ChatGPT can reduce critical thinking skills

By Alimat Aliyeva

A recent study conducted by scientists from the Massachusetts Institute of Technology's Media Lab has uncovered some concerning results regarding the impact of ChatGPT on critical thinking skills.

In the study, 54 participants aged 18 to 39 from Boston were divided into three groups. Each group was asked to write several essays based on the SAT test, using OpenAI’s ChatGPT, Google Search, or simply doing the task without any tools, respectively. The researchers used an electroencephalograph (EEG) to record brain activity in 32 regions, and the results were telling. Of the three groups, ChatGPT users exhibited the lowest brain engagement and showed consistent lagging on neural, linguistic, and behavioral levels.

Over the course of several months, the group that used ChatGPT became increasingly disengaged with each essay. By the end of the study, many participants resorted to copying and pasting, with little effort to engage their own critical thinking. The study suggests that the use of large language models (LLMs) like ChatGPT may impair the brain’s ability to engage in deep cognitive processes, which is especially concerning for younger users still in the process of developing their cognitive and problem-solving abilities.

The study, still awaiting peer review, has garnered attention, as the sample size was small and focused primarily on ChatGPT’s impact. However, the primary author of the paper, Natalia Kosmina, believes it’s crucial to raise awareness about these potential downsides. She stresses that, as society becomes more reliant on AI tools for convenience, long-term brain development might be compromised.

The MIT Media Lab has recently allocated significant resources to investigate the broader effects of using generative AI tools. Earlier research found that spending time chatting with ChatGPT led to increased feelings of loneliness among users, adding further concerns about the mental and emotional impact of AI on users’ well-being.

Kosmina, a full-time researcher at MIT since 2021, aimed to study the effect of AI tools on schoolwork as the use of AI in education continues to rise. The subjects were tasked with writing 20-minute essays based on typical SAT assignments.

The group that used ChatGPT produced highly similar essays, with little to no originality. The essays were described as “soulless” by two English teachers who evaluated them. The electroencephalogram (EEG) data also showed low levels of executive control and attention engagement. By the third essay, many participants simply gave ChatGPT a hint, and the model completed most of the work for them. As Kosmina puts it, “It became more like, ‘just give me an essay, finalize this sentence, edit it, and I’ll finish.’”

On the other hand, the group that worked without any technological assistance showed the highest neural connectivity, especially in the alpha, theta, and delta brainwave ranges, which are associated with creativity, memory recall, and semantic processing. These participants were more engaged and curious, taking greater responsibility for their work and expressing higher satisfaction with their essays.

The third group, which used Google Search for research, also showed high levels of brain activity and satisfaction. However, the cognitive pattern was slightly different from that of the brain-only group. Many participants used AI-powered chatbots for information retrieval instead of traditional search engines, a trend that has become increasingly common.

After writing three essays, participants were asked to rewrite one of their previous essays. The ChatGPT group had to do this without the tool, while the brain-only group could now use ChatGPT for assistance.

The results were striking: the ChatGPT group struggled to recall details about their previous essays, showing weaker alpha and theta brainwaves, likely indicating a bypass of deeper memory processes. In contrast, the brain-only group showed significant improvements in brain connectivity and cognitive function when using ChatGPT to revise their essays, suggesting that AI, if used correctly, can actually enhance learning.

Despite the compelling results, the study has its limitations. Critics have pointed out the small sample size, the focus exclusively on ChatGPT, and the narrow scope of tasks tested. Additionally, there were concerns about the lack of clear functional separation between the tasks being analyzed. Further research with larger and more diverse sample sizes will be needed to validate these findings.

Kosmina’s research highlights an important question: how should we integrate AI into learning without sacrificing critical thinking and cognitive engagement? As AI tools become more integrated into everyday life, it is essential to consider their long-term impact on both learning and brain development. While AI offers undeniable benefits in terms of efficiency, it’s important to use it responsibly to ensure that we don’t lose the mental engagement that fosters deep learning and critical thinking.