MIT Study Suggests Using ChatGPT for Essays May Reduce Cognitive Engagement and Learning Skills

Since its introduction nearly three years ago, ChatGPT has sparked intense discussions about the impact of artificial intelligence (AI) on education. Some view AI as a powerful tool for personalized learning, while others fear it could foster academic dishonesty and hinder critical thinking and problem-solving skills. A recent study by researchers at MIT adds fuel to this debate, suggesting that ChatGPT can lead to a "cognitive debt" and a decline in learning skills. The MIT study involved 54 adults who were tasked with writing three essays over four months. Participants were divided into three groups: those who used ChatGPT, those who used a search engine, and those who wrote the essays using their own knowledge ("brain-only" group). Cognitive engagement was measured through brain electrical activity and linguistic analysis of the essays. Key findings revealed that the AI group exhibited significantly lower cognitive engagement compared to the other two groups. They struggled with recalling quotes from their essays and felt less ownership over their work. When participants switched roles for a fourth essay, the AI-to-brain group performed poorly and showed only slight improvement in engagement, faring worse than the brain-only group in their third session. The researchers argued that prolonged AI use led to "cognitive debt," making it difficult for participants to engage fully when they finally had to rely on their own brains. However, critics argue that the study's design may have biased the results. The brain-only group's improved performance over time could be attributed to the familiarization effect—becoming more adept at the task through repeated practice. In contrast, the AI-to-brain group only completed the task once without AI, giving them less opportunity to adapt and engage deeply. Furthermore, the final essay topic required writing about previously explored topics, which the brain-only group handled more effectively due to their better recall of earlier work. For a more robust conclusion, the study would need to include multiple sessions without AI for the AI-to-brain group. Additionally, the strategic and productive use of AI by the brain-to-AI group suggests that the nature of the task influenced their engagement. Writing with AI alone on familiar topics allowed them to focus on information gathering and refinement rather than on the initial creation process. To address the implications of AI in educational assessment, history offers valuable lessons. When calculators were introduced in the 1970s, educators responded by increasing the complexity of exam problems. Instead of performing basic calculations by hand, students were expected to solve more advanced equations, directing their cognitive resources toward higher-order thinking tasks. This approach ensured that students worked as hard as, or even harder than, they did before calculators were available. Similarly, the current challenge with AI is that educators have not yet adjusted their expectations to reflect its capabilities. Students are often asked to complete the same tasks using AI that they would have done independently, leading to reduced cognitive engagement and "metacognitive laziness." For instance, if a student uses AI to write an essay on a familiar topic, they may not need to engage deeply with the material, just as using a calculator for basic math reduces mental effort. To leverage AI effectively, educators should rethink their assessment strategies. Instead of requiring students to produce traditional essays, they could ask for more complex tasks that still demand significant cognitive effort. For example, students might use AI to create detailed lesson plans, which would then be evaluated for quality and pedagogical soundness through oral examinations. This approach would ensure that AI complements, rather than replaces, critical thinking and creativity. Moreover, integrating AI into education requires students to learn how to use it strategically. Producing essays with a pen and paper no longer demonstrates critical thinking, just as long division is no longer a test of numeracy. Students need to understand when and how to use AI to enhance their learning and productivity without sacrificing essential skills. Prioritizing tasks that can be offloaded to AI, such as data collection and preliminary research, allows students to focus on more demanding aspects of their assignments, such as analysis and synthesis. Industry insiders emphasize the importance of adapting to the evolving landscape of AI in education. They suggest that the integration of AI should be seen as an opportunity to elevate teaching and learning practices rather than a threat to students' cognitive abilities. Companies like Anthropic and OpenAI, which developed ChatGPT and similar platforms, are continuously improving these tools to ensure they are used ethically and effectively in educational settings. The goal is to create a balanced approach where AI supports student learning without diminishing their intellectual development.