PrettyJaded

Assess the Thinking, Not Just the Output: Reflection and Explain-Your-Choices Tasks in Canvas

4 min read

A polished piece of work no longer tells you much about how the student got there. Instead of asking whether AI was used, you can ask a better question: can the student explain their choices? This post shows how to add a short reasoning component to an open assessment in Canvas, without turning marking into policing.

Learning outcomes

By the end of this post you’ll be able to

  • Explain why reasoning can be a better signal than a polished output.
  • Design a short reasoning component for an open assessment.
  • Choose a validation method that suits the task.

Why assess the reasoning

In a HEPI guest post dated 9 October 2026, Roney Lima do Nascimento argues that the test of AI capability should not be whether a student can produce a polished output. It should be “whether they can explain why they accepted, rejected or modified what the system produced”. The post is mainly about a minimum AI learning entitlement, and readers’ comments question whether the proposal is proven.

Comparison of judging output only with judging output and reasoning
Look at the thinking too.

A Wonkhe article by Lisa Abrahams and Debbie McVitty, dated 4 October 2026, looks at validating learning. It is produced in partnership with Studiosity, so read it with that in mind. It names AI declarations and vivas as ways to validate learning, and notes the challenges: declarations can clash with a supportive approach, and vivas are hard to scale.

Three ways to add reasoning

You do not need all of these. Choose the lightest one that gives you the evidence you need.

Three cards for ways to add reasoning: an AI declaration, a decision log and a brief explanation
Pick the lightest option.
  • AI declaration. The student says which tool they used and what for. It suits small, named uses.
  • Decision log. The student records three or four moments where they kept, changed or rejected AI output, and why.
  • Brief explanation. Two short paragraphs on the key choices in the work, written in the student’s own words.

A short viva is a fourth option. It lets a student show ownership of the work, but it takes time, so keep it for tasks where it matters.

Put it into practice

Your turn

Choose one open assessment you will mark this term. Write the one reasoning prompt you would add, in a single sentence, for example “Describe one suggestion you rejected and why”.

Try it yourself

Which validation method fits?

Read each assessment and choose the reasoning component that suits it best. Each answer comes with a short reason.

Setting it up in Canvas

You can build this into an existing Canvas assignment. Keep the reasoning short, and mark it, so students see it matters. 5 Tips for Building Effective Rubrics in Canvas explains how to build a rubric.

Four steps for setting up a reasoning task in Canvas: the brief, submission types, a rubric row and feedback
Build it into the assignment.

Put it into practice

Try it in Canvas

  1. Open the assignment and select Edit.
  2. Under Submission Type, choose Online, then tick Text Entry and File Uploads.
  3. In the description, add your reasoning prompt and the word limit.
  4. Add a rubric row called “Explains choices”, then select Save.

When you mark, use SpeedGrader comments to respond to the reasoning, not just the result. How to Use SpeedGrader in Canvas and Why shows how.

Keep the tone supportive

The Wonkhe authors say unclear or inconsistent guidance on AI makes students anxious, and that students doubt institutions will trust their declarations and fear that declarations could expose gaps in their understanding.

Comparison of a policing tone with a supporting tone in a reasoning task
Choose the supporting tone.

Keep that in mind when you design the task. Say what the component is for. Mark the quality of the reasoning, not a confession. Do not treat a declaration as evidence of wrongdoing. Pair it with a clear statement of what is allowed; Tell Students Where They Stand on AI: An Assessment-by-Assessment Statement in Canvas shows how.

Check your understanding

Knowledge check

What test of AI capability does Lima do Nascimento propose?

Whether students can explain why they accepted, rejected or modified what the system produced.

Name two challenges the Wonkhe article gives for validation.

Declarations can clash with a supportive approach, and vivas are hard to scale.

How would you mark a decision log fairly?

By the quality of reasoning in a rubric row, not by whether AI was used.

Remember

Key takeaways

  • Ask whether students can explain their choices, not only whether AI was used.
  • Choose the lightest method: a declaration, a decision log, a brief explanation or a viva.
  • Mark the reasoning with a rubric row.
  • Say what the task is for, and keep the tone supportive.

Your next step

This week, pick one open assessment and add a one-sentence reasoning prompt to its Canvas brief. When you mark, notice whether the explanations tell you more than the finished work.

Reflect: what would you most like to know about how your students reached their answers?