Alert Budget¶
Policy — instantiates Attention Budgeting
Caps how many signals are allowed to interrupt a person or team, treating interrupt-rights as a scarce quota that must be earned rather than assumed.
An Alert Budget treats the right to interrupt as a finite quota. Instead of asking of each new alert "is this worth showing?" — a question that always answers yes in isolation — it fixes a ceiling on how many interruptions a person or on-call rotation may receive in a period, and forces every candidate signal to compete for a slot under that ceiling. The defining move is the cap: interrupt-rights are scarce and rationed, so adding a new page means justifying it against the budget, and a channel that blows the budget must lose signals, not gain volume. This is what separates it from a mere classification scheme — there is a number, the number is enforced, and when the total exceeds it the response is to cut, not to relabel.
Example¶
A site-reliability team is paged 140 times a week. Most pages are non-actionable — a disk hits 80% and clears itself, a latency blip resolves before anyone opens the laptop — and the on-call engineers have started swiping them away half-asleep, which means the two pages a week that do matter now arrive in a stream that no one reads carefully. The team sets an alert budget: no more than fifteen interrupting pages per week per on-call shift. Every existing alert must now justify its slot. Anything that fired in the last quarter without a human taking action is either deleted, converted to a metric on a dashboard, or auto-routed to a digest that is read on a schedule rather than at 3 a.m. Two months later the on-call is paged eleven times a week, every page is something a human should look at, and the pages that matter are read on the first buzz again — not because the systems got healthier but because interrupt-rights stopped being free.
How it works¶
- Set the ceiling. Fix an explicit interruption quota — pages per shift, push notifications per day — sized to what a human can absorb without habituating.
- Make every signal compete. Each alert must earn its slot by demonstrating that a human reliably acts on it; unproven alerts do not get to interrupt by default.
- Demote, don't relabel. Signals that lose the competition are not made quieter-but-still-interrupting; they are moved out of the interrupt lane entirely — auto-filtered to a digest, a log, or a dashboard.
- Watch the overflow. When the budget is repeatedly blown, that is itself the signal that the alert set needs pruning, not that the ceiling should rise.
The budget's teeth are in enforcement: the total is measured, and exceeding it triggers removal of the least-actionable signals rather than a quiet expansion of the cap.
Tuning parameters¶
- Ceiling height — a tight quota protects attention hard but risks suppressing a rare real event; a loose one is safer against misses but slides back toward fatigue. Set it from the cost of a missed signal versus the cost of overload.
- Budget window — per-shift, per-day, or per-week. Shorter windows react faster to a storm; longer ones tolerate bursts but can hide a slow creep.
- Actionability bar — how much evidence an alert needs (last-N-fires action rate) before it keeps its slot. A high bar prunes aggressively but can retire a genuine tail risk.
- Overflow discipline — what happens when the budget is exceeded: auto-suppress the noisiest, freeze new alerts, or force a review. Stricter discipline keeps the cap honest but needs someone empowered to cut.
- Exemption list — which few signals bypass the budget entirely (page-always safety alarms). Every exemption is attention you can no longer protect, so the list must stay short.
When it helps, and when it misleads¶
Its strength is that it attacks the root cause of ignored warnings — volume — rather than the symptom. A capped, competed-for alert stream stays credible, and credibility is the whole point of an interrupt: an alert nobody reads is worse than no alert, because it costs attention and delivers nothing. The budget also converts "we have too many alerts" from a complaint into an enforceable constraint with a number attached.
Its failure mode is a threshold set by comfort rather than by risk: squeeze the budget too hard and a rare, high-consequence signal gets pruned along with the noise, so the very event the system exists to catch arrives silently. This is the danger behind alert fatigue[1] — the cure (fewer alerts) and the disease (missing the real one) live on the same dial. A classic misuse is to hit the budget by moving noisy alerts to a "warning" tier that still interrupts, which relabels the problem instead of solving it. The guarding discipline is to prune by demonstrated non-action, keep a short exemption list for safety-critical signals, and periodically audit what was suppressed to confirm nothing important is dying quietly in the digest.
How it implements the components¶
attention_budget— the explicit interruption quota is the budget, denominated in interrupt-rights rather than time or display space.interruption_policy— the competition for slots is the policy for what may override the current focus and what may not.overload_signal— a repeatedly-blown budget is the built-in indicator that demand has outrun capacity and the alert set needs cutting.automation_filter— losing signals are auto-routed out of the interrupt lane to digests, logs, or dashboards.
It does not classify every signal into a graded display lane (salience_hierarchy) — that's Notification Tier System, which sorts rather than caps; nor does it weight which content categories deserve attention (priority_weighting_rule) — that's Dashboard Triage View.
Related¶
- Instantiates: Attention Budgeting — Alert Budget is the interruptive-signal instantiation: it rations the scarce right to interrupt.
- Sibling mechanisms: Notification Tier System · Dashboard Triage View · Attention WIP Limit · Quiet Hours Policy · Watch Rotation · Executive Attention Review · Editorial Priority Calendar · Focus Block Schedule
Editorial Notes¶
Form Classification¶
Form family: Rule, Policy & Commitment
Rationale: Caps how many signals are allowed to interrupt a person or team, treating interrupt-rights as a scarce quota that must be earned rather than assumed, making its operative form a standing constraint, permission, threshold, obligation, or conditional rule.
Independent corroboration: The frozen evidence defines Alert Budget as 'Caps how many signals are allowed to interrupt a person or team, treating interrupt-rights as a scarce quota that must be earned rather than assumed', so its operative form is Rule, Policy & Commitment.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Computer Science & Software Engineering
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Site-reliability and software-operations practice made actionable paging, on-call interruption load, and pruning non-actionable alerts into a governed engineering concern.
Related originating lineages:
- Cognitive Science — Limited attention, habituation, and vigilance capacity explain why an interruption quota is necessary.
- Human-Computer Interaction — Notification design treats interruptions and attentional switching as scarce interaction costs.
- Organizational & Management Science — Teams allocate interrupt rights and on-call capacity as scarce organizational resources.
Review resolution: The enforced interrupt quota is implemented in computer operations, with human-computer interaction and cognitive attention limits supplying the burden model and organizational management supplying rotation capacity. Clinical alerting is an important application, not a necessary additional origin.
Attribution caveat: The hard quota is a policy synthesis over several alert-fatigue traditions rather than a universally standardized named control.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Reconciled after independent review; high confidence.
References¶
[1] Alarm/alert fatigue — the desensitization that sets in when people are exposed to a high volume of alerts, so that they respond more slowly or ignore them, including the ones that matter. It is a documented patient-safety hazard in clinical monitoring and a well-known on-call problem; reducing alert volume is the standard corrective, which is exactly what a budget enforces. withdrawn registry ↩