University turns to oral exams as AI forces professors to rethink learning

Prof. Matthew Montebello realised something was wrong when a group of students submitted a written assignment that made no logical sense.

The third section contradicted the second, and none of the students could explain the inconsistency. 

“The university has a problem,” he thought, after discovering they had used generative AI to complete the work.

Montebello, a professor of artificial intelligence at the University of Malta, responded by overhauling his assessments.

Since then, he has adopted a new approach: “Assume students are using [AI]. We need to upskill and raise the bar on what we’re doing.”

He now favours oral assessments, such as public debates and presentations, and he is not alone. More lecturers across departments are using oral assessments to curb students’ over-reliance on generative AI.

Assume students are using AI- Matthew Montebello

As tools such as ChatGPT become widespread, university policy encourages educators to embrace them, reasoning that students will need them after graduation. But policies vary across departments, and academic staff are finding their own ways to respond.

Patrick Camilleri, doctor of education, has also observed a shift toward oral assessments, portfolio work and real-world problem solving. The Classics and Archaeology department is also using presentations, creative work and hands-on tasks as alternatives to written assessments.

While AI professor Alexiei Dingli still asks students to write reports, he has added a debating component to ensure they develop a deeper understanding of the material.

Identifying plagiarism

The English department is also adapting. Some ask students to submit drafts and journals, while others continue using oral presentations, already common before the rise of AI. According to department chair James Corby, only one or two instances of AI-related academic dishonesty have surfaced.

While these numbers may seem low, Montebello said identifying AI plagiarism has become harder.

In 2023, the university briefly tested software to detect AI-generated text in students’ writing. They have since abandoned the practice due to a lack of reliability and privacy concerns.

Now, if a student’s work is suspected of AI plagiarism, the university calls the student for an oral interview to test whether they can demonstrate a clear understanding of the work they submitted.

If unable to defend their work, the case is sent to a disciplinary board, as in any other instance of plagiarism.

Many students see oral exams as a helpful alternative. “They make sure you keep thinking,” said physics student Miguel Muscat. He sees this as an ideal option for university education in the AI era.

But oral assessments come with practical challenges, especially for large class sizes. Montebello is developing a technology that would make oral testing more efficient.

‘AI has ruined critical thinking’

Meanwhile, students are grappling with AI’s influence on their education. Suzanne Gafà uses AI to sift through sources more efficiently when doing research.

Others are more critical. “AI has ruined critical thinking,” said Daniele Caruana, a Master’s student in physics. Caruana thinks AI can help make simple tasks like coding more efficient but only after a student has mastered the skill. Most are not disciplined enough to use AI wisely, he said.

Miguel Muscat, who also studies physics, shares his concern. He is thankful that AI was not advanced enough when he was doing his undergraduate maths. He worries he may not have properly learned the course material if it had been.

Emma, a first-year student studying primary education, uses AI to summarise notes.

She also worries about being disadvantaged if she doesn’t use it. In junior college, she followed her teacher’s guidelines to not use AI for her assignments. She says her classmates who ignored the rules scored better and were never caught. 

Lecturers remain uneasy about the long-term impact of AI.

“My greatest concern arises when AI models are used as substitutes rather than assistants,” said Carmel Serracino, chair of the Classics and Archaeology Department.

Philosophy lecturer Francois Zammit agrees. “One of the difficulties that has arisen is the over-reliance on AI.”

Perhaps what is most at risk is the ‘disciplinary intimacy’ of learning that is the slow, uncertain process of grappling with complexity. AI’s speed and fluency can short-circuit the discomfort that often produces insight- Patrick Camilleri

Even when it comes to simple tasks, many people no longer attempt to do things without AI, he said.

Just because people know how to use technology doesn’t mean they understand how it works, Zammit added. He believes students should learn how the technology works, especially since generative AI produces biased results.

Zammit is also worried about declining attention spans. “Deep reading is suffering,” Zammit said, noting that people are becoming too reliant on fast technologies.

“People are prone to reading headlines and not reading the full article.”

He blames both AI tools and the rise of short-form digital media. 

This trend poses a serious learning barrier to a discipline like philosophy, which requires sustained engagement with complex texts. 

Camilleri expressed a similar fear. “Perhaps what is most at risk is the ‘disciplinary intimacy’ of learning that is the slow, uncertain process of grappling with complexity. AI’s speed and fluency can short-circuit the discomfort that often produces insight,” he noted.

Corby put it simply: “I really believe that if we explain properly what it means to study English, then we don’t have much to worry about.”

More training needed

To support lecturers, Montebello runs voluntary workshops on using AI ethically in teaching and assessment. But he believes the university should go further. AI training, he argues, should be mandatory, just like training on plagiarism.

“Once we abandon the idea of ethics and just blindly use the technology, the structure will fall on us,” he said.

As chairperson of the University of Malta Gen AI sub-committee, Victor Buttigieg wants students to develop their ability to judge the output of AI.

While the fast pace of AI advancements is difficult to keep up with, Buttigieg is committed to AI policies that enhance student learning and aims to involve students in the decision-making process as much as possible

The learning process is not a one-way road, he noted. “It’s a partnership more than anything,” he said.