Does use chatgpt to do a school work has a negative effect on the ability to think for yourself? This is what a study by researchers at the Massachusetts Institute of Technology (MIT) seems to confirm. According to their research, not only people who use Chatgpt puts their brain less, but in addition, they are becoming more and more lazy during the experience.
The 54 participants in the study, residents of the Boston region aged 18 to 39, were divided into three groups. Each group was to write several writing works similar to those requested during standardized tests for admission to American universities.
To write, one group used Chatgpt, another was entitled to Google’s search engine and the last was not entitled to any tool.
The researchers recorded neural activity in the different regions of the participants of the participants during this time thanks to an electroencephalogram (EEG). The results were quite homogeneous inside a given group.
The group that was to use Chatgpt presented the lowest brain activity.
As the study progresses, the members of the group using Chatgpt have increasingly started to simply copy and paste the responses of the large language model rather than being inspired. At the end of the exercise, they were the least capable of quoting their own work.
The group which was to trust only its brain presented the greatest neuronal connectivity in the regions associated with creative ideation, memory and semantic treatment. The researchers noticed that this group was more curious, motivated and presented greater satisfaction with the content of their dissertation.
As for the group that used Google, it also had a high brain activity and good satisfaction with their work. The researchers explain it by the fact that they had to seek information themselves thanks to their own keywords.
The fact of systematically less use your brain, and therefore make less new connections between its neurons, could have long -term effects on the development of the brain of the youngest.
• Read also: Can AI replace a shrink?
Critical loss
If the work of the groups using Google or just their brain diverge between them, those made with the help of Chatgpt were extremely similar, even if of good quality according to the teachers responsible for evaluating them. The latter described them as “soulless”.
Participants of the Chatgpt group were less likely to assess the response given by the great model of language in a critical way, which is worrying when we know that Chatgpt responses reproduce the biases and stereotypes introduced by the data with which it is taken.
Researchers point out that tool responses like Chatgpt can also be very easily influenced by the interests of companies who own them and their shareholders, which makes the risk of “echo chamber” even more worrying.
• Read also: “Starter Pack”: this is how much water we “waste” when we generate an image with AI
Publish
The results of this study are to be taken with a grain of salt, however, as they were published before having gone through a peer review. We can also notice that the size of the sample is quite reduced and that the participants all came from the same geographical region.
It is a conscious choice on the part of researchers who wanted to make their results known without delay to highlight the potential problems linked to an increasing use of large language models as a chatgpt.
“What really motivated me to take out the study now before waiting for the whole review by peers is that I am afraid that in six to eight months, there is a politician who decides to launch something like a” GPT maternal “. I think it would be completely bad and harmful, ”explains the main author of research, Nataliya Kosmyna.
She believes that if we want to one day integrate this kind of tools in education, we must better test the effect of these on developing brains in long-term studies and adopt laws to supervise them accordingly.
– with MIT and time information
• Read also: Artificial intelligence to better detect emergency fractures
