This week, nearly one in 10 people across the world will use a tool that did not exist just a few short years ago — ChatGPT. According to a September 2025 working paper by economists and researchers at OpenAI, more than 700 million people actively use ChatGPT each week and that “for a new technology, the speed of global diffusion has no precedent.” The multifaceted benefits and costs of artificial intelligence have been fiercely debated, and this debate has been particularly passionate when it comes to the role artificial intelligence plays in student learning and education.
As the Daily has reported, many Tufts professors have changed exam formats in response to increased student AI use, while the Medford School Committee debated an AI use policy earlier this fall. AI has become ubiquitous within education among students — a global survey by the Digital Education Council found 86% of students use AI in their studies, with nearly one in four using it daily. This can be beneficial if AI is used responsibly, in accordance with school or syllabi policy. There is no doubt that large language models can be incredibly good at breaking down difficult concepts simply and clearly. A June 2025 Nature article found that AI tutoring outperforms in-class active learning, with college students also feeling more motivated and engaged when the content was presented by AI.
However, there is growing evidence from recent studies that students ought to be aware of when opting to use AI. A prominent study from the MIT Media Lab split subjects into three groups and asked them to write SAT essays: The group that had access to ChatGPT had the lowest brain connectivity when writing, struggled to accurately quote their own work, had the lowest self-reported ownership of essays and “consistently underperformed at neural, linguistic, and behavioral levels.” Another study also found that while students who used ChatGPT to research a scientific topic did have a lower cognitive load, their reasoning and argument quality were less thorough compared to those who were only allowed to use Google Search. Furthermore, researchers at the Wharton School of the University of Pennsylvania found that while students who had access to AI for practice math problems did better on those problems, those same students did worse on math tests relative to their peers who never had AI access. This research starts to paint a picture: AI has the potential to negatively impact our critical thinking and problem-solving skills, even while saving time or making tasks easier.
Of course, one can reasonably argue that new technology has always led to some skills becoming obsolete. The rise of calculators has led to a decline in the importance of our ability to do mental arithmetic, while the rise of laptops has led to a decline in the importance of our ability to write cursive. However, AI differs from these technologies. Working to reason out a problem or research a topic not only creates better topic understanding, but it reinforces the problem-solving and critical thinking skills we use outside class and in every aspect of our daily life. Thus, the harm of skipping this process with AI is more than simply a reduced understanding of an academic concept or a skill not learned.
We should continue to research the effects of AI on learning. Artificial intelligence likely has an important role to play in education, and I am not arguing that anyone should shun AI. It is an incredibly potent tool that can be used to both enhance understanding when used correctly and reduce understanding when not. When permitted, being able to use AI for research, conceptual understanding, or even coding can save significant time. However, even in a context where the use of AI is allowed, we should take a moment before using AI to determine if we actually need it, or are simply using it as a crutch.
