Earlier this month, Ofqual’s chief regulator Jo Saxton expressed her opinion that coursework-style assessments would be stronger if carried out under controlled conditions to prevent candidates cheating by using tools such as ChatGPT.

“If I was running a centre now, I wouldn’t be asking for the pieces of coursework or the essays that contribute to the grade to be done at home or in school holidays - I’d be doing them in invigilated conditions and in my centre,” she said at the Association of School and College Leaders conference.

For many teachers, this will have raised the spectre of GCSE controlled assessment (CA) that became so utterly unworkable and open to all kinds of breaches that it was finally dropped in 2015 under the Govian reformed specifications. 

While the GCSE situation persisted for some time before being dropped, CA has never been tried at A level - for very good reasons, not least the time needed to complete an assignment from start to finish. 

Yet Saxton’s comments could be taken to mean that if coursework is to remain a part of the assessment portfolio students are assessed by, the rise of AI could see CA back with a bang and the end of any form of non-examination assessment (NEA).

Why do we need non-examination assessment?

Of course, the easiest way of countering the threat of cheating would be to abolish coursework completely.

After all, the main skills of my subject (English literature) at A level - comparison, analysis, synthesis of sources, contexts and texts - are already covered under exam conditions. 

From the regulator’s perspective, the assessment objectives have already been met by the existing exam papers.

Instead of bothering with all the trouble of NEA, another examination paper could cover the remaining set texts and consolidate the AOs. Job done.

However, the educational argument for NEA cannot be overlooked. 

Exam performance is based on pre-digested texts, critics and contexts that are often provided by awarding bodies and teachers, where candidates respond to unseen but reasonably predictable questions. This requires limited initiative.

NEA is much bulkier, messier and altogether more challenging. Projects are large scale and require planning, time management and independent research and decision making about what to select and how to include it. 

The most stretching form of coursework is where a student chooses texts (preferably ones for which there are no published study guides) then devises the question and researches independently, creating their unique bibliography as they go. 

The process is as valuable as the outcome. Students are being prepared for university and the world of work where they will have to independently set up projects and research unfamiliar sources in depth. 

How feasible would it be to use CA?

Because the scope of NEA is so large, it cannot in its entirety be fitted into lessons. 

A-level students are expected to study five hours a week per subject in addition to timetabled lessons.

This time is often used for NEA, with some class time so that work can be overseen (but not directed) by teachers, who can then authenticate the final piece. Teachers won’t see everything but should see enough to detect potential malpractice.

Controlled assessment takes away some independence and is not cheat-free. Moreover, because ChatGPT can purportedly assist with anything from sifting through research, planning and structuring an A-level essay to producing the final essay in suitable format and style, schools would need to cover the whole process from first reading of texts, selection of detail and essay planning right through to the final draft.   

Clearly, then, this would be a massive change that would put pressure on the existing timetable and have a huge impact on teaching time and workload.

What’s more, it could have an unfair advantage for independent schools, which run longer school days to include lunchtime clubs or boarding schools.

They would therefore be less affected by any such change. Urban schools in areas with good transport links could also accommodate longer days better than rural areas with spasmodic transport links.

How much, or how little, control?

First and foremost, how long would the CA need to take and how would the time be distributed? 

Reading and research, planning and structuring would need to be monitored to avoid conferring a time advantage on those who use AI to find and collate sources. And since ChatGPT can churn out superficially plausible format and style, CA should cover drafting and redrafting.

If school leaders decided to leave CA to teachers to manage in timetabled lessons, how would work done in between be policed to ensure no unacceptable assistance was provided? Online storage of work would need to be secure to prevent loss of work. 

Alternatively, running NEA over one or two whole days - as some schools do for art exams - might be better because it allows nothing to slip between the gaps in sessions and all students work under the same conditions. 

Would a period of between eight and 16 hours suffice to get the whole thing written? Specialist teachers would be most able to spot malpractice during the process at a sophisticated level. 

But what impact might that have on lesson provision for their other classes? Whichever model is chosen should be the same across all centres to keep conditions the same.

The greatest nightmare in the CA era was unpredicted student absence, which led to lunchtimes and after-school time being taken from teachers to supervise. 

Because of increased absence owing to Covid, more than one reserve day might be needed. Directed time would have to be set aside to avoid additional workload.

What about if candidates can take notes into the exam room? The most legitimate action would be to insist students go in with clean texts and have access to school internet devices only, with ChatGPT blocked. 

Students would produce bibliographies and their search histories could be monitored. This would ensure that students would conduct authentic searches, read relevant contextual and critical texts and plan without unauthorised assistance.

That again, though, would be hugely time intensive and cumbersome for schools to implement and oversee.

What about students with additional needs?

There is an argument that using ChatGPT can level the playing field for students who struggle with organisation in writing and have slower processing. 

But since extra time is permitted elsewhere, this could be added on as “special consideration”. Of course, that would cause concern about how long the school could stay open to accommodate 25 per cent extra time and mitigate student fatigue.

Would CA be organised into sessions to fit the timetable? If so, how would the work be stored between sessions? And how could students be prevented from using AI to devise a structure for use in the next session? 

Control might be needed over how material is taught before NEA titles are set. Currently, there is a range of practice in terms of preparation for texts, with some centres teaching one text lightly and giving students choice of four partner texts.  

Others provide completely free choice with consultation on the title.

English literature comparative essays are the most complex NEA assignments. As such, they exemplify the considerable logistical difficulties of room and IT allocation, staff supervision, timetabling and contingency planning associated with using controlled assessment as a way of eliminating malpractice of all kinds, including ChatGPT.

But to eliminate NEA completely would be to rob students of the educational opportunity to master the processes needed for successful long project completion, the skills prized by universities and employers. 

Yvonne Williams has been a head of English for 22 years and is chair of the National Association for the Teaching of English’s post-16 and higher education working group