Photo by Growtika on Unsplash
A student opens ChatGPT for an English essay to “just get started.” Within seconds, the blank page is filled. It’s faster writing, less stress, and polished results. AI feels like a miracle, but what is convenience costing us cognitively? Does this introduce cognitive debt where short-term ease creates long-term intellectual consequences?
A study conducted by Cornell University, entitled “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task,” provides a sobering answer. In this experiment, brain connectivity levels were measured through electroencephalography (EEG) during essay writing tasks to determine the level of critical thinking required with and without AI. There were three measured writing conditions: Brain-only, Search Engine, and large language models (LLMs) like ChatGPT, which determined the tools participants could use to create essays. Some participants engaged in a Session 4 where Brain-only users switched to LLM users, and LLM users switched to the Brain-only condition. Interestingly, these essays were evaluated by humans and AI.
Participants who wrote without external tools exhibited the strongest and most distributed neural networks. In simpler terms, their brains were more active across regions responsible for memory, reasoning, and synthesis, which are hallmarks of critical thinking. Search engines showed moderate engagement, likely because searching still requires evaluation and decision-making. LLM users, however, displayed the weakest brain connectivity. The AI was doing all the heavy lifting, and the brain responded by doing less.
The system was acting as their brain for them, creating overreliance. This overdependence risks losing critical thinking skills over time, without proper regulation. Since thinking is a skill that strengthens over time, when AI generates ideas, structures arguments, and fills in language, the brain loses opportunities to practice these cognitive processes. Over time, this creates what the researchers describe as cognitive debt. When these skills aren’t exercised, they don’t just stagnate; they erode.
In Session 4, LLM-to-Brain users struggled with the essay-writing task, as they showed lower alpha and beta connectivity, signals associated with attention and active engagement. Essentially, the brain struggled to re-engage. In contrast, Brain-to-LLM users showed increased activation and memory recall. AI can reduce the mental load of relying solely on brain power, which may feel helpful at first, but over time, reliance makes independent thinking harder. AI is easiest to use when you don’t need it, and most damaging when you do.
Beyond brain activity, the study uncovered a more subtle but equally concerning effect: a loss of ownership. LLM participants reported feeling less connected to their essays and struggled to accurately quote their own work. Brain-only writers felt the strongest sense of ownership of their own work. Writing, it turns out, is not just about producing words, but about how we process, internalize, and understand ideas. When that process is outsourced, learning becomes shallow.
Over four months, these effects compounded, with LLM users consistently underperforming at neural, linguistic, and behavioral levels compared to their peers. These findings raise serious concerns about what happens when AI becomes a default learning tool rather than a support. If students no longer struggle through ideas, wrestle with phrasing, or sit with uncertainty, they may graduate having produced more work but understood far less.
None of this means AI should be eliminated from education. Like powerful economic or political systems, its danger lies not in its existence, but in its unchecked use. AI can be transformative when regulated: refining drafts, offering feedback, or supporting research after foundational thinking has occurred. But when it replaces thinking entirely, it undermines the very skills education is meant to build.
The discomfort of thinking, of not knowing what to say next, is not a flaw in learning. That is the point. If we prioritize speed over struggle and output over understanding, we risk raising a generation fluent in prompts but unfamiliar with thought. Convenience may save time today, but the cognitive debt it creates could cost us our ability to think tomorrow.
Be First to Comment