Moderators should not assist before a predefined intervention point
Aliases: moderator intervention · rescue protocol · unassisted testing
What it is
In evaluative usability testing, a moderator intervention protocol withholds path-changing help until predefined failure, request, or safety conditions. “Never help” is not absolute: formative diagnosis can add graded prompts after recording initial failure, while safety and participant welfare always take priority.
Why it happens
Assistance supplies cues absent from the interface, turning system failure into moderator success and reducing comparability. Total nonintervention can waste time, increase distress, or create harm. Timestamped graded prompts preserve unassisted performance while enabling observation of recovery.
Studying it
Prespecify silence, task repetition, generic encouragement, directional hint, and direct instruction plus stopping rules. Record first help request, intervention level, pre-intervention outcome, and assisted completion. Train moderators on one script and audit nonverbal cueing in recordings.
Where it stops holding
Teaching systems, support tools, and collaborative products may include real assistance, which should be tested as a system condition. Accessibility accommodations are not cheating. Safety, distress, and irreversible actions require immediate intervention.
Applying it
- Define a prompt ladder, time threshold, and safety stop before sessions.
- Record unassisted failure before applying the minimum necessary prompt.
- Separate unassisted, assisted, and incomplete outcomes.
- Review wording, gaze, and cursor movement for leaked direction.