As large language models take on an increasing share of cognitive work, researchers are warning that this growing reliance on “mental outsourcing” may come with hidden costs.
Research scientist Nataliya Kosmyna has warned that if we become too reliant on AI, it could influence the language we use and even weaken our ability to perform basic cognitive tasks.
There is growing concern that as we delegate even more thinking to large language models and other AI systems, the impact on memory and problem-solving skills could deepen. AI tools can now produce persuasive poetry, offer financial guidance, and even provide a sense of companionship. At the same time, students are increasingly relying on these systems to complete their own work.
ChatGPT, Google and no-tech essay writing
Kosmyna and her colleagues at the MIT Media Lab recruited 54 students to write short essays, dividing them into three groups. One group was instructed to use ChatGPT, another was allowed to use Google Search with AI-generated summaries disabled, and a third completed the task without any technological assistance. The students’ brain activity was monitored throughout the exercise.
The essay topics were intentionally broad and open-ended, requiring minimal research. Prompts included questions about themes such as loyalty, happiness, and everyday life decisions.
Participants who relied only on their own thinking showed highly active brain patterns, with widespread engagement across multiple regions. The group using search engines also displayed strong activity in visual processing areas. In contrast, the ChatGPT group showed significantly lower overall brain activity, with reductions of up to 55%.
After submitting their essays, participants in the AI group were unable to quote from what they had written, and several said they felt little sense of ownership over their work.
ChatGPT essays described as “soulless”
Kosmyna also warns that outsourcing work to AI risks diminishing the creativity that drives original thinking. In her study, she says, the essays produced with ChatGPT were strikingly similar and were described by teachers who marked them as “soulless,” lacking originality and depth.
Kosmyna cautions that we should be mindful of cognitive shortcuts, something our brains are naturally drawn to. She argues that, for long-term brain health, it is important to keep challenging ourselves.
In doing so, she suggests, we can better support our minds, creativity, and overall cognitive well-being.