Unit 4 · Lesson 16 of 20
Revise the interaction and retest
Trace a design revision to findings and state what the next test must resolve.
01 / Learn
The idea
A revision is a hypothesis expressed as a design. Link each changed behavior to a finding: what observed failure should it prevent or make recoverable? Preserve what worked. Do not combine so many changes that the team cannot learn which one helped. After revision, choose tasks that expose both the former failure and any new cost.
For voice systems, feedback should work in the modality the person is using. Naming the affected zone aloud can reveal a wrong action; a short correction path can let the person recover. Yet every added spoken phrase takes time and attention, so test routine commands as well as ambiguous ones. An initial positive result does not end iteration if noisy rooms, multiple users, or disabled audio were not represented.
A design change is justified because it follows a test. It is justified only when the finding and the proposed mechanism are connected and the new tradeoff is evaluated.
02 / See the reasoning
Worked example
Fictional scenario
A voice-controlled heating system says 'Okay' after 'turn off the heat,' but sometimes acts on the wrong zone.
A fictional prototype test finds that people in another room miss the display and believe 'Okay' confirms a whole-home action. An expert changes the response to 'Bedroom heating off' and supports 'Undo that' for ambiguous actions. They retest whether people identify the zone, correct a wrong one, and tire of extra speech during routine use.
Try a decision →03 / Decide
Guided decision
Users miss the on-screen zone and have no hands-free correction. Which next revision would you test?
Feedback on your choice
What could change the answer: If wrong-zone action has serious safety consequences, a confirmation before action may be warranted even at the cost of extra steps.
This feedback helps you examine the decision; it does not grade your written reasoning.
Apply it somewhere new →04 / Apply elsewhere
Independent application
Fictional scenario
A smartwatch accepts 'pause workout'; a fictional six-person test found three audio pauses instead of workout pauses, two noticed only after another minute, and four never looked at the display.
Revise the command, response, and correction behavior. Tie each change to a supplied finding, describe the new tradeoff, and propose a retest with tasks that could disprove the revision.
Make: A before-and-after interaction script plus a findings-to-change matrix and retest plan.
The revision follows specific observations, handles ambiguity without assuming a display, and the retest measures correction and added friction.
Unit complete
You completed test and improve.
You finished this unit's four independent tasks. The next unit builds on your judgment with a new capability.
05 / Keep working