Protocol
Give feedback at the smallest actionable unit
'Needs improvement' names a direction; useful feedback names the next controllable change.
When it fits
- When feedback is accurate but too broad to guide the next attempt.
When to avoid it
- Not every performance problem has one local cause. Escalate to prerequisite, environment or task-design issues when the same correction repeatedly fails.
Why it matters
Locate the earliest decision, behavior or output feature that materially affected the result. State what was observed, why it matters for the criterion, and what the learner should try on the next case. Keep identity judgments and unrelated preferences out of the correction.
Steps
- The learner can execute the feedback on the next attempt without asking what 'better' means.
An example
Instead of 'be more confident,' feedback says the recommendation appeared after three minutes of background; lead with the decision and evidence, then answer detail questions.
Check your result
The learner can execute the feedback on the next attempt without asking what 'better' means.
Keep this limit in mind
- Not every performance problem has one local cause. Escalate to prerequisite, environment or task-design issues when the same correction repeatedly fails.
Connected ideas
Use beforeMake feedback end in another attempt
Evidence and sources
A meta-analysis of 435 feedback studies found a medium average effect on learning (d = 0.48) with substantial heterogeneity, and the information content of feedback materially moderated impact.
More feedback is not automatically better, and effects differ by outcome and feedback form.
The Power of Feedback Revisited: A Meta-Analysis of Educational Feedback Research · Abstract