Learning brief
Decision: [ ]
Riskiest assumption: [ ]
Intended participant: [role/context]
Prototype fidelity needed: [concept / clickable / working flow / technical spike]
Neutral task: [goal without naming the button to press]
Evidence supporting continue/change/stop: [ ]
Build specification
Starting state: [ ]
Core steps: [ ]
One difficult moment: [ ]
Success state: [ ]
Real behavior: [ ]
Simulated behavior and fictional data: [ ]
Out of scope: [ ]
Session guide
- Explain the purpose, consent arrangements, and whether recording is involved.
- Ask about the participant's current process before showing the solution.
- Give the neutral task and observe without teaching the interface.
- Ask what they expected at moments of confusion.
- Ask what they believe happened and how they would recover from a mistake.
- Explain any intentional simulation or safe test error that needs debriefing.
Observation log
| Participant ID | Observed behavior | Exact quote if captured | Interpretation | Alternative explanation |
|---|---|---|---|---|
| [ ] | [ ] | [ ] | [ ] | [ ] |
Record recruitment limitations. Do not generalize a convenience sample to a population conversion rate.
Difficult-state checks
- Empty or missing input
- Invalid or contradictory information
- Loading and slow response
- Partial result or timeout
- Wrong or unsupported AI output
- Permission denied
- Edit, cancel, retry, and undo where appropriate
- Clear explanation of whether an external action happened
- Keyboard access, readable labels, and small-screen behavior
Readout
What we observed: [ ]
What we infer: [ ]
What remains unknown: [ ]
Decision and why: [ ]
Next test: [ ]
Production requirements not covered by this prototype: [security, permissions, persistence, monitoring, reliability, support, other]