AI poses risk of surge in cheating in online MBA programmes

In 2023, when Christian Terwiesch, a Wharton School professor, asked ChatGPT to explain an industrial bottleneck concept, he gave its response an A+. “Not only is the answer correct, but it is also superbly explained,” he wrote in a paper on his experiment.

Since then, generative AI’s growth in terms of capabilities and numbers of users has left schools, particularly those offering online MBAs, worrying about the potential for a tidal wave of cheating. “The student population is faced with a choice they’ve always had,” says Megan Leroy, assistant dean at University of Florida’s Warrington College of Business. “But it’s now easier to make the wrong one.”

Daniel Pearson, director of academic environment at Warwick Business School, says that AI will not only create slide decks and analyse company reports, but it can write essays using the student’s tone of voice. “It will probably advise you that cheating is bad. But it will enable that, if that’s what you want to do.”

For business schools, this revolution in the use of technology raises many questions. What is appropriate use of AI, and what is not? How can schools detect cheating, and what should they do when they encounter it?

The biggest challenge is assessing the size of the problem. Australia’s AGSM is looking at results as one possible indicator of misconduct. “Our thinking is that if students are using AI to get better grades, we’d see grade inflation,” says Michele Roberts, head of school at the AGSM and associate dean at UNSW Business School, of which AGSM is part. “But our grades are not going up.”

Others are less confident. “This fear of assignments getting dropped into ChatGPT and coming out the other side, that’s a really tough one to measure,” says Bradley Staats, senior associate dean for strategy and academics at University of North Carolina’s Kenan-Flagler Business School.

Detecting individual incidents of cheating is even harder, particularly among online MBA students. “In a classroom, I can walk about and see what’s on your screen or that you’re not using a tablet,” says Emily DeJeu, assistant teaching professor at Carnegie Mellon’s Tepper School of Business. “In an online environment that just breaks down.”

Schools have been trialling machine learning tools that distinguish between written assignments generated by AI and those created by humans. These include CrossPlag, Copyleaks, Turnitin, GPTZero, and Originality.ai.

Schools have been trialling machine learning tools that distinguish between written assignments generated by AI and those created by humans © Casimiro/Alamy

However, most are seen as ineffective in the face of generative AI. “Typically, these provide some sort of score. That may be suggestive, but rarely is it conclusive,” says Staats. “It’s not just: push a button and this goes away.”

Studies have also shown that these types of tools produce inconsistencies such as false positives and struggle to keep up with rapidly evolving AI text generation models.

Moreover, it is hard to confirm misconduct beyond reasonable doubt. “AI introduces a new level of he-said-she-said,” says DeJeu, adding that the awkwardness of questioning students can make faculty hesitant to report potential breaches.

All this is prompting schools to rethink their assessments, particularly for online MBAs.

The rethink includes shifting the balance between essays and group projects, asking students to defend their written work and using oral assessments to test their critical analysis abilities.

Warwick Business School is also piloting methods for scaling up oral assessments. “While it’s a nice idea to have a viva for every student, when you’re bringing in 400 online MBAs every year, doing that for every module creates a huge workload,” says Warwick’s Pearson.

One option, he says, would be for students to submit recordings of oral assessments where they answer random sets of questions or participate in simulations, presentations or debates. “You have the same issue as any asynchronous submission — they may use AI to help them,” he says. “But we are experimenting with different approaches.”

When it comes to clarifying what constitutes cheating, schools often leave this up to faculty. At the AGSM, university guidelines set broad expectations but MBA academics design guidance for each assessment project, says Roberts. “Carnegie Mellon is very committed to faculty independence,” says DeJeu. “They tell faculty to make their expectations clear and design assessments to support that.”

Growth happens when you’re working on challenging problems — not when you’re looking for shortcuts

Michele Roberts, AGSM

However, some believe that only a change in student attitudes can deter AI-driven cheating. “We talk to MBA students about a growth mindset and that really helps,” says Roberts. “Because growth happens when you’re working on challenging problems — not when you’re looking for shortcuts.”

But if cheating has been confirmed, the next question is what to do about it. Here, solutions already exist. Most schools turn to their university’s office for academic integrity, which employs the same processes as it would for any academic misconduct — gathering and documenting evidence and imposing penalties, from failing students for an assignment to expelling them.

“If students have copied and pasted text from ChatGPT, they’re going to be dealt with as we’ve always dealt with students who’ve copied and pasted from unacknowledged sources,” says Roberts.

Nevertheless, schools are having to move fast to keep pace with the implications of AI’s widespread use.

“We can’t ignore the fact that cheating is now scalable,” Warrington College’s Leroy says. “But AI angst has caused us to re-evaluate what our responsibility to students is and to make sure we’re creating ethical future leaders.”