Adjacent targets with very different consequences need extra spacing or isolation
Aliases: risk isolation · target spacing · error consequence
What it is
The more different neighbouring actions' consequences are, the more extra spacing or isolation they need. Delete versus keep, pay versus cancel, and send versus draft should not be arranged by generic density alone; loss and recoverability change the needed protection.
Why it happens
Touch error cannot be eliminated — that is a starting premise, not something to be solved. What can actually be controlled is how much harm results when an error lands on a neighbour of a given consequence. A mistake landing on a similarly reversible neighbour (tapping the wrong filter tab, say) usually costs one more tap to fix; one landing on an irreversible or expensive action (triggering delete or a payment by mistake) can cost far more than "tap again." This means spacing here is not doing simple false-tap prevention — it is allocating defensive resources according to how asymmetric the consequences are. Physical spacing is only the most basic layer; visual grouping, confirmation, and delayed execution are further layers stacked on top of it, and together they determine whether a false tap can turn into real harm — not whether the false tap happens in the first place, since the latter is nearly impossible to eliminate while the former is the achievable target.
Studying it
Test adjacent layouts by consequence and recoverability, recording false taps, recovery success, anxiety, and time. Do not inspect total errors only: the same rate is entirely differently acceptable for delete versus view switching, so results need to be reported stratified by severity of consequence rather than pooled into one generic false-tap rate. One concrete approach measures two separate rates: "a false tap occurs" and "a false tap turns into unrecoverable harm." A well-isolated design should tolerate the former while pushing the latter toward zero; if the two rates are nearly equal, the layered defenses are not actually intercepting harm after a false tap happens.
Where it stops holding
Isolation should not create needless confirmation fatigue or hide frequent safe work. Low-risk reversible commands can remain compact; high-risk commands need not rely only on whitespace, as labels, delay, and undo also protect. Simply widening physical spacing has limited value for actions that already have undo: if a clear, timely undo path exists after an accidental delete, pushing spacing wider than usability requires can cost more — through a looser layout and more distant everyday actions — than the extra protection it buys.
Applying it
- Move destructive or irreversible actions away from frequent safe actions and leave real buffer.
- Use grouping, labels beyond colour, confirmation, or undo as layered protection, treating spacing as one layer among several rather than the only one.
- How to check: measure the false-tap rate and the rate at which a false tap turns into unrecoverable harm separately; if the two are close together, the defenses beyond spacing — confirmation, undo, delayed execution — are not actually doing their job, and those are what need reinforcing rather than widening the spacing further.