A recent study out of MIT Media Lab shows that students using ChatGPT and other AI tools to write essays may be acquiring “cognitive debt” at a higher rate than students using searching engines or only their brains.
According to the study, “Cognitive debt defers mental effort in the short term but results in long-term costs, such as diminished critical inquiry, increased vulnerability to manipulation, decreased creativity.”
The study divided participants into three groups. One was allowed to use LLMs (large-language models) to write their essays, another was allowed to use search engines, and the other was only allowed to use previous knowledge. The study refers to the latter as the “Brain-only group.” Researchers asked each group to write three essays using their designated tool. For a fourth essay, LLM users were only allowed to use their brains, and brain-only writers were allowed to use LLMs. One AI judge and several human teachers scored the essays. The researchers measured the electrical activity of the participants’ brains during each stage of the study.
The study showed significantly weaker brain connectivity in LLM users than in the brain-only group. In the fourth essay, LLM users continued to struggle with brain connectivity and struggled to quote their own work, while the brain-only group exhibited better brain connectivity and memory recall. “LLM users consistently underperformed at neural, linguistic, and behavioral levels,” the study says.
The study is not yet peer-reviewed and has a small sample size of 54 students. The cognitive function required to write an essay may not be broadly applicable to all life tasks, and the familiarization effect may account for some of the study’s results. Nevertheless, they are worrisome.
Despite concerns, AI use in education continues to grow. Sixty percent of teachers integrate AI into their lesson plans, according to an informal poll by Education Week, and “English language arts or social studies teachers in middle and high schools are most likely to use AI for instruction.” President Trump recently established an Artificial Intelligence Education Taskforce to train students and teachers in how to use AI to improve classroom outcomes.
While data has shown that AI use in the classroom can positively affect learning outcomes through personalized learning, many have raised concerns about the negative effects the technology could have on students’ cognitive ability and learning environments. Public policy advocates have also raised ethical questions about how the technology poses other risks to children’s well-being.
Jacqueline Annis-Levings is a correspondent for the Federalist. She is a rising junior at Patrick Henry College, where she is majoring in English.
