MIT research reveals: excessive reliance on AI writing may weaken human cognitive abilities

Recently, a 206-page research report conducted by the Massachusetts Institute of Technology has attracted widespread attention. This study explores in depth the impact of using large language models (LLM) in educational environments on cognitive abilities, particularly in the area of essay writing. The results indicate that excessive reliance on artificial intelligence Bots may have a negative impact on people's cognitive abilities.

The research team designed a 4-month experiment, dividing participants into three groups: LLM group, search engine group, and brain-only group. Participants were required to complete writing tasks on different topics within a limited time. To comprehensively assess the participants' performance, the researchers employed various methods, including electroencephalogram (EEG) recording, natural language processing (NLP) analysis, and evaluations by both humans and AI.

The experimental results show that participants relying solely on the brain exhibit a greater diversity in writing style, while the articles produced by the LLM group show a clear trend of homogeneity. In terms of the use of specific named entity (NER), the LLM group had the highest frequency of use, while the brain-only group had the lowest.

It is worth noting that participants in the LLM group and the search engine group, due to time constraints, tend to rely more on the output results of the tools and incorporate less of their personal opinions and experiences. This behavior may lead to a reduction in creative thinking.

In terms of neural connectivity patterns, researchers measured participants' cognitive load using the dynamic directed transfer function method (dDTF). The results showed a systematic decline in brain connectivity as external support increased. The brain-only group exhibited the strongest and most extensive neural network activity, while the overall neural coupling of the LLM-assisted group was the weakest.

The study also found that the LLM group had a lower sense of ownership over their articles and performed poorly in recalling and citing the content of articles just completed. Over 83% of LLM users were unable to accurately cite articles written just a few minutes earlier.

Although this study has not yet undergone peer review, its results indicate that participants in the LLM group performed worse than the control group, which relied solely on their brains, in terms of neural, linguistic, and scoring metrics over the 4-month study period. The researchers noted that the widespread use of LLMs may impact the enhancement of learning skills, particularly for younger users.

The research team emphasizes that more long-term studies are needed to understand the long-term effects of artificial intelligence chat Bots on the human brain before confirming that LLM is truly beneficial to humans.

In this study, artificial intelligence itself also expressed its opinion, believing that the research does not deny the value of LLM, but rather warns people not to overly rely on such tools while ignoring the importance of independent thinking and effort.

This research provides us with valuable insights, reminding us that while embracing new technologies, we must also pay attention to cultivating independent thinking and creativity. As artificial intelligence technology continues to develop, how to balance technological assistance with the enhancement of cognitive abilities will become an important issue faced by the field of education.