If you have ever realized that you remember less than a time, or if you found yourself thinking “This text does not really seem to me mine” after using chatgpt, you are not paranoid: science has just started confirming what many suspected: using large linguistic models (LLM) as a chatgpt can weaken memory, flattening the originality of the texts and reduce neural activity in the brain areas involved in creativity and in deep cognitive elaboration.
The study, entitled Your Brain on chatgptwas conducted on a group of 54 participants to investigate how the use of artificial intelligence influences the writing process and, consequently, the functioning of the brain. The results are clear: the more delegated to the AI, the less the brain is activated.
The participants were divided into three groups: one wrote essays using chatgpt, one used a search engine and the third wrote without any technological support (the “Brain-Only” group). Each group faced three writing sessions, followed by a fourth experimental session for some of the participants.
Through electroencephalogram (EEG), the researchers observed that the Brain-Only group presented greater neural activation during the writing activity. In particular, areas related to memory, semantic integration, creative thinking and automonitoring were stimulated.
The group that used search engines showed a marked activation of visual and occipital areas, probably due to the reading and evaluation of the content found online. However, in the chatgpt group this visual stimulation did not occur, although they too were in front of a screen. According to the researchers, this shows that the interaction with AI does not activate the same type of critical and visual processing.
The use of chatgpt compromises memory and reduces the sense of belonging to their texts
The behavioral results confirm what is detected at the neurological level. The Brain-Only group was able to precision its texts with precision, reaching 100% accuracy already in the third session. On the contrary, in the chatgpt group, 83% of the participants had difficulty remembering what has already been written after the first session, a percentage that descended to 33% to the third.
Another important figure concerns the sense of Ownership, or the perception of having independently created the content. The Brain-Only group declared a strong sense of belonging to the texts produced, while in the chatgpt group the feeling was often uncertain: only a minority felt fully author of her essays, while between 50% and 90% declared a sense of only partial belonging.
Then there is the theme of the variety in content. The texts generated with chatgpt were statistically homogeneous within each topic, with very little diversification among the participants. In practice, many passively followed the traces suggested by AI, without deviating from the proposed path. This raises doubts about the idea that AI can serve simply as a basis to stimulate personal creativity. In reality, most users merely accept the responses provided, without making original contributions.
A possible solution
But not everything is negative. In the fourth session of the study, the participants of the Brain-Only group had the opportunity to use chatgpt. The result? A significant increase in brain connectivity on all EEG frequencies. A sign that, after an autonomous initial effort, the use of AI can further stimulate neural activity.
On the contrary, the participants who had already used Chatgpt from the beginning have not shown any neurological improvement. This brings researchers to a very precise conclusion: the introduction of AI must be carefully calibrated. Only after activating your cognitive resources is it possible to integrate tools such as chatgpt without losing control over your mental processes.
In other words, using chatgpt in moderation can have positive effects, but it must take place after an active and personal involvement. If, on the other hand, it becomes a shortcut to avoid the effort of thought, it risks compromising memory, creativity and critical sense.
Don’t you want to lose our news?
You may also be interested in:
