Middlebury College Research Reveals Nuanced Story About AI Use in Education

Artificial intelligence has become an integral part of education, capable of accurately summarizing novels, writing essays, solving math problems, and coding – all within seconds. However, a popular narrative online suggests that most college students use AI to automate their work and cheat. Middlebury College assistant economics professor Germán Reyes disputes this notion, citing research conducted with colleague Zara Contractor that reveals a more nuanced story about student AI use.

Reyes and Contractor’s two-part study aimed to understand how students are using AI, what they think they’re learning from it, and what they’re actually learning. The first part involved surveying Middlebury students from December 2024 to February 2025 about their AI use at school. The results showed that 80% of Middlebury students use AI for academic work, but surprisingly, the majority do so as an augmentation tool rather than automation.

Augmentation refers to working with the AI tool to enhance learning, whereas automation involves relying on the AI to do the work for you. Reyes describes this distinction as crucial in understanding how students are using AI. To verify their data and ensure its applicability to other educational scenarios, the research team asked Anthropic, the company behind the AI software Claude, about student usage patterns with college email addresses. The response confirmed that most students were using AI for ‘technical explanations,’ rather than automating work.

A comparison of Middlebury’s survey results with global data from over 50 countries further supported the notion that college students are not primarily using AI to automate their work. Reyes notes that higher education institutions need a deeper understanding of how AI is being used by students before creating policies about its use. He warns against banning AI altogether, as this could inadvertently harm students who benefit from it.

If some students are using AI to learn more and an institution decides to ban the technology due to concerns over its potential misuse, then those students who rely on AI for learning will be negatively impacted. Reyes emphasizes the importance of having a good understanding of ground-level facts when making policy decisions about AI use in education.

The second part of the study aimed to investigate how AI impacts learning. Contrary to expectations based on other research at the time, the findings revealed that students who used AI for augmentation purposes performed better than those without access to AI tools. However, it was not simply using AI that determined student performance – rather, it was how they used it.

Reyes and his colleagues found that when students automated AI to write their essays, they initially did well but ultimately underperformed compared to those who used AI for augmentation purposes. The researchers classified prompts as either augmentation or automation based on the conversations between students and the AI tool. They discovered that augmentation users had a smaller effect on the essay in week one but a larger impact in week two.

This suggests that relying on AI to do the work for you may provide short-term gains, but it comes at the cost of lower learning in the long run. Reyes stresses that the effects of AI largely depend on how students use the tools and explains why similar studies can have different outcomes. He believes that many people tend to be pessimistic about student AI usage, but when speaking with students, they reveal a nuanced and sophisticated understanding of both its benefits and drawbacks.

The study’s findings also highlight the importance of considering factors like grade inflation in educational settings. Reyes notes that if there is high grade inflation at an institution, ‘there is less value to augmentation because you need to differentiate yourself from all other students who also have a very high GPA.’ He suggests that this could lead students to use AI for automation purposes, which may ultimately hinder their learning.

Reyes and Contractor are currently finalizing their paper on the experimentation portion of their study. However, they plan to continue researching student AI usage in education, with plans to conduct another survey in 2026. They aim to explore new questions, such as whether students are paying for premium versions of AI assistants.