History & Social Studies: Draft-and-critique loops

What you'll be able to do: Find out what your essay actually argues, as opposed to what you meant it to argue, and close the gap yourself, in repeated rounds, without ever letting the tool write the fix for you.

The gap you can't see from inside it

Ask a student what their essay argues and they'll tell you, fluently, immediately, and often accurately, about the essay in their head. The essay on the page is a different, usually weaker object, because writing a first draft mostly succeeds at getting your own thinking down for yourself, and that's a different job from making the argument legible to someone who doesn't already have it.

Composition researchers have a name for the gap: writer-based prose, text organised around the sequence in which the writer discovered their own ideas rather than around what a reader needs, in what order, to be persuaded. You cannot proofread your way out of it, because rereading your own draft, you supply the missing argument from memory without noticing you did it. Every sentence sounds like it says what you meant, because you're reading it with what you meant already loaded.

The fix needs an outside reader. That's expensive to get repeatedly from a person. It's cheap, and available every time you draft, from a tool that hasn't read your intentions, only your words.

The one question that does the diagnosis

Read my essay [paste the whole thing]. What is my argument, in one sentence, as it actually appears on the page, not what you think I probably meant?

Say it back to yourself before comparing: what did you intend to argue? Now look at the gap. It is almost never that the tool got it wrong. It's that the essay, read cold, makes a narrower, flatter, or subtly different claim than the one in your head, and a marker reads it exactly as cold as the tool did.

The rule that makes this a learning method rather than a rewriting service

It identifies. You write.

Ask for critique of the argument's structure, where the evidence doesn't support the stated claim, where a counter-case goes unaddressed, where the conclusion doesn't follow from what came before. Do not ask it to fix any of that. The moment it rewrites the paragraph, two things are lost: you no longer know whether you can close that kind of gap, and the next "what's my argument" extraction is now reading the tool's prose, not yours, which tells you nothing about your own writing.

Don't rewrite anything. Tell me: where does my evidence not actually support my stated claim? Where is there a strong counter-case I haven't addressed?
Where does the argument just restate the claim instead of building on it?
I'll revise it myself.

The loop, in order

1. Finished draft in. Not a plan, the actual prose.
2. Extract the one-sentence argument as it appears. Compare to your intention. Note the gap; don't explain it away.
3. Get structural critique: evidence, counter-case, logical flow. Prose quality is explicitly out of scope at this stage.
4. Revise the structure yourself.
5. Resubmit and re-extract. The second extraction is the real test: has the one-sentence argument moved toward what you intended, or just changed?
6. Repeat until the extracted sentence is the argument you meant to make.
7. Only then, move to prose-level critique, word choice, transitions, signposting. Polishing an argument you're about to restructure is wasted effort, which is why this comes last, after article 08's structural check and before any line-editing.

Worked shape

First extraction, on a WWI causation essay: "The alliance system caused the war." The student meant to argue a multi-causal, contingent case, alliance structure, mobilisation timetables and the July Crisis's specific failures, none of them sufficient alone. Nothing in that intention is visible on the page, because the essay's topic sentences all lead with the alliance system and never return to weigh it against the others.

Second round, after restructuring the topic sentences and adding a paragraph that explicitly weighs the three factors against each other: "The war resulted from several converging factors, of which the alliance system was necessary but not sufficient." That's the argument that was always intended and, for the first time, the one actually on the page.

Guarding against a false positive

If every extraction agrees with what you already believed, be suspicious before you're relieved. A tool that has just read your framing and reflects it back isn't confirming your argument is on the page, it may be pattern-matching to your own phrasing nearby in the text. Paste the essay into a fresh session with no other context, or ask a person who hasn't discussed it with you, before trusting an extraction that arrives too easily.

Across the subjects

History. As above.

Political science. Policy essays where the intended argument is "this policy failed for reason X" and the page argues, unintentionally, "this policy had several problems": a list standing in for a claim.

Sociology. Theory essays where the extracted sentence is a restatement of a theorist's position rather than the student's own evaluation of it.

Anthropology. Essays intending a methodological argument that, extracted, turn out to be a description of the fieldwork instead.

Geography. Evaluative "to what extent" essays where extraction reveals an unweighted list of factors rather than a position on which mattered most.

International relations. Essays on a crisis's causes, prone to the same gap as history, intended contingency, delivered inevitability.

Pitfalls

1. Letting it rewrite. The single failure that empties the method.
2. One loop and done. The informative divergence is usually the second or third, once the obvious problems are fixed.
3. Prose critique before structural critique. Wasted effort on sentences you're about to cut.
4. Trusting an extraction that flatters you. Check it with fresh eyes, yours or someone else's.
5. The tell: the extracted sentence stays the same loop after loop while your wording keeps changing. You're editing the surface; the argument underneath hasn't moved.

Try this today

Take a finished draft. Ask only for the one-sentence argument as it appears on the page.

Compare it to what you meant. Write down the gap in one sentence of your own (not the tool's rewrite, yours) and revise one paragraph to close it.