← All posts

For tutors

Receipts: the AI's raw draft next to what I actually published

One lesson, eight drafted questions, and every edit I made before a student saw them, including the one I got wrong and a student caught.

Abbas Ahmadi · 9 min read

What this is, and what it isn't

This is the full edit log for one set of questions, from the draft as it arrived to the six questions a student actually sat. Nothing tidied up. Where I made a mistake, it's here, including the one a student found.

The lesson was the first and second conditional. "If it rains, we'll stay in" against "If it rained, we'd stay in". One hour, one adult learner, a two-page handout I've used for years. I asked for eight questions so I could cut to six.

I'm publishing this because "the AI drafts it, you approve it" sounds tidy and abstract. It's neither. It's four minutes of specific, slightly grumpy editing, and the only way to show what that means is to show it.

Check: What lesson were the questions for, and why did the tutor ask for eight questions rather than six?

The brief I wrote

Here's the line I gave with the handout, word for word:

Eight questions on the first and second conditional. Adult learner, intermediate. Two written answers. Target the mix-up between the two forms.

Decent, not great, and the gap shows up later. I didn't say that the learner is a Portuguese speaker, or that the mistake she actually makes is putting "would" inside the if-clause ("If I would have time"). The draft had no way to know that, so it didn't target it. Question seven below is the one I added to fix my own omission, not the draft's.

The lesson from the brief, which I relearn every few weeks: name the learner's actual mistake. The draft will go after what you name.

Check: What did the brief leave out, and what effect did that have on the draft?

Question 1: kept exactly as it came

Draft: Choose the correct form. "If it ___ tomorrow, we'll cancel the picnic." (a) rains (b) will rain (c) rained (d) would rain Published: Identical.

Good question. The stem gives a clear time ("tomorrow"), the right answer is (a), and the three wrong options are three real mistakes: the future in the if-clause, the second conditional's past, and "would" in the if-clause. That last one is my learner's own error, which the draft found on its own from the handout's examples. The explanation was also fine: "A real future possibility takes the present tense in the if-clause and 'will' in the main clause."

Three in ten questions are like this. I read them, nod, move on.

Check: Why was question 1 kept exactly as drafted?

Question 2: the stem I reworded

Draft: Which sentence is correct? Published: Which sentence uses the second conditional correctly?

Same four options. The draft's stem is a guess at what "correct" means, and one of the options was a perfectly correct first-conditional sentence, so a student could argue that two were "correct" and be right. Adding "the second conditional" to the stem fixes it with five words. The options didn't need touching.

This is the most common edit I make: not changing what is asked, but saying it.

Check: What was wrong with "Which sentence is correct?" and what did adding five words fix?

Question 3: the distractor that was true

Draft option (c): If I were you, I would take the job. Published option (c): If I would be you, I would take the job.

The question asked for the sentence with a mistake in it. Option (c) in the draft had no mistake. "If I were you" is textbook second conditional. A sharp student would pick (c) as correct, be told it was wrong, and be right to complain. I swapped it for the mistake I actually wanted to test, which is "would" in the if-clause.

This fault is sneaky because the draft's sentence is good English. The draft wasn't wrong about English. It was wrong about what the question needed the option to do.

Check: Why was the draft's option (c) a problem even though it was correct English?

Questions 4 and 5: the two I deleted

Draft Q4: According to the handout, the second conditional is used for ___ situations. Published: Deleted.

That's a question about my handout's wording. The answer is "unreal" or "imaginary" or "hypothetical" depending on which line of the page you remember, and it tests whether the student was looking at the page, not whether she can use the form.

Draft Q5: Rewrite in the third conditional: "If I had known, I would have come." Published: Deleted, kept for later.

Perfectly good question. Not today's lesson. The handout mentions the third conditional in one line at the bottom, and the draft, reasonably, assumed it was fair game. It wasn't, because I hadn't taught it. This is the context the draft can't have unless I give it, and I hadn't.

Check: Why was each of the two deleted questions removed?

