The Borrowed Essay

About a year ago a student posted a short message on The Student Room, a website where students ask each other for help. This is the whole message, word for word, except for two spellings we have corrected.

“How do I get over my ChatGPT addiction. I rely on it too heavily to write all my essays and when I actually have to write essays under exam conditions, I literally can’t”^[1]

The strange word in that message is “essays”. This student had plenty of essays, all of them finished and all with their own name at the top. Yet under exam conditions they could not write one.

Picture the exam room for a moment. The front page of AQA’s GCSE English Literature Paper 1 gives the instructions: “Use black ink or black ball-point pen. Do not use pencil.” A few lines further down it adds, “You must not use a dictionary.”^[2] Edexcel’s International GCSE English Literature Paper 1 prints a similar rule: “Dictionaries may not be used in this examination.”^[5] On the desk there is a question the student has never seen, an empty answer book and a clock. Before the first sentence can be written, someone has to decide what the essay will argue, which parts of the text will prove it, and in what order they should come. In that room, who is going to decide?

A strong tool with a limit

Start with what chatbots are good at, which is a great deal. Chatbots such as ChatGPT and Claude write computer code, operate a computer, and search the internet, gathering what they find far faster than a person can.

Chatbots can help with an essay, too, up to a point. Give a chatbot the question and the mark scheme, and it will give you an essay structure. There is no guarantee, though, that it will be a good one. A good structure is one the student can use again for different questions and different texts, and a chatbot may build one for that single question instead. A new structure for every question means more for the student to remember in the exam. One day it may suggest five paragraphs, and the next day seven or ten, so the student has no way to know which one to aim for.

A chatbot can look at an essay from a technical point of view. It has never marked an exam or taught a student through one. It may even produce an essay that could score a 9, yet a 9 on the page does not mean a student could repeat it in the exam hall.

What we keep finding, however many criteria we give it, is that the writing inside the structure still does not read like a person’s. We know this because we tried it ourselves, on an essay our own students now read.

The draft that needed about 96 revisions

This year we wrote a model essay on Mrs Birling in An Inspector Calls for our library. We began with a chatbot, gave it the criteria and asked it for a draft. When we saw how weak that draft was, we essentially redid it ourselves, over about 96 revisions.

Most of those revisions were small. One of them changed what the essay was about. In an early draft, the first paragraph of analysis was built on the moment when Mrs Birling scolds her daughter for using a slang word. Her line is “What an expression, Sheila!” The paragraph argued that her prejudice shows first in the way she polices how people speak. We decided that idea was weak, and no amount of polishing the sentences could have made it strong.

So we moved the paragraph onto a different quotation. Early in Act One, Sheila teases Gerald about “all last summer, when you never came near me”, and her mother replies that “men with important work to do sometimes have to spend nearly all their time and energy on their business.” The audience later learns that Gerald’s “important work” was an affair. With that line at its centre, the paragraph could argue something far stronger: Mrs Birling will ask her own daughter to accept being neglected, in order to protect a social order that rewards the family.

The chatbot did not find that idea. It was ours, and in the end we wrote the paragraph ourselves. Its first sentence now reads:

“A first instance of Mrs Birling’s prejudice is, ironically, directed towards her own daughter when she encourages Sheila to accept Gerald’s infidelity as the price for the family’s devotion to protecting and climbing the social order.”

By the end, the essay was ours: its ideas, its argument and most of its wording. When we added up the time the corrections had taken, we could have written the whole essay ourselves from the start.

Every revision was a judgement

Look again at what that one change needed. To reject the first paragraph, someone had to see that its idea was weak and understand why. To replace it, someone had to know the play well enough to find a better line, and then see what that line would allow the paragraph to argue. Each of those steps is a judgement about an argument and its evidence. We made every one of them, while the chatbot waited for its next instruction.

Those judgements are the skill an English exam tests. AQA marks this kind of GCSE Literature essay in levels, and its top level is called Level 6. At the top of that level, AQA says, the response “is likely to be a critical, exploratory, well-structured argument.”^[3] A chatbot can supply the “well-structured” part in seconds, but in the exam hall that is exactly the part the student must build alone.

Other boards use their own words for the top of the mark scheme. Edexcel GCSE’s Level 5 says: “A critical style is developed with maturity”.^[6] OCR’s Level 6 is headed “Sustained critical style in an informed personal response to both text and task”.^[7] Eduqas’s Band 5 asks students to “convey ideas with consistent coherence”.^[8] All three describe the response that the student writes.

The rules for the exam hall come from the Joint Council for Qualifications, which publishes the regulations that exam centres follow. It says students sit most of their exams “under close staff supervision with limited access to authorised materials and no permitted access to the internet.” It adds that students must not be able to use chatbots and similar tools during those exams.^[4]

So when a chatbot writes a student’s essays, the student receives the finished essays and misses the decisions inside them. Reading a good essay can feel very like being able to write one, especially when your own name is at the top of it. We think that feeling is the trap, because in the exam hall the decisions are the one part that nobody else can make for you.

The job decides the effect

So what should the student on the forum do? A chatbot can be a powerful tool, but what it does to a student depends on the job it is given, because the work still has to be done by the student’s own brain. A student who uses a chatbot to make the challenge easier, and to skip the part where their brain has to struggle, is handing the thinking to something else. It feels like help at the time. In the exam hall, where nothing else can do the thinking, it becomes exactly the struggle that student described.

The other way to use a chatbot is to extend ideas that are already the student’s own. It can test a student on the order of events in a play, ask them questions about a character, or explain an old word they have not met before. Each of those still leaves the thinking with the student, and anything a chatbot says about the text should be checked against the book itself.

How we keep the thinking with the student

This is how we work with our own students. They write each essay completely by themselves first. We then assess that writing to find exactly where its strengths and weaknesses are. Only then do they redraft it, and during the redraft they are asked questions rather than given answers. Which quotation have you chosen, and why? What idea does it convey, and which technique carries that idea? Do your three quotations come from the beginning, the middle and the end of the text? If an idea looks weak, the student is asked whether a better quotation might carry it.

Those questions help a student check whether the essay meets the top level, and whether their own ideas could be stronger. What the questions never do is make the decision for the student. Choosing the quotations, structuring the paragraphs and understanding what the examiner rewards are all decisions the student has to make. Nobody can hand that thinking to a chatbot for two years and then expect a great result in the exam.

A skill, not only a grade

Our aim at Sophicly is mastery. We want every student to leave us able to structure an argument and to analyse evidence, because those two skills are what A level, university and most careers will go on asking of them, long after the exams are over. The grade is how the exams measure those skills. A chatbot can hand a student something that looks like the evidence for a grade, such as a folder of finished essays, with none of the skill underneath it.

Four things to do this week

1. Ask to see the plan before the essay. Next time your child has an essay to write, ask them to show you the plan first, on paper: what the essay will argue, and which quotations will prove it. If they can explain the plan to you without a screen, it belongs to them.

2. Ask “Why this quotation?” Pick one paragraph and ask why your child chose that quotation and not another. Then ask what the paragraph would argue if it used a different one. That second question is the one that rebuilt our Mrs Birling paragraph.

3. Time one answer by hand. Choose a real exam question, put the phone in another room, and give your child a pen, paper and the time the exam allows. Whatever appears on the page is what they can do on the day.

4. Let a chatbot ask the questions, and never write the answers. If your child uses one, suggest they ask it to test them, for example with questions about a character. The person who writes the answers is the person who builds the skill.

If you would like to know where your child’s own writing stands, our free assessment takes five minutes.

Sources