AI has turned university assessments into a ‘wicked problem’ – and we don’t have all the answers
Unis must ensure degrees demonstrate students’ learning, rather than how well they’ve used artificial intelligence. But it’s an ongoing challenge.
The emergence of ChatGPT in late 2022 has sparked concerns about university assessments, questioning whether they are testing students' learning or just their ability to use generative AI. Universities and teachers have been striving to reform assessments to ensure degrees genuinely reflect students' knowledge and skills. However, this presents an ongoing challenge for the sector.
Research indicates that nearly 54% of Australian university students' submissions from October 2025 to April 2026, processed through Turnitin, contained some form of AI use. About 10% of these submissions had over 80% of their material generated by AI. This doesn't necessarily mean all students using AI are cheating, as it can be employed selectively to help develop ideas, edit sentences, enhance drafts, or provide feedback on arguments. Many students also use AI to aid their assessments before starting to write.
Despite this, the rules regarding acceptable and unacceptable AI use remain unclear to both students and teachers. It's unclear whether merely improving a draft with AI means correcting grammar alone or restructuring the argument. The line between appropriate and inappropriate AI use can vary between assessments and even among individual students within the same assessment.
Australian universities have a regulatory obligation to ensure assessments confirm students have achieved the required knowledge and skills. Various approaches have been taken, with the University of Sydney's "two-lane approach" being influential. This approach divides assessments into two categories: "secure" assessments require students to demonstrate their skills and knowledge without AI use, while "open" assessments permit AI use, typically for feedback and learning purposes.
However, keeping up with increasing reliance on in-person assessments is proving difficult, as they can be expensive, limit assessment methods, and unfairly bias against certain student demographics.
Monitoring AI-enabled smart glasses in exams, which can resemble ordinary eyewear and provide assistance without drawing attention, also poses a challenge for universities. The idea of treating assessment solely as a security issue to deter AI use may focus on control measures rather than students' education. Instead, universities should focus on which problem each assessment solution solves and whether it creates new issues.
For example, if the goal is to keep AI out of assessments, more supervision and tighter controls may be seen as progress, but if the aim is to prepare students to use AI effectively in the workforce, such restrictions could hinder the desired learning outcomes. Therefore, understanding which assessment approaches work in specific disciplines and contexts is crucial for progress.
Written by urgent.news from The Conversation AU's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
