Opinion Piece: AI and tertiary education

In last week’s article AI is changing university. Does a degree still matter?, Hunter Yates argues that without drastic change, tertiary education risks becoming a relic.

His argument turns on the fact that cheap and easily-accessible AI models have now developed to the point that they can automate the majority of being a student.

Where unsupervised assessments – essays, tests, etc are concerned, it is now possible for an AI model to log in to a university’s learning management platform and complete these assessments on a student’s behalf, he suggests, without that student ever needing to become aware of it.

Hunter is right that this is a pivotal moment for the tertiary sector. Yet the conclusion that this is an existential risk for the tertiary sector relies on certain assumptions that, when analysed, suggest that the situation may not be as it first seems.

The first of these assumptions is that the sole purpose of a university education is to get a degree.

That is without a doubt, a point of a university education, but it is far from the only one. Many students come to university to learn about a subject that interests them, not just to be able to write on their CV that they have learned something. Even if AI models can be set up to complete assessments for you, they cannot learn on your behalf too.

Other benefits of a university education include the non-academic aspects of the experience: the experiences that you have and the people that you meet. The adage “it’s not what you know it’s who you know” still embodies an important truth. The social networks that you build at university can yield intrinsic and extrinsic benefits throughout your life.

Perhaps it is not so much tertiary education that risks becoming a relic, but that the qualification students leave with can no longer be relied upon to signal what it used to.

This is indeed a significant challenge for the tertiary sector and it is one that we are highly conscious of. This leads me to the second assumption, though: that universities are not adapting, and are maybe not even able to adapt, to this new environment.

To ensure that the qualifications we issue remain meaningful, we need to be confident that the students themselves possess the skills, knowledge and expertise that they are being taught.

In the short term, we will see a return to the more traditional invigilated exam to provide this confidence. Yet, as Hunter notes, this is not a silver bullet. This is partly because wearable AI tech is likely to eventually become undetectable, but it is also because traditional exam conditions require students to possess other skills, such as the ability to manage anxiety and recall information under pressure, that are not necessarily reflective of how well they know and understand the subject matter being tested.

For these reasons, the sector is not assuming that exams alone will solve these problems for us. Instead, it is working hard to develop both individual assessments and assessment schedules that render AI use, if not impossible, at least far less easy and tempting.

For instance, “embodied” assessments, in which doing something is an intrinsic part of the assessment, cannot yet be completed by AI (even if AI can be prompted to pretend it has). Examples of embodied assessments include such things as conducting interviews, collecting specimens in the field, group presentations, role play, Objective Structured Clinical Examination in healthcare and vet and placements in a range of work environments.

Not only can these kinds of assessment not be completed by AI on a student’s behalf, they can also change the motivational structure behind other assessments. If you know that the readings you are being asked to write about are going to give you the knowledge that you will subsequently use in an embodied assessment, you are only going to be cheating yourself if you get AI to do it for you. Your lack of learning is likely to be found out in the real world application. Assessment structures that involve independent learning activities that are then applied to the real world could make the choice to use AI, as opposed to doing that learning for yourself, far less tempting.

This is only one example of the kind of response under development. Others are also being considered, and we are definitely not there yet. But the sector is laser focused on developing our practice to ensure that the badge of tertiary education – the qualification that you leave with – remains meaningful in an age of AI.

This takes me to the final assumption underlying this approach: that AI won’t change. I don’t doubt that AI is here to stay, but to plan your future on the assumption that these capacities will remain reliable, usable, and available for $40 a month would be a significant gamble.

The total market capitalisation of companies that are engaged in AI technology reaches trillions of dollars, despite many of these companies not being profitable at present.

These high capitalisations represent the market’s belief that these companies will eventually yield vast future profits. How will they do this? One worryingly dystopian possibility is that AI models will eventually be able to perform the roles of significant swathes of the workforce at a lower cost than employing humans to perform those tasks.

A more likely future scenario is the one that we have seen with other technology platforms. AI companies that are currently running a deficit will seek to turn a profit through increasing subscription fees “to reflect the true value of their services” and/or making them less accessible by introducing advertising to their platforms.

An additional risk is that the owners of AI tools can affect their outputs. In 2025, the New York Times reported that Elon Musk’s Grok model had been adjusted to more closely reflect his own political views[DW1] How Elon Musk Is Remaking Grok in His Image’. Although Large language models are generally fairly reliable now (albeit still prone to “hallucinations”), to assume that this will remain the case when the companies that supply them are desperate to turn a profit, is a high-risk strategy.

For an individual to plan their future around the assumption that AI tools will remain cheap, accessible and reliable is therefore a huge gamble. We know that these companies will look for a return on their investment. If you outsource your skills and expertise to the systems they control, you lose your autonomy. To retain a measure of control, you need to acquire knowledge, skills and expertise for yourself rather than placing your trust in AI companies to be so good as to provide them for you. This remains the goal of going to university and learning about a subject that interests you.

The challenges raised by AI are real, but rather than seeing them as an existential threat to the tertiary sector, they are an opportunity for universities to fundamentally rethink what we do, how we do it and how we assess it in a way that will ultimately lead to an enhancement in the quality of tertiary education. Rather than rendering the whole idea of a university education redundant, AI is going to make properly-certified qualifications that have been carefully designed to accurately reflect actual learning even more valuable.

Professor William Fish is the Deputy Pro-Vice Chancellor of the College of Humanities and Social Sciences at Massey University.