Opinion
As AI changes assessment, are universities doing enough to ensure students are ready for what comes next?
I remember the architecture students from my years at UNSW. They were a passionate bunch. They carried their portfolios around campus, always ready to show someone what they had designed.
I still remember the computer labs and design studios, where students sometimes worked for 48 hours straight. Pizza boxes everywhere. Drawings, models and unfinished designs covering every available surface.
What struck me was how much they cared about their work. They wanted to get things right. They knew what they could do, what they could not yet do, and what they needed to improve. They would spend hours changing something because they were not satisfied with it. The mark mattered, of course, but it rarely seemed to be what drove them.
Their portfolios meant something beyond university. They were building a record of what they could do, something they would eventually show to employers or use to apply for further study.
I noticed something different among engineering students. They were just as capable and often worked extremely hard, but many approached assessment more pragmatically. How much was the assignment worth? What did they need to do to get a distinction? Was it worth spending another ten hours on something that contributed only 10% to the final mark?
I used to wonder why the two groups seemed to have such different relationships with their work. Looking back, I suspect we, as educators, had something to do with it.
Perhaps it begins before university. In the final years of secondary school, enormous effort goes into examination preparation and achieving the results needed for a competitive ATAR. Much of that effort involves genuine learning, but students also learn to pay close attention to what is assessed and how it counts.
Then they arrive at university, where we encourage curiosity, independent thinking and experimentation, while continuing to assign weightings to every assessment. Students learn where to concentrate their effort. I cannot say I blame them. They have learned how the system works.
Perhaps we, too, have become accustomed to treating a passing mark as the end of one subject, without paying enough attention to what students need to carry into the next.
Recently, I read an article by Elizabeth Angstmann and Helen Gibbon, When Marks Stop Driving Learning, in Needed Now in Learning and Teaching. It describes a UNSW pilot of competency based assessment involving 25 courses and more than 5,000 enrolments. Students reported less anxiety, more cooperation and greater attention to learning.
The findings are encouraging, although they do not tell us whether students retained more knowledge or were better prepared for subsequent study.
But changing how we assess achievement does not necessarily resolve another concern: what happens to essential knowledge after students pass and move on?
That brings me back to a question I have asked myself many times: what does it actually mean when we say a student has passed?
Take civil engineering, the discipline I know best.
In first year statics, students learn the first principles of structural engineering: equilibrium, forces, moments, free body diagrams and support reactions. I have always thought of these as the alphabet of civil engineering.
By the time students reach solid mechanics and structural analysis, they must bring these principles together to understand increasingly complex problems. If statics is learning the alphabet, those later subjects are the equivalent of reading novels.
But how can we expect someone to understand a novel if they have not learned the alphabet?
A student may pass statics with an overall mark of 50%. That does not mean they understand only half the subject. They may have a sound grasp of every essential concept, but they may also have performed well enough in some assessments to compensate for serious weaknesses elsewhere.
A student might pass without being able to construct a reliable free body diagram or apply equilibrium equations independently. To me, that is rather like passing the alphabet without having mastered the vowels.
Yet in the following year, we expect that student to analyse beams, calculate stresses and understand structural behaviour.
When students struggle, we tend to look at their mathematical ability, study habits or the increasing difficulty of the material. I wonder how often we ask whether the first principles were secure in the first place; or whether they have been revisited before students are expected to apply them to more complex problems.
This is where our approach to assessment and progression deserves closer attention.
Universities understandably want to make courses more flexible, reduce unnecessary hurdles and allow students to progress without avoidable delays. These are legitimate aims. Not every assessment hurdle has a sound educational justification.
But in engineering, knowledge is cumulative. What students learn in one unit is often essential to understanding what follows.
We can simplify progression requirements, but we cannot simplify away the intellectual foundations of a discipline.
A student may have completed statics and satisfied the formal prerequisite for structural analysis. That tells us they passed the earlier unit. It does not necessarily tell us whether they can still construct a reliable free body diagram or apply equilibrium equations independently.
I sometimes wonder whether we spend more time checking that individual units meet their learning outcomes, than examining what happens to that learning between units. A degree is not simply a collection of completed subjects.
AI makes this question more difficult.
I have written previously about AI and the development of engineering judgement. Here, my concern is what happens much earlier, when students are still learning the alphabet.
A student can use AI to complete a statics problem, then a beam analysis, and eventually a much more sophisticated design task. Each submission may meet the assessment requirements. The student receives a mark and moves on. But what happens to the knowledge they were supposed to acquire along the way?
AI has not created the problem of students progressing with gaps in their understanding. It may, however, keep those gaps hidden longer as students produce increasingly sophisticated work in successive years of their degree.
A convincing technical solution does not necessarily establish that the student understands the principles behind it. Nor does being able to explain an AI assisted solution immediately after receiving it tell us whether the student can apply the same knowledge independently months later.
Students need to learn how to use AI well. But the more capable these tools become, the more important it is to distinguish between completing a task and developing the knowledge needed for the next one.
AI is, in important respects, a black box. Engineering education must give students the knowledge to evaluate what comes out of that box, even when they cannot inspect every aspect of how it works.
The concern is not simply whether we can trust an individual assessment. It is whether we can trust the progression of learning across the degree.
If a student passes statics but cannot apply equilibrium in structural analysis, whose responsibility is that? The student who has forgotten? The lecturer teaching structural analysis? The academic who taught statics the previous year?
I think the responsibility belongs to the course as a whole.
Academic teams need to agree on which concepts are indispensable, where they are introduced, how they are assessed and when students should demonstrate them again. This does not require more hurdles everywhere. It requires carefully placed checks of essential knowledge, meaningful feedback and opportunities to improve.
Imagine a lecturer discovering at the beginning of structural analysis that a substantial number of students cannot reliably apply equilibrium equations.
A short diagnostic assessment could identify the gaps, but what happens next matters more. We could leave the lecturer to address the problem while delivering an already demanding subject or tell students they should have remembered what they learned the previous year.
Alternatively, the academics responsible for both units could work together to understand where students are struggling, provide targeted support and reconsider how those principles are taught, assessed and revisited.
If the same gaps appear year after year, that tells us something about the curriculum, not just the students.
We can still recognise different levels of achievement and provide flexible pathways. But we should have good reason to believe students possess the knowledge on which subsequent learning depends. An assessment completed a year earlier may not provide sufficient assurance.
I think again of those architecture students.
Their portfolios let them see their work developing, recognise what had improved, and understand what they still needed to learn. They were accumulating evidence of capability, not simply marks.
I would like engineering students to leave university with a similar body of work: designs, investigations, calculations and projects showing how their thinking has developed. Work they can explain and defend. Mistakes they have recognised and corrected. Decisions they have reconsidered.
AI will be part of that work, as it should be. But we must also be able to establish what students have learned through doing it.
We have well developed systems for recording when students complete a unit. We are less certain about what happens to the knowledge after that unit is finished.
A passing grade closes one academic requirement. But the knowledge it was intended to represent is needed again, often in more demanding circumstances.
Perhaps we should spend less time asking whether students have accumulated enough marks to pass, and more time establishing whether they are ready for what comes next.
If we expect students to understand increasingly complex novels, we have a responsibility to make sure they know the alphabet. And to make sure they have not forgotten it by the time they turn the page.
Professor Zora Vrcelj is Head of Built Environment & Engineering, Chair of the Academic Board and a member of Council at Victoria University.