Question 6: the expected answer I cut from 71 words to 18

Draft question: Explain the difference between the first and second conditional. Draft expected answer (71 words): The first conditional is used to talk about real and possible situations in the future, where the condition is likely to be fulfilled, and is formed with the present simple in the if-clause and "will" plus the base form in the main clause. The second conditional is used for hypothetical, unreal or unlikely situations in the present or future, formed with the past simple in the if-clause and "would" plus the base form in the main clause. Published expected answer (18 words): First conditional: a real possibility, present tense plus "will". Second conditional: an unreal or unlikely one, past tense plus "would".

The draft's answer is correct and would make a fine paragraph in a textbook. As a marking standard it's a trap, because the marker holds the student to it, and a student who says the right thing in two lines looks like she's missed most of it. The published version is what I'd accept from a student out loud. The marking follows what I write, so I write what I'd accept.

I rewrote both written expected answers this way. I always do. It's the edit with the biggest effect on the scores students see.

Check: Why was a correct 71-word expected answer a problem, and what did the tutor replace it with?

Question 7: the one I added

Draft: Nothing. This question didn't exist. Published: Write one sentence about your own work using the second conditional. Expected answer: Any sentence with "if" plus a past tense and "would" plus a verb, about her work. Example: "If I had a bigger team, I would finish faster."

The draft had no question that asked her to produce the form about her own life, and that's the question that tells me most. Recognising the second conditional in four options is one skill. Making one up, under mild pressure, about something real, is the skill the lesson was for.

I added it because my brief hadn't asked for it. That's on me, not the draft.

Check: Why did the tutor add a question the draft hadn't written?

Question 8: the one I got wrong

This is the wart.

Draft stem: Which sentence describes an unreal situation? My edit: Which sentence describes an imaginary situation?

I changed "unreal" to "imaginary" because it's the word I use in class. What I didn't notice is that one of the options was a first-conditional sentence about an unlikely but possible future ("If I win the lottery, I'll buy a house"). Under "unreal" it's clearly not the answer. Under "imaginary", my learner argued, winning the lottery is pretty imaginary. She picked it, was marked wrong, and told me why she disagreed.

She was right. My edit had made two options defensible. I put "unreal" back, which is the lesson's own word, and apologised. Her attempt is still in her record with that question marked wrong, because every attempt is kept and I wouldn't want it otherwise. The next time she sat the set, she got it.

The moral isn't "don't edit". It's that editing a stem means re-reading every option against the new stem, and I hadn't.

Check: What mistake did the tutor make in question 8, and what rule does the post draw from it?

The explanations: three of eight rewritten

I haven't shown the explanations question by question, so here's the pattern. Five were fine. Three restated the right answer without saying why the tempting wrong one was wrong.

Draft explanation for Q1: The correct answer is (a) because the first conditional uses the present simple. Published: A real future possibility takes the present tense in the if-clause: "if it rains". "Will" belongs in the main clause, not after "if".

The second version is read by a student who has just put "will" after "if", and it tells her exactly what she did. That line is teaching. It's worth thirty seconds each.

Check: What was wrong with the draft explanations that were rewritten?

The tally

  • Kept exactly as drafted: 3
  • Reworded the stem: 1
  • Replaced an option: 1
  • Deleted: 2
  • Added: 1
  • Expected answers rewritten: 2 of 2
  • Explanations rewritten: 3 of 8
  • Edits I got wrong: 1

Four minutes, by the clock on the wall. The thing to notice is that the faults are the same faults every time: expected answers too long, a distractor that's true elsewhere, a question about the page, something I didn't teach, explanations that restate. Once you know the five, the pass is fast, and the draft is doing the part that used to take me twenty minutes: writing eight plausible questions from my own material in the time it takes a student to finish an exercise.

Check: Which faults does the tutor say recur in every draft, and why does knowing them make the reading pass fast?

If you're in Australia

If your learners are preparing for a local English test or a school's own assessment, write the expected answers in the register that test rewards, and in Australian spelling where that's what's being produced. The draft follows the material you give it; if the handout is in one convention and your expected answers in another, the marking will wobble between them. Swap this section for your own country's conventions; nothing above depends on it.

Check: Why should the handout and the expected answers use the same spelling convention?

Try it with your own notes →

Comments

No comments yet. Be the first to say what you think.

Join the conversation

Sign in to comment and reply. Your comment shows with your name.

Sign in to comment