Why Universities Must Radically Transform Exams in AI Era

Rethinking university assessment is no longer a theoretical debate but an urgent necessity. Since generative AI tools like ChatGPT became widely accessible, their use among students has become the norm. Surveys indicate over 90% of undergraduates now utilise AI for academic work. This shift forces a fundamental rethinking of university assessment models to preserve academic integrity and ensure evaluations truly measure human learning and critical thinking.

Moving Beyond Traditional Written Exams

Current written assignments, like essays, are increasingly vulnerable. AI can produce high-quality text, making it difficult to gauge a student’s genuine understanding. While short-term fixes like oral exams or AI detectors exist, they offer limited solutions. A deeper, structural change in how we evaluate learning is required. We must shift from assessing the final product to evaluating the thinking process itself.

Embracing Conversation-Based Assessment

One promising approach adapts the ancient Socratic method for the AI age. Rethinking university assessment through dialogue means using AI to engage students in dynamic, context-sensitive conversations. Unlike simple chatbots, advanced AI can ask follow-up questions, provide personalised hints, and adapt in real-time to a student’s knowledge level. This turns assessment into a collaborative, exploratory process that reveals how a student reasons, explains concepts, and applies knowledge.

However, challenges remain. AI must balance open-ended exploration with structured dialogue to gather meaningful evidence of understanding. We also must guard against AI misunderstanding student intent and ensure these methods can be standardised fairly where needed, such as in university admissions.

Implementing Continuous and Low-Stakes Evaluation

The high-stakes exam model increases stress and the temptation to misuse AI. Rethinking university assessment effectively involves replacing high-pressure finals with continuous assessment. This model builds a comprehensive picture of student progress through frequent, low-stakes tasks.

AI makes continuous assessment feasible by recording and analysing ongoing student interactions. Specialised learning platforms, rather than general tools like ChatGPT, are needed to track growth and identify persistent misconceptions over time. This approach reduces anxiety, discourages cheating, and encourages authentic engagement, providing a richer, more holistic view of student development.

Focusing on Higher-Order and Human-Centric Skills

The final pillar of rethinking university assessment involves valuing uniquely human skills. Assessments should focus on creativity, collaboration, and critical thinking—areas where humans excel. Students could use AI as a trusted partner in projects, such as designing a sustainability plan or analysing a policy problem. The assessment would then focus on the student’s higher-order thinking, leadership, and ethical use of technology, rather than the output AI can easily generate.

Developing consistent frameworks to evaluate these complex skills across disciplines is a key next step for universities.

The Path Forward for Universities

Successfully transforming assessment requires concrete action. Universities must invest in AI literacy for both educators and students, providing training and support. Clear, updated policies are needed to define acceptable AI use, originality, and critical thinking in this new context. Ultimately, fostering an institutional culture that embraces thoughtful AI integration is crucial. This cultural shift will ensure that AI enhances genuine intellectual development, rather than undermining the core values of education.