Undo fits this population better than confirmation dialogs

What it is

When motor misses are frequent, a ordinary confirmation dialog is not protection. Confirming is still a pointing act, a switch beat, or a spoken command—the same channel that already errs. Doing the action and taking it back with a large undo moves the gate from “aim again” to “is this result right?” That judgment is visual and cognitive, usually not the damaged motor link.

Why it happens

A confirm dialog parks the safeguard where error is still happening. The user just missed; the system immediately demands another hit of the same spec. OK and Cancel often sit side by side, so the next jolt can land on OK just as easily. A switch user pays another full scan for the dialog. Protection that is isomorphic with the original act does not cut error in proportion to the number of layers.

Undo parks the safeguard after a result exists. The user sees “not what I wanted” and hits a recovery control that can be large, stable, and kept away from the dangerous neighbour. What motor conditions share is that precise re-pointing is unreliable—not one tremor’s time curve. The fix is the structure of the safeguard, not another precision movement.

Studying it

Run one task under two safeguards: a pre-commit dialog versus a large post-commit undo. With motor-impaired participants, compare irreversible errors, time, and frustration—not whether they hit Confirm.

Independent variables: confirm size and spacing, undo size and dwell, whether irreversible acts still use a low-precision gate. Dependent variables: irreversible errors, second misses on the dialog, time to find undo.

Code second misses on the dialog separately. They show the safeguard copied the original problem, not that “nobody read the text.”

Where it stops holding

Content that has left the device, or a payment already cleared, cannot be saved by an undo window; a gate before entry remains. That gate must change channel or precision: a much larger target, far from Cancel, speech or dwell—not two ordinary buttons in a row. Undo also needs a visible result and enough time to notice; a delayed result or a flashing toast breaks the path. Keeping undo always on screen to reduce operation anxiety is a different causal story. Do not merge it with motor-error structure.

Applying it

• Default reversible acts to “do it, big undo.” Do not confirm every archive, move, or flag.
• If an irreversible act must be gated, rebuild the gate for a high miss rate: separate, enlarge, do not mirror Cancel.
• Verify: run error-prone tasks on both a confirm flow and an undo flow, and count only irreversible errors that remain. If the confirm group misses again on the dialog, do not add another dialog of the same kind.

Related

• Same group: J3.09.1 Motor impairments raise accidental activation well above typical rates · J3.09.2 Destructive actions that complete on a single activation carry the highest risk
• Nearby: A8.16.4 This population has a higher mis-tap rate, and undo beats confirmation · A11.03.5 A persistently visible undo affordance to reduce operation anxiety · J3.13.4 After a miss-tap, recovery should be one undo, not repeating the whole flow
• Search terms: undo versus confirmation · motor impairment · error recovery