Knowunity: A Self-Check Beyond the Score
Check right after revision
The brief left placement open. I put the activity in the study plan, immediately after the section revision, because that moment gives students a clear next step: check whether they can explain what they have just studied. It also gives the microphone request a reason, since students know what they are speaking for.
The flow stays simple: read a prompt, explain the concept, and receive written feedback. Students can use a hint, try again, or skip. A summary shows how they did and offers another round of practice.
Keep the summary honest
The main design challenge was encouraging students without overstating what they knew. An independent answer, an answer with a hint, and an answer repeated after seeing it should tell different stories.
During iteration, I found that the summary used the same wording after a first attempt and after practising again. A student could read an answer, repeat it, and receive a result that sounded as though they had known it all along.
I kept the score and changed the wording. The second summary now names the topics recalled during that later attempt. Keeping the score acknowledged real progress, and the new wording made clear that the answer was still fresh in memory.
Keep a typed path open
Speaking aloud is not always practical, so typing is available throughout the activity. Students who chose typing originally had to choose it again each time they returned to practise. I made the flow remember their input preference, so someone in a shared room can keep typing without interruption.
When microphone access fails, students can retry it or continue with text, so the activity never ends in a dead end.
Build to test transitions
I built a working prototype because transitions, such as feedback arriving and retries, were hard to judge in static screens.
I defined the flow before building with Claude Code, reused existing components, and documented component states in Storybook. An automated check flagged design-token violations, keeping new screens consistent with Knowunity’s design system. Human walkthroughs identified confusing moments, and separate AI reviews using a shared rubric added a second check.
The sprint delivered the prototype, a component library, and reusable AI workflow instructions.
