Dr Leon Furze

Earlier this month, teachers at Mazenod College in Melbourne’s outer east became suspicious about a batch of Year 12 English oral assessments. The school reviewed the submissions and found, in the principal’s words, evidence “that suggested artificial intelligence tools were used by several students”. Up to 50 boys, almost a quarter of the school’s Year 12 cohort, had marks deducted.

The story arrived amidst an already hostile media environment. In May, the Sydney Morning Herald covered a new report on AI in NSW schools under the headline “AI poses ‘urgent threat’ to student learning and the HSC”, and EducationHQ ran with “Get AI out of schools”. I wrote a few weeks ago about how students around the world are experiencing this discourse, and how conflicted they are about a technology they use constantly and largely resent.

In this post and the next I’ll focus on another aspect of the story: what the cheating panic is doing to the conversation about assessment, and whether the response that seems to be taking shape is actually the one our assessment systems want.

Trust issues

The reasoning kicking around in news stories goes something like this: if students can use AI to complete coursework (or any out of school assessment), then coursework can no longer be trusted, and the only assessment we can rely on is the kind done under exam conditions, where the timing is fixed, a supervisor is watching, and nobody has seen the questions in advance.

It sounds like common sense, it sounds “rigorous”, and, framed like this, the exam sounds like the only option. In K-12, we often look to our neighbours in tertiary study for answers, and universities around the world have already acted on the reflex. Princeton has scrapped its 133-year-old honour code in favour of proctored exams, and plenty of Australian universities have moved towards so-called “secure assessment”. Despite the framing of secure assessment as “typically assessment types or tasks that are “observed” in some way and/or have a significant dialogic component“, the pull towards the exam hall is strong, and AI has given it a fresh justification.

This reflexive turn towards exams gets one thing right and almost everything else wrong. In part two of this article, I’ll make the case that a wholesale retreat into exam conditions would actually put K-12 schools in Australia at odds with the published policies of every senior secondary assessment authority in the country.

First, though, a closer look at that headline grabbing report from May.

The Learning First Report

Learning First’s report, AI Use in Schools: Taking Action Now, is based on NESA survey data from around 3,400 NSW teachers and 750 school leaders. It reports widespread student use of AI for assessment, teachers who feel unable to stop it, and a phenomenon the report calls cognitive outsourcing (drawing on language from this UTS paper), where students delegate their thinking to a chatbot rather than working with it. None of that will surprise anyone who has spent time in a school recently, and the report is worth reading in full.

The SMH described it as calling for “an urgent review of the use of take-home assignments in year 12”. The report’s own recommendation is worded differently:

“system leaders need to set the direction of all their schools by reviewing and strengthening the parameters for internal assessments that contribute to student results and qualifications”.

It asks systems to evaluate, subject by subject and year by year, which assessment tasks are susceptible to AI interference, and to redesign the vulnerable ones. That is a review of how tasks are designed and administered, which is not quite the same thing as a review of whether take-home tasks should exist.

More interesting is that the report explicitly warns against the over-correction its media coverage seems to invite. On the risks of retreating to supervised conditions across the board, it says:

“A hasty shift to fully in-class, device-free assessment without deliberate design risks stripping away important elements of tasks, limiting the value of assessment.”

And on the tasks the headlines imply we should be abandoning:

“Tasks like longer essays and larger projects should remain, but as part of an assessment program that also contains other methods of assessment that are less susceptible to AI interference.”

The same section argues that “Students still need opportunities to write at length, complete substantial projects and conduct research.” So the report itself is not asking for the end of coursework. It is asking for assessment programs designed so that teachers can still trust the evidence they collect.

The report does, of course, focus on the benefits of external assessment and supervised tasks. It states that “external assessment tends to better protect assessment integrity than does internal assessment”, and its selection of options for vulnerable senior tasks includes redesigning a take-home essay as “an in-class essay where students respond to an unseen prompt”, converting tasks to pass/fail hurdles, reducing weightings so that “a task once worth 30 per cent may be reduced to 10 per cent”, and, described in the report’s own words as “the most extreme response”, removing internal assessment from some subjects entirely.

