Students scored 96 with take-home exams and 48 in class, professor blames AI

Ever since ChatGPT was launched in November 2022, AI tools have been used for different purposes, be it research, coding, or automation. But chatbots like ChatGPT may have been particularly useful in one area – cheating. And now a professor in Brown University in the US has claimed that he has proof to show his students used the AI chatbot to cheat.

Professor Roberto Serrano teaches an economics course at Brown University. He alleges that his class used AI tools to cheat in take-home exams. And once he switched the exams back to in-person, the class average fell from 96 to 48. With 18 students dropping the course right before the exam.

Serrano believes that this needs to stop as it could impact the future of our society. He told Inside Higher Ed, “We cannot afford to have a society in which a significant fraction of our best young minds think that cheating is okay.” He added, “That leads to a declining society, to a failed society. We cannot choose to become idiots.”

More students enrolled for take-home exam course

To give you some context, Robert Serrano’s course ECON 1170 is an advanced undergraduate economics course. In December, he decided to allow take-home exams for students after a shooting in the campus left two students dead and nine injured.

As per reports, historically the course would draw a small group of students, with the number never crossing 30 in the professor’s long tenure. One year, he even had just 8 students. But this time around the enrollment reached a record 86 students. Robert Serrano told Inside Higher Ed that this was likely linked to take-home exams.

Then it was time for the mid-term exams. The professor found that the class had an average of 96, with 40 students scoring a perfect 100. In the past, the course average usually ranged between 65 and 80.

On top of that, Serrano says that he actually prepared harder questions for this year since “take-home was an opportunity to challenge the class a little bit more, given that you’re giving the students unlimited time.

And the answers seemed odd. Robert Serrano said that many responses were correct but had a “convoluted style”, and that when he and his graders ran the questions through ChatGPT, they received similar answers.

Students drop course, average falls

After the midterm, Serrano wrote to students saying he suspected many of them had used AI to cheat. He then changed the final exam to an in-person test and told the class that if the distribution of marks was similar to the midterm, he would keep the midterm. If not, he would void it and reweight the final.

He said he received no response from students, but 18 later dropped the course and another nine, though still enrolled, did not sit the final. Of those 27 students, 22 had scored 100 on the midterm, according to El Pais.

Among those who took the final, the average fell sharply to just 48, and three students scored zero. That is, the average of the class fell below the historical average of the course that never fell below 65.

Serrano then voided the midterm, made the final worth 80 per cent of the overall grade, lowered the passing mark on the final to 40 per cent from 50 per cent. He added that 19 students failed the class.

Brown University professor files complaint

After this ordeal, Serrano submitted the data to Brown's Standing Committee on the Academic Code in May and said he initially received no response. After he spoke publicly about the case, the university asked him to file individual complaints against each student he suspected, along with copies of their exams.

Serrano added, “Their response, I must tell you, is seen as appalling and insufficient by hundreds of people who have emailed me in support, many of them Brown alumni.” The professor also raised concerns on AI detection tools which may flag false positives even while assessing AI-generated text.

Brown University spokesperson Brain Clark said in a statement that the university was in touch with Robert Serrano regarding the allegations. Clark added, “To date, the faculty member has not provided the necessary details to the Standing Committee on the Academic Code to pursue this path toward resolution.”

The case has also fed into a wider debate over AI use on campuses. A recent Brown report on generative AI said most undergraduates use such tools weekly or daily, while many also worry about the effect on their learning and cognitive capacity. The report recommended changes to academic rules to address misuse.

Previously, Princeton University decided to break its 133-year rule of not having invigilators during exams following concerns of AI tools being used for cheating.