Past a threshold, hints are ignored as a class
Aliases: tip fatigue · banner blindness · overlay habituation
What it is
Dots, bottom bars, first-visit bubbles, “what’s new” strips are hints: the system inserts a teaching or promotional sentence into work already underway. When the count in a stretch of time crosses a threshold, people stop judging content per item and ignore the class. Restraint here is the dose of hints as a stimulus class inside one product; past the threshold the whole class dies, not only the last item. It is not how several teams rank claims on one quota, and not whether progressive intros need a per-session cap—those are how a quota is shared. This is the dose poisoning the channel itself.
Why it happens
Repeated, fixed-place, similar-skinned stimuli are down-weighted by the sensory system: habituation. They are not unseen; they no longer enter the rank of attention that reads. Banner blindness is the same mechanism on a UI—the top strip is coded “ad zone,” and a useful sentence placed there is skipped too. Loss after the threshold is channel-wide: a team still thinks it can slip in one high-value hint, but the channel is closed, high and low die together. The threshold counts what the person lived through, not the release calendar; three features hinting once in a week are three events. Habituation also leaves a residue: after a few quiet days, the same skin returning is still classed together. Restraint is therefore not manners; it is whether the hint channel remains usable.
Studying it
In a live UI, split hint frequency into low, medium, high; measure effective reading per item (dwell past readable duration, not instantly dismissed) and reading of the next item.
Independent variables: items per unit time, whether skins are isomorphic, whether position is fixed. Dependent variables: index where reading collapses, share of a clearly useful hint still ignored after collapse, recovery after a quiet interval.
A single lab task is too short for habituation. Do not treat clicks as reading: clicks to dismiss also look like success. Isomorphic skins collapse earlier; separate skin change from count.
Where it stops holding
Hints on errors and blocking failures belong to another channel (persistent, different skin) and should not share the teaching dose; sharing would fill the threshold on incident days. User-opened help is not a hint. Safety alerts at high frequency are alarm fatigue: similar mechanism, different cost (missed danger); do not reuse a teaching-hint threshold. A professional monitoring wall is itself a hint surface whose threshold is set by watch procedure, not by onboarding policy.
Applying it
- Inventory every automatic teaching or promotional layer (dots, bars, bubbles, update strips) and count them as one dose, not as separate “just once” events.
- Put a hard cap on that dose; after it, new hints may not reuse the same skin and place.
- Split channels that must remain usable (true errors) out of the teaching skin so they are not habituated together.
- Verify by plotting reading-rate curves of consecutive automatic hints per user. When almost everything is dismissed from item N, pull the cap below N, then place one high-value hint after collapse—if it also dies, the channel is spent; cut dose before rewriting copy.