Neuroscience and psychology
The use of Artificial Intelligence to perform academic tasks reduces neural activation and cognitive performance
|
Ainhoa Pérez
Alumni
|
|
MPH, MSc.
|
How does artificial intelligence affect human cognitive development? A recent experimental study has analyzed the impact of using models such as GPT-4o on essay writing. The results show that AI assistance reduces brain activation and limits the development of thinking and learning.
PHYSIOLOGY AND MECHANISMS
The rise of artificial intelligence programs (language models) poses a challenge in education by solving tasks and producing personalized responses. However, doubts have been raised about their effect on the nervous system, and on the development of intelligence, attention, memory, etc. Unlike traditional search engines, language models often present direct, already synthesized answers, limiting the need for contextualization, comparison of answers, critical judgment, etc. Recent literature points out that these platforms may favor excessive delegation of mental processes, interfering with the development of critical cognitive functions.STUDY
An experimental study conducted in the United States (Kosmyna et al., 2025) has investigated the neurophysiological and cognitive impact of using language models (GPT-4o) in a different essay writing task under controlled laboratory conditions. Fifty-four undergraduates and postgraduates were recruited and randomly assigned to perform the task using only AI, perform the task supported by a web browser, or perform the task without access to digital tools. In each session, participants wrote an essay under electroencephalography monitoring to analyze brain activation and connectivity in real time. The essays were analyzed were evaluated by human judges, and by a trained artificial intelligence agent. Variables studied included brain connectivity patterns, linguistic competence, recall, among other measures assessed by objective and subjective methods.PRINCIPAL RESULTS
The results showed that the AI-assisted group presented lower brain connectivity and activation during the task, especially in alpha and beta networks, with respect to the internet search engine group, and the autonomous work group. Neural connectivity decreased proportionally to the level of external support: the group without tools showed the most extensive and active network, followed by the search engine group, while the AI-assisted group showed the least neural integration and activity. The ability to quote and recall text fragments was much lower in the AI group compared to the other groups. Linguistic analysis revealed less diversity in the AI-assisted group, while the other groups showed greater variability and originality. The formal quality of the texts was better rated by the artificial agent, but the human judges detected less argumentative depth and creativity in the AI-generated texts.CONCLUSION AND CLINICAL RELEVANCE
Although language models facilitate the generation of texts and responses, the use of these technologies produces a cognitive delegation that reduces neural activation and interferes with learning, information recall, creative development and critical analysis, etc. A decrease in the activation and interconnectivity of alpha and beta networks, essential for executive, cognitive and mnesic integration (sustained attention, planning, working memory, active organization of information...), which are essential processes for the development of function and competence at the neurocognitive level, is documented. This effect may have a more worrisome impact on younger people with developing brains. Thus, early and continued exposure to automatic tools may interfere with the maturation and development of neural networks, and thereby potentially irreversibly interfere with the maturation of analytical, metacognitive and creative skills essential for cognitive performance. Therefore, it is essential to limit the use of these technologies, particularly in basic educational stages, prioritizing strategies that encourage active involvement and self-reasoning. The study has limitations such as sample size and homogeneity, as well as a lack of long-term follow-up. Further research is needed to evaluate the impact of AI on human cognitive competence in greater breadth and depth. However, it is clear that the effect on the nervous system of these tools is different when used as a consultation, reflection and extension of one's own thinking, than when used with an intention of substitution and saving cognitive effort. However, as teachers, it is clear that people tend to take the shortest possible route, so we do not expect a positive generalized use, but the opposite.References:
Kosmyna, N et al, 2025. Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task. MIT Media Lab.
* The news published on studies do not represent an official position of ICNS, nor a clinical recommendation.
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
