Research at Middlebury College reveals nuanced story about artificial intelligence use

Researchers at Middlebury College have released a two-part study challenging the narrative that students primarily use artificial intelligence to cheat, instead identifying a distinction between augmentation and automation. The study found that while 80% of students utilize AI for academic work, the majority use it to enhance their understanding of complex concepts rather than to bypass effort. These findings suggest that the long-term impact of AI on learning depends heavily on user intent, providing critical data for educational institutions developing AI governance policies.

Middlebury College assistant economics professor Germán Reyes and colleague Zara Contractor conducted research between late 2024 and mid-2025 to investigate the reality of AI adoption in higher education. Their survey of Middlebury students revealed that 80% use AI for academic tasks, but contrary to popular belief, most utilize it as an augmentation tool for technical explanations rather than for automation. To verify these findings, the researchers collaborated with Anthropic, the developer of the Claude AI software, which confirmed that students using college email addresses were primarily seeking conceptual explanations rather than automated output.

The second phase of the research involved a controlled experiment where participants were split into groups with and without AI access to research and write about CRISPR gene-editing technology. The results showed that students who used AI for automation—having the tool write their essays—saw significant short-term grade boosts on their first assignments but performed significantly worse on follow-up tests and essays conducted without tools a week later. Conversely, students who used AI for augmentation performed better on average in the long term than those who had no access to AI tools at all, indicating that the technology can facilitate deeper learning when used as a supplement.

Professor Reyes emphasized that these findings have major implications for how higher education institutions should approach AI policy. He argued that blanket bans on AI could inadvertently harm students who use the technology to improve their learning outcomes, as the negative effects are largely confined to those who use it to replace their own cognitive effort. The study highlights that the impact of AI in the sector is highly nuanced, suggesting that administrators must understand ground-level usage facts to create effective and fair academic guidelines.