Reversible actions let users learn by probing

What it is

A reversible action lets people return to a prior state after acting in a clear, reliable, acceptably costly way—undoing an edit, restoring a move, cancelling a draft, or recovering from trash. Reversibility lets people try an action, observe a result, and correct their understanding without being completely certain beforehand. It turns some learning from high-risk guessing into recoverable exploration.

Why it happens

Under uncertainty, people assess error consequence. When a result is visible, a recovery entry is near the current context, recovery scope is clear, and there are no hidden side effects, people are more willing to try and learn from feedback. If undo is absent, time-bounded unclearly, partial, or affects others, psychological risk stays high. Reversibility is thus not only a technical capability but whether people understand when, how, and what they can restore.

Studying it

In tasks containing uncertain action, observe willingness to explore, ability to find recovery, prediction of undo scope, and verification after restoration. Record probes, abandonment, error-recovery time, unintended loss, and trust in recovery. Test multistep operation, cross-device sync, collaborative editing, and asynchronous completion, because these often make the scope of “undo” ambiguous.

Where it stops holding

Not every action can be safely reversed: sent content, external transactions, regulatory records, seen publication, and physical processes may be remediable but not erasable. A false promise of undo is more dangerous than none. Even when technically reversible, undo should not replace clear feedforward, constraints, or error prevention; it is a safety net, not justification for ambiguous operation.

Applying it

• Provide adjacent, visible, result-checkable undo, redo, cancellation, or draft restoration for frequent probing and common error.
• State the object, scope, time limit, and nonrecoverable effects of restoration, distinguishing undo from remediation in collaborative or externally consequential cases.
• Test recovery on real failure paths, not normal flow alone, ensuring people can find and trust reversible entries under pressure.

Related

• Same group: B2.12.2 Irreversible actions must be explicitly marked · B2.12.3 Reversibility is a prerequisite for exploratory interfaces
• Nearby: B2.07.2 Feedforward reduces the cost of probing · B2.05.5 Constraints can make the correct action the only feasible action
• Search terms: reversible action · undo · error recovery