounded, stable, and self-contained. AI disrupts this illusion. It does not make students less intelligent. It exposes the obsolescence of a paradigm that still evaluates thought by what can be internalized and recalled. The old measures were never neutral; they were logocentric, privileging presence over relationality, origin over trace, the mind over the network. In this sense, the end of the book is unmistakable, and the beginning of a new writing — a distributed, trace-dependent intelligence — has already arrived.
From Book Economy to Writing
Schools were conceived to cultivate a certain kind of mastery, appropriate for scarcity. Books were limited, information traveled slowly, and intelligence was defined by what could be memorized and reproduced independently. Tools were merely auxiliary; cognition was assumed to reside entirely in the individual. This “book economy” valued presence and retention. The student was a self-contained author, producing work that proved possession of knowledge. Errors were failures of the mind itself.
AI overturns this arrangement. It operates in a field of abundance, where drafting, summarizing, and reframing are tasks performed instantaneously. The question is no longer whether the student “knows” in isolation, but whether they can navigate, evaluate, and synthesize. Intelligence is no longer internalized; it is relational. It emerges in the interplay between human judgment and machine-generated trace. Where the book demanded an originary subject, writing, in the Derridean sense, is always already a system without a fixed center. Mastery is no longer the ability to reproduce, but the capacity to engage critically with an unfolding network of signs.
What the MIT Study Reveals…and Hides
The MIT study places participants into three groups: writing unaided, writing with search engines, or writing with a large language model. Participants using AI performed worse on recall tests, leading to the headline conclusion that AI diminishes learning. Yet the study presupposes that learning exists in isolation, and that tools are distractions rather than constitutive elements of cognition. By removing AI and measuring what remains, it punishes delegation, precisely the moment where distributed intelligence is enacted.
Participants were not trained to collaborate with AI, nor were they fluent in prompting or evaluation. The study, therefore, measures the failure of first-time outsourcing under conditions designed for solitary authorship. Its findings are unsurprising, they confirm the logic of the book-based paradigm, not the realities of AI-mediated thought. What it does not measure is the development of judgment, discernment, and fluency, capacities essential for navigating a world where intelligence is enacted across humans and systems alike.
Chomsky and the Residue of the Book
Noam Chomsky’s association with MIT naturally colors public perception of AI debates. His critiques emphasize plagiarism and internalized knowledge, framing AI as a potential threat to learning. Yet his arguments are grounded in a theoretical vision that treats cognition as a bounded, individual computational system. Within this frame, tools may assist performance but cannot constitute thinking. Meaning resides in the mind; the subject authorizes the text.
The MIT study, though not directly influenced by Chomsky, operates under the same logic. Intelligence is treated as something present, measurable, and detached from its material and technological environment. AI exposes the limits of this paradigm: the authority of the “originary” mind is gone, the subject is no longer central, and what counts as intelligence must be reimagined.
From Ethics to Ontology
Much of the discourse on AI centers on ethics: fairness, authorship, and integrity. These debates are important but misdiagnose the problem. Moral panic substitutes for ontological insight, defending a model of thought that is already dead. Treating AI as a threat to virtue preserves familiar assessment practices, but it does so by insulating students from the very environment they will later inhabit. The ethical debate masks an epistemic one: what does it mean to know, to reason, to produce in a world where thought is distributed, trace-dependent, and emergent?
A more productive approach distinguishes AI literacy from AI fluency. Literacy is understanding how systems operate, their limits, and their risks. Fluency is the capacity to engage with them critically and creatively, to generate judgment through iterative interaction. Fluency is writing, not the book: it recognizes that authority and intelligence are not fixed in the mind, but emerge relationally, in the space between human and system.
The End of the Book and the Beginning of Thinking
AI does not diminish intelligence. It unmasks the obsolescence of metrics inherited from a book-based economy, where scarcity justified internalization and authorship guaranteed mastery. The danger lies not in reliance on AI, but in continuing to measure intelligence as if presence were possible. Drafting an essay without assistance once demonstrated mastery because it tested skills that were rare. Today, mastery lies in framing problems, interrogating outputs, and assuming responsibility for action across systems of trace.
The book is over. Its metrics, its values, and its subjects are no longer sufficient. The beginning of writing calls for education to embrace intelligence as relational, emergent, and interactive. Only then can learning remain relevant in a world where the marks left on a page, whether human or machine, are traces of something larger than any individual mind.
Chomsky, N. (1965). Aspects of the theory of syntax. MIT Press.
Clark, A., & Chalmers, D. (1998). The extended mind. Analysis, 58(1), 7–19. https://doi.org/10.1093/analys/58.1.7
Hutchins, E. (1995). Cognition in the wild. MIT Press.
Kosmyna, N., Hauptmann, E., Yuan, Y. T., Situ, J., Liao, X.-H., Beresnitzky, A. V., Braunstein, I., & Maes, P. (2025). Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing tasks (arXiv:2506.08872) [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2506.08872
Comments
Post a Comment