Why I read every question before a student sees it (and what I change)

The two-minute rule

Every set of questions my students see has been read by me first. All of it, every time, in a two- or three-minute pass. I've kept that rule through a year of being told it isn't necessary, and I've become more sure of it, not less.

Here's the stance in one line: a question is a promise to the student that it is answerable, fair and about what they were taught. A draft can't make that promise. Only the person who taught the lesson can. So the draft is a draft, however good it is, until I've read it.

This isn't a complaint about the drafts. They're good. Most questions I read, I keep. But "most" is doing a lot of work in that sentence, and the ones I change are exactly the ones that would have cost me a student's trust.

Check: What is the one-line stance the post takes on who makes a question's promise to the student?

The mark is the AI's; the question is mine

I want to be precise about the division of labour, because it's the part people argue with.

What I own: what gets asked, how it's worded, what the options are, what the right answer is, and, for a written answer, what counts as right. That last one is the expected answer I write while approving. It's the standard the marking follows.

What the machine owns: applying that standard, the same way, to every student, every time.

What I don't do is change a mark after it's given. If a student's written answer scores six and I think it deserved eight, the problem is upstream: my expected answer was too narrow, or the question invited an answer I hadn't allowed for. The fix is to widen the expected answer so the next student is marked the way I meant, not to hand-adjust this one. I resisted this for a month. Then I noticed that the students who used to "improve" when they moved to a softer tutor had stopped improving, because the marker didn't care whose student they were. I'd been the softer tutor. That was the end of my objection.

Check: What does the tutor own, what does the marking own, and what does the tutor do when a mark seems wrong?

What a draft gets wrong, and why that isn't stupid

A drafted question comes from the material you gave it and the brief you wrote. It doesn't know which half of the handout you spent forty minutes on and which half you skipped. It doesn't know that your students are Spanish speakers who put the present perfect where a finished time goes, unless you told it. It doesn't know that example three on the page has a typo, and it will cheerfully build a question on the typo.

None of that is a failure of intelligence. It's a failure of context, and context is the thing the tutor has and the draft doesn't. A good brief closes most of the gap. The reading pass closes the rest.

So the edits below aren't corrections of a bad writer. They're the adjustments any writer's draft needs from the one person who was in the room.

Check: What does a drafted question lack that the tutor has, and what two things close that gap?

Edits 1 to 3: the question itself

1. An ambiguous stem. "Which of these is correct?" with no context is a guess at what "correct" means. I add the context: "Which sentence uses the past simple correctly?" One clause, and the question now has one reading.

2. Two defensible answers. The most common serious fault, and the one students will catch. Usually one option is right in the lesson's terms and another is right in a wider sense. I either change the second option so it's clearly wrong, or I rewrite the stem so only one fits.

3. Negative stems and "all of the above". "Which is NOT an example of..." tests reading more than knowing, and "all of the above" rewards the student who notices two right answers and stops. I rewrite the negative as a positive and replace "all of the above" with a real option.

Check: What are the three faults in a question's stem that the tutor fixes first, and how is each fixed?

Edits 4 to 6: the options

4. A distractor that is true somewhere else. The wrong option that would be right in a different lesson. "He has worked here since 2019" is a fine sentence; if it's a distractor in a question about finished time, a sharp student will defend it and be right to. I swap it for something that is actually wrong.

5. The long option is the right one. Drafts, like tired tutors, write the correct answer carefully and the wrong ones quickly. Students learn this within a fortnight. I trim the right answer or lengthen a wrong one until they're the same shape.

6. Distractors nobody would pick. A wrong option no student has ever chosen is dead weight. The best distractors are the mistakes your students actually make. I keep a short list of those per topic, and I write them in. That's where the brief earns its keep: if I've named the mistakes, the draft usually targets them already.

Check: What makes a good distractor, and what are two common faults in drafted options?

Edits 7 and 8: the written answer

7. An expected answer that's too long. The draft writes a model paragraph. The marking holds the student to it, so a student who says the right thing in fewer words scores low. I cut the expected answer to the idea, in plain words, two lines at most, in the language the questions are in. The marking follows what I write, so I write what I mean.

8. A question that invites one word. "What tense is this?" is a written question in form and a choice question in fact. I turn it into something that needs production: "Rewrite this sentence so it's correct, and say why." Now the written answer is doing what only a written answer can do.

I'll say it plainly because it took me a while to see it: for a written question, the expected answer is the question. Spend your editing time there.

Check: Why does the length of the expected answer matter, and what does the tutor change a one-word written question into?

Edit 9: the explanation

Every question carries a line of explanation that the student reads after answering. Drafts tend to write it as a restatement of the right answer. That's a lost opportunity.

A good explanation says why the right answer is right and why the tempting wrong one is wrong. "Yesterday is a finished time, so the past simple. 'Have seen' would need a time that reaches up to now." That line is teaching, and it's read at the exact moment the student is most ready to learn: just after getting it wrong. I rewrite about a third of the explanations, and it's the edit I'd defend hardest.

Check: What should a good explanation do beyond restating the answer, and when is it read?

What I delete outright

Some questions don't get edited. They go.

• Questions about the page, not the idea. "What is the title of section two?" tests whether the student was looking at the handout. Gone.
• Things I didn't teach. A good question on next week's topic is next week's question. I keep it aside and delete it from today.
• Trivia. A date, a name, a number that happened to be in the material and has nothing to do with the point.
• Questions that need outside knowledge. If the answer depends on something the student would have to know from somewhere else, it isn't a check of this lesson.
• Loaded examples. A sentence about a situation that lands differently for different students. There are a thousand neutral examples; use one.

Check: Name three kinds of question the tutor deletes rather than edits.

What I don't change

Honesty requires this section. Most questions survive the pass untouched. I'd guess seven in ten.

I don't rewrite questions to my taste. If the wording is fine and the answer is right, it stays, even if I'd have phrased it differently. I don't reorder for the sake of it. I don't make everything harder, which is a habit I had to break: the check is meant to show what landed, not to prove the tutor is clever.

The test I apply is simple. If a student questioned this question, would I be embarrassed? If no, leave it. The two minutes are for the ones where the answer is yes.

Check: What test does the tutor use to decide whether to leave a question alone?

Why "it's just a check" is the wrong argument

The argument against reading every question is that the stakes are low. It's a ten-minute check, not an exam.

The stakes for the grade are low. The stakes for trust are not. A student who finds a wrong answer marked as right learns two things: that the check can't be trusted, and that the tutor didn't look. They stop reading the explanations, because why would you learn from something that's wrong. And because every attempt is kept, the bad question sits in their record, forever, scoring them on a mistake that wasn't theirs.

Three minutes buys all of that back. It's the cheapest thing I do all day.

Check: What two things does a student learn from a wrong question marked as right, and why does it matter that attempts are kept?

If you're in Australia

Two small local habits. Write expected answers in Australian spelling if that's what your students are being taught to produce, because a marking standard in one spelling and students writing in another is a gap you created, not them. And where a question touches a subject name, use the one the student's school uses, so the check reads as part of their week rather than something imported. Swap this section for your own country's conventions; nothing above depends on it.

Check: Why should the expected answer be written in the spelling students are taught to use?

Try it with your own notes →