Insider Brief
- Researchers from Cornell University and the University of California, Berkeley found widespread use of generative AI among college students and warned that universities may need to rethink how they assess learning as AI becomes more common in coursework.
- The study, published in Science, analyzed responses from more than 95,000 students across 20 U.S. public research universities and found that 37% used generative AI at least monthly, with usage reaching 62% among computer science students.
- Researchers estimated that 9% of students used AI to cheat on coursework, with misuse rates climbing to 26% among daily users, leading the authors to call for changes to testing methods and clearer AI guidelines.
New research found widespread use of generative AI among college students and warns that current methods of evaluating student performance may no longer accurately measure learning outcomes and raises concerns about the long-term credibility of university credentials.
The study, conducted by Cornell University’s Future of Learning Lab and researchers affiliated with the Student Experience in the Research University Consortium at the UC Berkeley, found more than one-third of students surveyed regularly used AI tools, prompting researchers to suggest that universities may need to rethink how students are assessed.
“Assessment reform is necessary and urgent,” noted study co-author Rene Kizilcec, associate professor of information science and director of the Future of Learning Lab. “The fact that students are misusing GenAI is a problem for assessment validity, and that’s a problem for the credibility of university credentials.”
Kizilec partnered with Igor Chirikov, director of the Student Experience in the Research University (SERU) Consortium for the research. Their study, published in Science, analyzed responses from more than 95,000 students across 20 U.S. public research universities during the 2023-24 academic year. Researchers claim the survey was the largest of its kind examining AI use among university students and offered a broad view of adoption patterns across disciplines and demographics.
What Did The Survey Show?
According to their work, 37% of students reported using generative AI tools such as ChatGPT at least monthly to help complete coursework. Usage varied significantly by field of study and students in disciplines involving more data analysis reported higher adoption rates, with 62% of computer science students saying they used AI regularly compared with 24% of students studying the arts.
The researchers also identified demographic differences in AI use. Thirty-three percent of female students reported regular use compared with 45% of male students. Students from underrepresented racial minority groups reported lower rates of regular use than white and Asian students. The authors suggested these gaps may indicate emerging inequalities in access or adoption and warned they could widen as AI systems become more specialized and potentially more expensive.
How Widespread is Cheating with AI?
The study also examined misuse. Since students may be reluctant to admit cheating directly, researchers used a survey technique known as list randomization. Rather than asking respondents whether they had cheated, researchers asked students how many statements applied to them without identifying which ones. By varying the list content, the researchers estimated rates of AI misuse while preserving anonymity.
Using that method, the researchers estimated that approximately 9% of students had used generative AI to cheat on coursework. Frequent AI users were substantially more likely to report misuse. Among daily users, the estimated cheating rate reached 26%, compared with 7% among monthly users, according to the researchers.
What Did the Researchers Conclude?
The results were lower than many anecdotal accounts had suggested, but researchers said rising AI use could create broader challenges for higher education if institutions continue relying on assessment methods developed before generative AI tools became commonplace.
The central concern, according to the researchers, is not as simple as misuse. It is also whether traditional assignments and testing methods remain effective measures of learning. If AI systems increasingly assist students with writing, coding and analysis, institutions may face difficulty determining which work reflects student understanding and which reflects AI support.
How Can Educators Address Concerns?
The researchers proposed three approaches for universities adapting to AI use in the classroom: returning to tightly controlled testing environments, such as using in-person, pen and paper exams with proctors; establishing clearer rules around acceptable AI use; or redesigning assignments to incorporate AI tools in ways that better reflect real-world professional skills. The authors pointed out that maintaining academic integrity may increasingly depend on adapting assessments rather than simply restricting AI use.
They also cautioned that universities should account for unequal access to AI tools and varying levels of AI literacy among students.
The authors noted that AI isn’t going away, so the technology isn’t solely a disciplinary issue. They suggested universities may need to adapt coursework and evaluation methods to account for AI’s growing role in how students learn and complete assignments.
“As we expect GenAI use among students to only grow, for better and worse, we also expect that GenAI misuse will grow, which is concerning,” Kizilcec added.
Ivan Smirnov of the University of Technology Sydney is a co-author on the paper, Cornell noted.
