The unseen cost of AI: How LLM dependence is hollowing out the next generation's minds

Khaleej Times

10-03-2025

While claims of an impending robotic uprising by super-intelligent AIs are just a bit overblown, there is no doubt that Large Language Models have already quietly taken over classrooms and college admissions offices. Microsoft recently partnered with Carnegie Mellon researchers in a large scale study revealing that AI dependency causes significant cognitive reduction in user performance. ChatGPT has redefined student writing capabilities while also reshaping their cognitive processes at an essential time when they need to develop skills that thrive on challenge. The study revealed that workers who delegated their routine tasks to AI systems gradually gained doubts about their problem-solving skills, beginning a spiral of self-doubt.
As an educational consultant at Hale Education Group, I've observed this process occur first-hand. Previously, my students' essays displayed the raw, untidy thinking of teenagers through awkward metaphors and half-baked epiphanies and occasional grammatical errors. The modern essays, meanwhile, resemble corporate mission statements. When asked to revise an essay without AI, many freeze. 'I don't know where to begin,' has gone from a rare utterance to a constant crutch. Students turn to tidy, AI-generated plotlines that often have very little to do with their actual lived experiences, as AI cannot possibly have the necessary context to flesh out one's life.
The tool's illusion of competence is seductive: Today's students find syntax wrestling distasteful as bots deliver generic passages like 'My journey taught me resilience, adaptability, and the transformative power of education' in mere seconds. But ease has a cost. Admissions teams at several top universities have noticed an increase in applications with perfectly written essays that lack distinctive personality and charm, and are fighting back both algorithmically and intuitively.
Educational institutions Duke and UC Berkeley use Turnitin and GPTZero detectors to screen submissions and impose expulsion penalties for excessive AI assistance use. Another admissions officer from UCLA recently turned to Reddit to point out all of the clear 'tells' of AI writing: words like 'delve' and 'tapestry' as well as overused grammatical elements like em-dashes and tricolons are obvious signals that admissions officers can recognize even without specific detection software.
More broadly, AI dependency hits at a critical developmental window. During adolescence, the brain eliminates unused neural connections but fortifies pathways that learned experience has established. Writing college-level essays requires students to demonstrate narrative structure and analytical thinking and provide evidence to support their argument, and ChatGPT omits providing students the learning experience of that process. Students who create drafts through AI tools miss opportunities to build 'intellectual muscle memory' according to Carnegie Mellon research by doing all their thinking for them. The consequences follow them to campus. First-year writing instructors at numerous universities report that their freshmen students struggle when they must adapt to changing seminar discussions because these students think in rigid patterns like chatbot outputs.
The real consequences at play here extend beyond college admissions because they represent how we sacrifice our internal development when we rely on external means. Writing holds a key function beyond conveying messages, serving as a tool for both identity creation and cognitive development. Every student who allows ChatGPT to express their 'passion for environmental justice' or 'lessons from failure' misses an opportunity to have a genuine encounter with their values and identity. The Microsoft study hints at this: Those workers who doubted AI came up with more varied, creative answers. Similarly, the admissions essays that move admissions officers come from genuine voices.
While artificial Intelligence will not likely bring about some cataclysmic event in the next century, it could create a false sense of success, making people believe that efficiency means assembling generic concepts. Modern technology's greatest risk doesn't lie in machine intelligence exceeding human capacity, but rather, that we unintentionally stunt our own development by over-adoption of artificial models. So to students: Keep the chatbot tab closed. Express your thoughts first (poorly, if need be) before revising them. Find your own words by struggling with them, as genuine expression requires time, and most importantly, ownership.
The writer is an education consultant at Hale Education Group.