Teaching & Learning Center

We support, inspire, and celebrate teaching excellence and innovation at WCU.

Michael Rosario

2.3-Assessing process, not just product

AI microchip over Applied AI Literacy Field Guide over WCU Teaching and Learning Center logo on a tablet screen.
Field Guide Entry — Assessing process, not just product
A Field Guide to Applied AI Literacy
WCUPA Teaching & Learning Center
Entry 2.3
Designing for It
Designing for It

Assessing process, not just product

Breaking down a major assignment into discrete stages gives students more opportunities to demonstrate their thinking (and makes it more difficult for AI to be the sole author of the final product). This entry will help you apply a multi-stage process without greatly increasing your workload.

The concept

The biggest problem with only assessing a final product is that AI is generally good at making polished final products with minimal input. A staged submission process, however, shifts the assessment towards the work that is done before and during the creation of the final product. This allows students to demonstrate their thinking throughout the process.

One of the most revealing stages is the “revision note” where students write a short commentary about what has changed between drafts and why. Students who have genuinely thought through the draft and revision stages will have a relatively easy time completing this assignment, which can reveal a lot about their thought process. These intermediate check-ins often tell you more about a student’s work than a final product alone.

The four stages — what to collect, what it reveals, what it costs

Stage 1
Planning document
Light grading load
What to collect
Thesis or central claim, 3–5 supporting points, at least two sources identified with a note on why each is relevant.
What it reveals
Whether the student has read enough to have a position. A planning document built from real reading looks different from one built from AI’s summary of a topic.
AI note
AI produces planning documents instantly. This stage is most useful when paired with a personal anchor or a source the student had to locate themselves.
Stage 2
Working draft with marked questions
Light grading load
What to collect
A rough draft — not polished — with 2–3 places where the student has marked a question, uncertainty, or decision they haven’t resolved yet.
What it reveals
Unresolved thinking is the evidence that real thinking happened. Students who used AI for the whole draft often can’t identify a genuine uncertainty — the draft has no rough edges.
AI note
AI drafts are smooth and confident. Requiring marked questions creates friction: either the student finds real ones, or the absence of rough edges is itself informative.
Stage 3 — Most revealing
Revision note
Light grading load
What to collect
A short written account (150–250 words) of what changed between the draft and the final, and why. What did feedback, reflection, or further reading cause the student to reconsider?
What it reveals
Real revision requires having made a decision. A student who genuinely revised something can describe the before, the trigger for change, and the after. A student who didn’t cannot construct this convincingly.
AI note
This is where AI-dependent work breaks down most visibly. AI can generate a revision note, but it won’t match the specific changes between the student’s actual draft and final — and that mismatch is detectable.
Stage 4
Final submission
Full grading load
What to collect
The polished final work, submitted alongside all prior stages. Evaluating the final in light of the process is more informative than evaluating it in isolation.
What it reveals
When read alongside the draft and revision note, a final product tells you whether the student developed across the assignment or just produced something polished from the start.
AI note
A final that looks dramatically better than the draft with a vague revision note is a flag worth examining — development that outpaces the described revision process is unusual.

The revision note — why it matters most

Writing about what you changed requires having made a change

The revision note is the hardest part of this process for AI-dependent students to fake convincingly — not because AI can’t write one, but because a convincing revision note has to match the specific content of the student’s own draft. A generic AI-generated revision note (“I clarified my thesis and strengthened my evidence”) won’t describe the actual changes between the two drafts in your inbox. That mismatch is usually detectable in under two minutes.

It also functions as an independent writing sample. A student whose final draft is polished and whose revision note is vague, generic, or inconsistent with the changes you can observe has told you something important.

Revision note prompt — give this to students
“In 150–250 words, describe what you changed between your draft and your final submission — and why. Be specific: name at least one passage or argument you revised, explain what caused you to reconsider it (feedback, further reading, your own rereading), and describe what the change was. ‘I improved my writing’ is not sufficient. ‘I cut my third paragraph because it repeated the claim in my introduction and added a counterargument I hadn’t addressed’ is.”

How to set this up

1

Choose one major assignment to pilot this semester

There’s no need to redo every assessment in the course. Just start with an assignment where you think it would be most helpful to see the student’s thinking process.

2

Set due dates at least 48 hours apart

For example, planning document due Monday, draft due Wednesday, revision note and final due Sunday. The gaps force actual time between stages, which is where real revision happens.

Tip: Make the planning document due early in the week when you can give brief, quick feedback. Students who get a response to their planning document are significantly more likely to produce a draft that shows genuine engagement.
3

Tell students explicitly what each stage is for

Students typically increase their engagement when they understand why you’re asking them to do things a certain way. Make it clear that you are using these different stages so that you can provide feedback quickly when it is most useful.

4

Read stages 1–3 looking for signals, not substance

Remember that you’re not evaluating these intermediate stages as final products. You’re trying to evaluate whether or not there is real evidence of thought or wether the intermediate outputs appear disconnected.

AI prompt — design staged submissions for your assignment

Step 1 — design the four stages
I teach a course on [course topic]. I have a major assignment that currently asks students to [describe the assignment]. I want to redesign this as a four-stage submission: planning document, working draft with marked questions, revision note, and final. For each stage, tell me: (1) exactly what to ask students to submit, (2) a specific prompt I could give them, and (3) what I should look for when I review it — keeping in mind that stages 1–3 are engagement checks, not full evaluations.
Step 2 — write the revision note prompt
Write a revision note prompt for this assignment that: (1) asks students to name something specific they changed, (2) requires them to explain what caused the change, and (3) is explicit that generic statements like “I improved my writing” won’t satisfy it. The prompt should be firm but not accusatory in tone — framed as helping students reflect, not as a lie-detector test.

Caveats

  • Staged submission only creates accountability if you actually look at the stages. If students learn that planning documents and drafts go unread, they’ll stop engaging with them. A brief, even one-sentence acknowledgment of each stage goes a long way.
  • AI can generate a plausible revision note for a student who submits AI-generated drafts at every stage — especially if the student instructs AI to “write a revision note explaining how I improved my thesis and added a counterargument.” The check is whether the revision note is specific enough to be verifiable against the actual changes in the two documents.
  • Students with accessibility accommodations that affect writing speed or executive function may struggle with multiple deadlines. Build in flexibility for submission timing — the goal is the evidence of process, not compliance with a specific schedule.
  • This approach works best for assignments long enough to actually require revision — papers, projects, analyses. Applying it to short responses or weekly reflections creates unnecessary overhead without proportionate benefit.

Reflection questions

  1. Which of your major assignments would benefit most from seeing the student’s thinking in progress — not just the finished product?
  2. Have you ever received a final paper that seemed dramatically better or worse than the student’s other work? What would a draft and revision note have told you about that assignment?
  3. What would you look for in a revision note to distinguish genuine reflection from a description that could apply to any paper on any topic?
  4. How would you frame staged submission to students so it feels like support rather than surveillance?

Categories:

Tags:

,