The Learning First report should be read as a situated document, not a neutral reporting of survey results. Its frame is the cognitive load and knowledge-rich curriculum tradition running through Kirschner, Wiliam, Willingham, Christodoulou, and Hendrick, and it converts teacher observations (from NESA data) into evidence for a prior theory of how learning happens. It also arrives at the same moment NSW has mandated explicit teaching across its public schools, with the department’s “What Works Best 2025” approach. Victoria has released VTLM 2.0 along the same lines. The report shares the intellectual framing that underwrites both reforms, so it’s not landing in a vacuum: it reinforces, and draws authority from, a pedagogy already ascendant in the two largest Australian systems. It should be read within that political economy.

Ultimately, the findings rest on a single NSW study, conducted as part of a larger syllabus implementation project, supplemented by selected research and expert opinion the report concedes is drawn partly from blogs and podcasts (not that there’s anything wrong with blogs and podcasts!). Student voice is largely absent.

Even with these caveats, the report makes useful observations and is more nuanced than its media reception: it does not call for AI to be removed, it preserves extended writing and research, it distinguishes between users and school levels, and it frames the work as systemic rather than a set of individual fixes. That nuance was of course the first thing the headlines stripped away, because a call for large-scale, calibrated, system-level review does not read as well as a “clear and present danger.”

An exam by any other name…

And all of this is why the Mazenod story is worth a second look. The assessment at the centre of it was an oral task, described by The Age as part of the school’s “year 12 oral English exam process”, and counted by the principal among the school’s “assessments and examinations”.

As a former Head of English, VCE English teacher, and VCAA examination assessor, I think it’s worth pointing out that there is no such thing as a “VCE English oral exam.” There is a coursework outcome which carries 20/100 marks (5% of the overall VCE English study score) for a 3-5 minute presentation on a persuasive topic. It’s a nightmare task, frankly, requiring schools to supervise hours of orals in an outcome barely anyone takes seriously because it’s not on the final, external examination.

This might sound like pedantry, but I think this is the whole point. Conflating the oral School Assessed Coursework (SAC) Outcome in Unit 4 VCE English with “exams” and then posing AI as a threat is a rhetorical sleight of hand that does both the school and the students a disservice.

These students did not cheat in an exam. They exploited the unsupervised preparation of work for something their school (or possibly just The Age) called an exam, but which was actually a piece of coursework that most schools in Victoria struggle with.

Coursework is not an exam

The bigger issue – the one that the media is fuelling, but which has been a problem for as long as I have taught – is the misconception that senior secondary coursework should mirror the exams.

Back in 2021, a year before ChatGPT existed, I wrote on this blog that we were doing coursework wrong: VCE SACs were routinely being run as miniature exams, timed and unseen, in a way the study design never asked for, at a cost to drafting, feedback and the students who don’t perform under exam conditions. The exam reflex I mentioned earlier was alive and well long before GenAI arrived. All AI has done is hand it a respectable-sounding justification.

That logic carries more weight in higher education, where a lecturer might see a student across one or two semesters, in large cohorts, with little one-to-one contact. In K-12 it falls apart. Teachers might work with the same students for years; they know their writing, their tics, their verbal reasoning, the gap between what a kid can do on Tuesday and what suddenly appears on Thursday. The relationship that makes authentication possible is the one schools are built around, and it is exactly the relationship every assessment authority in Australia points to when it asks teachers to attest that work is a student’s own.

Even the Learning First report does not ask schools to collapse coursework into exams; it asks them to design assessment programs teachers can trust. The retreat to the exam hall is not what the research says – it is what the headlines say, and the two have drifted a long way apart. The language we have been handed, the threat, danger, cheating, crisis, all points in one direction. But the people who set the rules and the people in the classrooms are pointing in the other.

I’ve spent some time recently going through the handbooks, manuals and syllabus documents of the Australian senior secondary authorities to see what they actually say about what exams are for, what school-based assessment is for, and where exam conditions are required. That will be part two of this article.