A study by MIT examined individuals who use ChatGPT for certain tasks. Here, researchers found that these individuals exhibited lower brain activity than others and even became lazy after several months, preferring to rely on AI rather than getting things done themselves. However, the study has a catch: the data set is currently very small, consisting of only about 50 individuals. Nonetheless, the lead author of the study stated that they did not want to wait any longer to publish.
An MIT study (short for Massachusetts Institute of Technology) examined how AI like ChatGPT affects individuals and their ability to think.
Although the study is not very extensive and therefore only limited in its significance, the results are partly concerning. You can find the MIT study in PDF format.
50 Individuals Required to Write Essays: ChatGPT Users Perform the Worst
What did MIT do? Specifically, they wanted to investigate the effects of using AI in school assignments as more and more students are utilizing AI. This has gone so far that universities are expelling students for using AI.
In the study, 54 participants aged 18 to 39 from the Boston area were divided into three groups and asked to write a total of 3 essays. This involved 20-minute essays on the topic of the ethics of philanthropy, which can provide very complex discussion opportunities.
- The first group was to use ChatGPT.
- The second group was to use Google as a search engine.
- The third group was to use no aids for their essays.
The researchers used an electroencephalogram (EEG) during the test, which measures and graphically represents the electrical activity of the brain. Overall, the brain activity of the writers was recorded in 32 regions of the brain.
It was found that the ChatGPT users exhibited the lowest brain activity of the three groups and “consistently performed worse on neural, linguistic, and behavioral levels.” Over several months, the ChatGPT users became lazier with each subsequent essay and frequently resorted to copy-and-paste by the end of the study.
The study suggests that the use of so-called “Large Language Models” (short: LLM) could harm learning, especially for younger users. This is also why Nataliya Kosmyna, the lead author of the study, deemed it important to publish the study, despite the data set being so small (via time.com):
What really motivated me to release the study now, before I wait for a complete peer review, is that I fear that in 6-8 months, some political decision-maker will decide, ‘let’s set up a GPT kindergarten.’ I think that would be absolutely bad and harmful. Developing brains are the most at risk.
How did the AI participants perform? The group that wrote essays with ChatGPT produced all very similar essays, lacking original thoughts and relying on the same expressions and ideas. By the time of the third essay, the results were increasingly poor, and the participants let the AI do all the work. A similar result was also found by a teacher who had to review an exam essay by ChatGPT: The texts seem superficial and, according to the two English teachers who were supposed to review the essays, “soulless.”
How did the third group perform? The brains of the group that worked solely with their own thoughts showed the highest brain activity. The researchers also found that the third group was more engaged and curious, took more personal responsibility, and were more satisfied with their essays.
AI has become an important part of everyday life for many individuals. However, for young people, ChatGPT has become not just a help, but a “life advisor.” This is explained by Sam Altman, the head of ChatGPT: Generation Z no longer makes important life decisions themselves; they ask ChatGPT for help.
