Teaching needs demonstration and live correction
Aliases: mirror feedback · live comparison · gesture coaching · augmented feedback
What it is
The effective loop for teaching a motion gesture is "watch the demonstration → perform → immediately see the gap between your body and the target → adjust and retry." The demonstration tells the player what the target is; live correction (mirror feedback) tells them where they currently are and how far off. Without the second half, players know only "pass/fail"—not which hand was too low, which joint never arrived—and learning degrades into blind trial and error.
Why it happens
Motor learning runs on a closed loop built from error signals: the nervous system revises its next output by comparing the intended movement against actual feedback, and the more specific the error signal, the more precise the correction. "Success/failure" is a binary signal carrying one bit; a skeleton overlay or deviation highlight is a continuous signal indicating the error's direction and magnitude. Proprioception provides internal sensing during execution, but its precision for absolute spatial position is limited—a player feels their hand at shoulder height when it is actually ten centimetres lower, and only external visual feedback reveals that gap. Immediacy is the loop's other condition: feedback delayed beyond a second arrives after the player has moved on to the next action, leaving the error signal without a correction target.
Where it stops holding
Live correction does not mean constant judgement. Synchronous scoring or deviation readouts during execution consume attention and disrupt the movement's own fluency—the "micro-managed performance" problem. Correction should arrive after the action completes (terminal feedback) or at key checkpoints, not every frame. Correction granularity must also match the learning stage: a beginner facing ten deviations cannot fix them all at once, so effective feedback names only the largest one per attempt; advanced players need not "what was wrong" but "how close," so granularity should refine as progress accumulates. Skeleton overlays carry their own cognitive cost for some players—identifying one's own body inside a mirrored rendering can confuse players with weaker spatial ability rather than help them.
Applying it
- After each attempt, immediately show a comparison view (the player's skeleton overlaid on or beside the reference skeleton), highlight the single most deviating body part, and attach one correction cue ("raise your hand a little higher").
- Adjust feedback granularity to progress automatically: beginners see one deviation per attempt; after three consecutive successes, allow two.
- Verification: compare learning speed (attempts needed to reach stable execution) between a pass/fail-only build and a comparison-feedback build. If the gap is small, check whether feedback latency exceeds one second.