Recognition difficulty and consequence severity should be designed independently, not treated as naturally bound

What it is

“This motion is hard, so use it for delete” binds recognition difficulty to consequence severity. The two should be designed independently: consequence from task risk, difficulty from learnability, comfort, and classifier separability. A hard motion is not necessarily hard to misrecognize—it may sit a millimetre from some easy everyday motion, so a high-consequence command is fired by a confusable trajectory. Difficulty is not insurance.

Why it happens

A recognizer sees distance from a template, not how effortful the user found it. An awkward crossed-hands pose can be, to the camera, only “two hands in front of the chest,” overlapping with smoothing a shirt. Effort also lowers repeat precision, inflates within-class variance, smears the decision boundary, and makes the high-consequence command easier to wander into. Conversely, an easy, morphologically unique short motion can be paired with a strict confirm flow; safety comes from the flow, not from soreness. Products that treat difficulty as insurance also lock people with less physical capacity out of high-consequence functions, an accessibility hole. Threshold bias can follow consequence; that is a cutting policy. It does not replace the structural rule “do not smuggle insurance into difficulty.”

Studying it

Have designers and users first rate candidate motions for “fits delete / payment” by intuition, then measure those motions’ confusion with everyday movement on the recognizer. Compare how much the “intuitively hard” and “measured unique” lists overlap. Bind the same high-consequence command to a hard-but-confusable motion and to an easy-but-unique motion plus confirm, and compare false fires and completability. Stratify by handedness, shoulder injury, and age, to test whether difficulty-as-insurance is locking some people out.

Where it stops holding

A ritual two-handed motion can itself mean “I am in earnest” in some cultures; fully decoupling then looks casual—that is a symbolic issue, and still wants a confirm rather than effort alone. On a toy vocabulary with huge between-class distance, difficulty and uniqueness correlate by accident and the coincidence will seem to support binding. Industrial gloves make every fine motion hard, so difficulty carries no safety information.

Applying it

• For each command, list a consequence tier and a reason for the motion choice. Forbid “this one is harder to do” as the only reason.
• Prefer morphologically unique, repeatable motions for high-consequence commands, and put safety in confirm or undo, not in soreness.
• In review, crash high-consequence templates into everyday-motion video. On a hit, change the motion; do not “add difficulty” by making it more awkward.

Related

• Same group: C4.27.2 High-consequence actions should not rely on a single one-shot recognition result · C4.27.3 Low-consequence, reversible actions can accept looser recognition thresholds · C4.27.4 When expanding the gesture vocabulary, re-check whether similar gestures map to different consequence levels
• Adjacent: C4.14 Gesture vocabulary size limits · C4.24 Recognition confidence and bias direction
• Search: gesture difficulty · consequence · false security