Destructive actions that complete on a single activation carry the highest risk

What it is

An action that commits on one pointer event and is hard to take back is the highest expected-loss case for motor error: delete, send, pay, call, overwrite. Motor noise decides whether the wrong target is hit. Single-activation commit decides whether anything is left afterwards. A high miss rate times irreversibility is a product, not a sum.

Why it happens

A discrete click quantizes continuous movement into one activation. There is no “halfway”: inside the rectangle, the whole command runs. If destruction is bound to that quantum, the motor channel has no protection—no second hit, no drag past a threshold, no dwell. The first event that lands is treated as a promise.

Expected loss is probability of a wrong activation times how irreversible the result is. The same miss that toggles a harmless sort is cheap; a sent mail or a charged payment is not. Giving a high-loss action the same one-click grammar as “mark read” drives the costliest state change with the cheapest input. That product is harshest where accidental activation is already frequent, and it is unsafe for everyone.

Studying it

Ship the same destructive function in two commit structures: fire on one activation versus require a second, distinct act. With participants who have unstable motor control, count irreversible errors—not whether the first tap hit the button.

Independent variables: activations required to commit, reversibility, distance between the destructive control and a high-frequency one. Dependent variables: irreversible error count, delay until the person notices, whether any recovery was still possible.

Hit-rate on the button will hide the bug. The button can be easy to hit. The mistake is settling the transaction on that one event.

Where it stops holding

Actions with instant, cheap reversal (switching a consequence-free view) should not be made heavier. True irreversibility (a transfer that has left the building) still justifies extra friction, but a second step that is the same-sized click only moved the risk. Bulk “select all, delete” destroys many objects with one activation; loss scales with count and is not “just one tap.”

Applying it

• Do not let delete, send, or pay mean “landed, done.” Either keep a short take-back window, or require a second act that is not the same click.
• Do not give a destructive control the same light tap as “open this row.” Opening is the row click; destruction lives in an explicit action menu.
• Verify: on a list with a destructive neighbour, ask someone with unstable motor control to open a named row ten times. Count how many times something was already sent, deleted, or paid. Any number above zero means the commit was too cheap.

Related

• Same group: J3.09.1 Motor impairments raise accidental activation well above typical rates · J3.09.3 Undo fits this population better than confirmation dialogs
• Nearby: J3.13 Pointer Cancellation · A10.08 Error Tolerance and Graceful Degradation
• Search terms: destructive action · single pointer activation · irreversible commit