Once fatigue sets in, it is hard to reverse
Aliases: learned ignore · channel distrust · fatigue persistence
What it is
Alarm fatigue is not a toggle. It is learned ignoring of a channel. Once ignoring is habit—mute on sight of the source, flip the device, twist the threshold until almost nothing sounds—response does not snap back just because the sender later cut false positives. What people updated is "this source is not credible," not "the last hour's statistics." Rebuilding is far more expensive than the first build.
This is fatigue as sticky learning. It does not explain how false alarms produce ignoring, and it does not argue for a louder rescue.
Why it happens
Habit short-circuits detection: attention no longer treats that sound as change, so the handling program never starts. Statistical improvement (false positives down) happens in the channel; habit lives in the person; the clocks do not match. Changing habit needs repeated samples of "it rang and it was worth it" to overwrite the old posterior. After losing trust, senders often dare not ring, so samples get rarer and the posterior is harder to flip.
There is a social layer too. One teammate saying "those can be ignored" spreads; newcomers inherit ignoring without ever seeing a true positive. Institutional muting (turning monitor volume down, batch-disabling rules) writes personal habit into the environment, and later calibration is harder still.
Studying it
Train ignoring with a high false-alarm stretch, then drop false alarms to baseline, and see whether true-event response returns to pre-training. The time constant of recovery matters, not satisfaction on the day noise drops.
Independent variables: length of the false-alarm period, whether sound or place is changed after the cut, whether checkable true events are interleaved. Dependent variables: number of true events needed to recover original response time, whether mute settings are reverted, whether newcomers still inherit ignoring.
Tens of minutes of lab annoyance is not months of field habit. In the field, look weeks after a large false-alarm incident at whether true alerts are still slow. Do not measure recovery on the day noise drops—people are still noticing "it's quieter," which is novelty, not habit reversal.
Where it stops holding
Someone new to a channel has no fatigue to reverse; early false alarms write the default. Rare, high-consequence alarms (a fire bell) sit on law and drills, so everyday ignoring is harder to freeze—but that institutional guardrail is not something consumer notices have. Changing sender identity, sound, and place is opening a new channel, sometimes easier to take seriously than "getting better" on the old one—that is migration, not reversal.
Applying it
- Treat "not yet distrusted" as a non-renewable resource: crush false alarms before launch, rather than promising to tune later.
- If ignoring is already visible in logs (true-event response near false, high mute rate), stop sending, move channel or sound, then restore true events—do not quietly improve in the original slot.
- On team products, inspect default mutes and folk rules of "don't look at those"; treat them as fatigue already institutionalized.
- Verify by taking true events after a false-alarm peak and plotting response time against the pre-peak baseline. If the gap does not close after noise drops, do not add louder stimulation on the old channel; migrate and measure hits on the new one separately.