ADVERTISEMENT
Don’t throw the formative baby out with the AI-tainted bathwater
The OfS’ call to ensure the integrity of assessment does not mean scrapping coursework and returning to traditional exams, says Chris Rust

Source: Emre Ceylan/Getty Images
“Office for Students says institutions in England should return to traditional exams and in-person presentations.” That is how The Guardian summarised the OfS advice last week on how universities should respond to the fact that, according to one survey respondent, artificial intelligence has made it “close to impossible” to assess coursework fairly. And I can already imagine quite a few faculty members cheering at this apparent green light to roll the clock back on assessment, and the thought of no more coursework to mark.
But the cheers are misplaced. For one thing, the OfS report mentioned the word “traditional” only once, in a very brief reference to the University of Melbourne’s “typology of secure assessments”, which “includes a range of practical options alongside more traditional approaches such as closed-book exams”.
The regulator’s view is that “institutions should develop accessible and robust approaches to assessment, which effectively assure student achievement”. And “secure approaches designed to prevent students from using AI will often be the most effective as they provide a high degree of confidence that students are demonstrating their own knowledge and skills”. But the OfS also advocates “the use of innovative assessment approaches that reflect the ‘real world’ application of students’ knowledge and responsibly integrate AI”.
This is quite right. It is a mistake to assume that exams must be “traditional”. And overturning decades of progress on reforming student assessment in pursuit of AI-proofing would be a textbook case of throwing the baby out with the bathwater.
If we go back to basics, let us remember that assessment has at least two distinct functions – formative and summative. And the increased use of coursework assessment has very much been driven by the desire to provide formative feedback opportunities – along with motivation to undertake the work. If we are creative, the existence of AI should not prevent this from continuing.
ADVERTISEMENT
We can still ask students to produce work that will not count for marks but does involve self-assessment and peer-assessment processes as a formative part of their learning. As to what will motivate them to do this seriously, or stop them using AI, what if they know this subject will definitely be in the final summative exam?
In this case, even using AI and then discussing the resulting product with a peer, considering its merits and flaws, can play a productive part in learning. So can individual or group presentations that are interrogated and both peer- and tutor-assessed but that do not count summatively.
The Guardian’s reference to “traditional” exams evokes, to me, the very worst characteristics of exams – regurgitating as much as you can on a subject in a short space of time. Even before AI, if a student could question-spot what was going to turn up in the exam, they could rote-learn an answer from a book – or one provided by a friend or an essay mill – with little or no understanding of what they were writing.
I fully accept the argument that AI means that summative assessment needs to be undertaken in controlled conditions to ensure it is the students’ own work. And obviously, the checks and preventive measures already needed to thwart the use of smart watches, phones and other means of cheating will need to be constantly monitored and strengthened.
But what is wrong with open-book exams, for example? What is wrong with letting students know, for both motivation and fairness, the topics that will be on the exam? The important point is that the tasks set in the exam should be focused on examining whether the students have acquired the skills, knowledge and abilities necessary to demonstrate the course learning outcomes – described in the pedagogical literature as “doing-it” exams.
In that context, why shouldn’t students take the coursework essays in with them; even if these have been written by AI, it would motivate the students to produce them. The students could then be asked to use what is in them to take a different viewpoint, or perspective, or develop and build the argument further, possibly having been given additional information to work with.
Other examples of “doing-it” exams could include asking for comment on a previously unseen article’s theoretical perspective; analysing a case and giving advice to a client; or commenting on a science, engineering or health article describing experimental work. None of this is particularly new, but I don’t believe it is what the term “traditional” brings to mind.
ADVERTISEMENT
And if vivas are good enough for PhDs, why don’t we make far more use of them in undergraduate programmes? If a student knows they will be interrogated about their coursework, there is good evidence that they will be far less inclined to cheat in its production. A common argument against this is the claim that staff don’t have time, but the time could be found if there were fewer but better assessments, such as a viva.
For a much more detailed consideration of how to sensibly deal with AI and assessment, I recommend that colleagues read Velma Adams’ recent paper. But the long and short of it is that while the OfS is right that summative assessment needs exam conditions, that should not be used as an opportunity to scrap coursework or to return to what many of us would see as the bad old days of exams.
Chris Rust is emeritus professor of higher education at Oxford Brookes University.

