On marking, machines, and the dog
In which a grieving man defends a machine, at greater length than he intended, and does not apologise for it.
Barnaby died three weeks ago on Tuesday. He was a lurcher, mostly, with some opinion in him about what else he might have been, and he had the run of my study for eleven years. I mention him at the top of an essay about marking software because he is the reason I am writing it, and because I have reached the stage of the night where the connective tissue between things feels obvious to me and will not, I suspect, survive the morning.
Here is the connection, while I can still find it. For eleven years, Barnaby lay under the desk while I marked. Thirty scripts a go, four times a term, across a career I am no longer young enough to describe as promising. He would sigh around the ninth one. I have come to understand that sigh as the most accurate assessment of secondary marking anyone has ever produced.
The guilt, and where it comes from
I have listened to colleagues talk about these marking tools the way one talks about a cousin who has done something unmentionable at a wedding. There is a lowering of the voice. There is the phrase I suppose it saves time, offered the way one offers a cigarette to a man on his way to the gallows.
The guilt, I think, comes from a category error, and I want to name it plainly, because it is late and I have run out of the energy for tact. We have confused the labour of marking with the judgement of teaching. They are not the same thing. They have never been the same thing. The labour is the transcription of a judgement into a form a fifteen-year-old can act upon, at eleven o'clock at night, for the twenty-third time, while a lurcher sighs.
The judgement is the thing you spent a decade acquiring. The labour is the thing that stops you using it.
What the ninth script costs
Let me tell you the honest thing that nobody puts in the prospectus. By the ninth script I am not marking. I am pattern-matching against the eight that came before and a memory of the mark scheme that has begun to soften at the edges. By the twentieth I am generous, because generosity is faster. By the twenty-eighth I am generous and vague, and a child who wrote something genuinely strange and good on page three receives four words about paragraphing.
This is not a confession. This is arithmetic. Every marker who tells you otherwise is either lying or has not marked a full set since the Blair government. The examiner's tolerance exists precisely because we all drift, and we drift in one direction, and it is downhill.
So when someone tells me that a machine marking a script is an insult to the profession, my answer is: compared to what? Compared to the marking I did at half past eleven on the Sunday before an inspection? That is not the comparison being offered. The comparison being offered is against an idealised marker who does not get tired and does not have a dog who has recently and permanently stopped sighing under the desk.
What it is actually good for
Not the number. I want to be clear about this, because it is where the enthusiasts and I part company. The mark a machine gives you is a second opinion from a colleague who has read the scheme very carefully and has never met the child. That is genuinely useful and it is not the same as being right.
What it is good for is the other thing, the thing that dies first when you are tired: the specific sentence about the specific line. Twenty-eight children each receiving a comment that could only have been written about their own paragraph. I cannot do that at script twenty-eight. I could do it at script three. The machine can do it at script twenty-eight because the machine does not know what number it is on, which is the single most useful property it has.
And then — and this is the part that matters — you disagree with it. You open the thing, you read the mark, you say "no, that's a fourteen, it's doing something with the caesura that you've missed," and you change it. You have now done the part of marking that is teaching, and skipped the part that is transcription. That is the whole trick. There is no other trick.
The objection I take seriously
There is one, and I will not pretend otherwise. If you stop reading the scripts, you stop knowing your class. The feedback goes out, the marks go in the book, and six weeks later you discover that nobody in the room understands what a dramatic monologue is, because the machine told each of them individually and nobody told you.
So: read them. Read the reports, skim the scripts, look at the shape of the thing. What you are buying is not the removal of your attention but its redeployment — from the eleventh transcription of "embed your quotations" to the question of why eleven of them needed telling.
Anyone who sells you the other thing is selling you something worse.
Barnaby
He was not a clever dog. I want to be honest about that too. He once spent a full afternoon in dispute with a watering can. But he was present for something like four thousand marked essays, which is more than most heads of department, and if he had a view it was that the ninth one takes something out of you that the first one does not.
I have marked one set since he died. I did it at the kitchen table because the study was unbearable. It took five hours and I do not believe for a moment that the child who came twenty-eighth got what she was owed.
Next time I am going to let the machine do the transcription of my judgement, and I am going to spend the time I save reading what they actually wrote, and I am going to feel precisely no guilt about it whatsoever. You shouldn't either. The guilt is not a sign of integrity. It is a sign that you have confused the hard part with the long part, and they were never the same, and a dog worked that out before either of us did.
— Written in one sitting, which will be obvious, and posted before better judgement could intervene.