Opinion
Generative AI has transformed one of higher education's most fundamental questions: how do we know that students have genuinely learned?
Universities across the world are rethinking assessment with unprecedented urgency. Institutions cannot afford to ignore this challenge. They have a responsibility to safeguard academic integrity, maintain public confidence in university qualifications, and ensure that graduates genuinely possess the knowledge, skills and judgement their degrees are intended to represent.
Yet amid this necessary and timely response lies a more fundamental question: Is our response to AI being driven more by fear than by educational evidence?
To understand what is at stake, we need to remember why assessment evolved in the first place. Assessment has never been static. It has evolved alongside our understanding of learning. For much of the last century, assessment focused primarily on measuring what students knew at a particular moment in time. Over the past several decades, however, educational research has fundamentally reshaped that view. Learning is now recognised as developmental, social and contextual, emerging through reflection, dialogue, feedback and practice. No single assessment, however rigorous, can capture the full breadth of what students know, understand and are able to do.
As our understanding of learning evolved, so too did assessment. Universities broadened the ways students demonstrate learning, recognising that different forms of learning require different forms of evidence. Authentic, reflective and collaborative assessments emerged because they aligned more closely with how learning develops. Assessment became more than a means of judging achievement; it became an integral part of the learning process itself.
Remembering why assessment evolved matters because it reminds us that educational change has traditionally been driven by a deeper understanding of learning. Today, however, AI risks shifting that conversation. Instead of asking how assessment can best support learning, we are increasingly asking how it can best resist AI. The distinction may appear subtle, but it is profound. One question begins with learning; the other begins with technological anxiety. And that difference has consequences for what we choose to value in higher education.
Why does this distinction matter? Because assessment does more than verify learning; it shapes it. Educational researchers have long argued that assessment influences what students pay attention to, how they study and what they ultimately learn. If assessment is redesigned primarily to authenticate learning rather than to advance it, we inevitably change not only how students are assessed but also what they value, how they engage and the kinds of learning universities ultimately encourage.
This shift also changes the relationship between students, educators and institutions. Beneath debates about academic integrity lies a more fundamental principle: trust. Effective education has always depended on trust, not instead of accountability, but alongside it. Trust enables students to take intellectual risks, make mistakes, seek feedback and grow. These are not signs of weak learning rather they are the very processes through which deep learning occurs. If assessment is increasingly designed around suspicion rather than trust, it raises the question: what kind of learning culture we are creating?
The implications extend beyond trust to questions of equity and inclusion. No assessment is neutral. Every assessment advantages some learners while creating barriers for others. Decades of research and practice have broadened assessment not only to improve learning but also to recognise the diverse ways students demonstrate knowledge and capability. If AI encourages a return to increasingly controlled and uniform assessment, we must ask who benefits, who is disadvantaged, and whether we are unintentionally undoing years of progress towards more inclusive assessment.
The greater risk is not AI itself. It is allowing the urgency of technological change to narrow our educational imagination.
If our primary goal becomes designing assessments that AI cannot touch, we risk overlooking a far more important question: what kinds of assessment best prepare students to think critically, exercise judgement, collaborate ethically and learn in a world where AI is becoming part of everyday professional life?
Generative AI undoubtedly challenges long-held assumptions about authorship, originality and evidence of learning. Universities cannot ignore those challenges, nor should they. Protecting academic integrity is essential to maintaining public confidence in higher education. But integrity and learning should never become competing priorities. The purpose of assessment is not simply to verify that learning has occurred; it is to create the conditions in which meaningful learning can occur.
This is why the question at the heart of our response to AI matters. Are we redesigning assessment based on what educational evidence tells us about learning, or are we allowing the need to secure assessment to become the dominant driver of change? If our response is shaped primarily by what technology makes vulnerable, we risk designing assessment around defence rather than education. But if it is guided by decades of scholarship on how students learn, assessment can continue to evolve while preserving the values that matter most: trust, inclusion, authentic engagement, critical thinking and human judgement.
At its core, the challenge universities are facing is not simply how to protect academic integrity in the age of AI. It is how to protect academic integrity without losing sight of learning itself.
The future of assessment should not be defined by what AI makes difficult to secure, but by what higher education exists to cultivate. The choices we make now will shape not only how students are assessed, but what universities ultimately choose to value.
Dr. Nira Rahman is an academic at the University of Melbourne.