Professors Rethinking Testing Amid Growing AI Use
About the case
Professor Roberto Serrano, who teaches economics at Brown University, replaced a traditional in-person midterm with a take-home exam after students expressed anxiety about returning to the classroom following a December mass shooting on campus. When grading the take-home exams, the class earned an average score of 96%, with many students receiving perfect or near-perfect grades. Historically, the average midterm grade in the course has ranged from 65% to 80%. Serrano noted the take-home exam was more difficult than previous midterms because students had unlimited time to complete it.
Serrano became concerned that some students may have used artificial intelligence to complete the exam after reviewing their responses. He compared answers using ChatGPT and believed many appeared to have been AI-generated, though AI tools cannot reliably determine whether text was written by a human or generated by AI.
Final exam and grading changes
For the final exam, Serrano returned to an in-person format. The results were dramatically different: 18 students dropped the course, nine did not take the final exam, and the remaining students earned an average score of 48.6%. Three students received zeroes. In response, Serrano voided the take-home midterm and made the final exam worth 80% of the course grade. Students who earned at least 40% on the final received passing grades under the revised grading policy, but 19 students ultimately failed the course. Serrano has brought his concerns to Brown University administrators, urging the institution to address the challenges artificial intelligence presents to academic integrity.
AI use among college students
The largest study examining undergraduate use of generative AI was led by Igor Chirikov, a senior researcher at the Center for Studies in Higher Education at the University of California, Berkeley. More than 95,000 students at 20 research-initiative public universities were involved. About two-thirds of respondents said they used GenAI and almost 40% used it monthly or more; at least 9% of students who used AI reported using it to cheat. Chirikov said there is no way to ban AI, but professors should find new ways to measure student knowledge that cannot be used with AI. The research concluded that AI initiatives need more resources and higher prioritization.
AI did not create cheating—it is a tool
In a paper by Austin Sarat, a professor at Amherst College, he argues that artificial intelligence did not create academic dishonesty but has simply become another tool students can misuse. Sarat wrote that students cheat for many reasons, including pressure to earn good grades, lack of confidence, and poor study habits.
"I don't think of my students as cheaters and I don't want to regard them with the kind of suspicion that turns teaching into a policing activity. But it is my job and that of the college where I teach to recognize that our students need a lot of help to develop good academic habits. Unless colleges acknowledge these facts, I believe they have little chance to curb the pervasiveness of cheating."
AI detection tools professors may use
Companies have developed tools that attempt to identify AI-generated writing for professors to use while grading. Turnitin, widely used by schools for plagiarism detection, also offers AI-writing detection. Other tools, including Scribbr and Pangram, claim to identify text generated by platforms such as ChatGPT and Microsoft Copilot. However, researchers and AI developers caution that AI detection tools are not fully reliable and can produce both false positives and false negatives. Many educators recommend using AI detectors only as one part of a broader review rather than as definitive proof of misconduct.
Looking ahead
As AI becomes a permanent part of the classroom, educators will need to balance its benefits with policies that uphold academic integrity and meaningful learning.