Skip to content

Refresh Cadence Adaptation

Adaptive controller — instantiates Activation Decay Measurement

Tunes how often a primed state is refreshed by comparing realised decay against prediction — faster fade earns earlier, more frequent reprimes, set per cohort.

Version
v1 · 2026-08-24 · History
Mechanism #
7291
Type
Adaptive Controller
Form family
Control, Automation & Runtime
Solution family
Attention, Salience & Focus
Problem family
Observability, Measurement & Feedback Gaps
Problem subfamily
Temporal Sampling, Decay & Transition Resolution
Origin domain
Cognitive Science
Also from
Psychology, Systems Thinking & Cybernetics
Instantiates
Activation Decay Measurement

A fixed refresh schedule is a bet that decay stays exactly as predicted — and it rarely does. Refresh Cadence Adaptation is the closed feedback loop that fixes the frequency of reprimes to the decay actually observed: when a group fades faster than the model expected, it shortens the interval and refreshes sooner; when a group holds longer, it lets the interval stretch and spares them the cue. Its defining move is control, not derivation — it is not a one-time window read off a curve but a running governor that keeps rewriting the cadence as new probe data lands, and it does so per cohort, giving fast-fading groups a tighter drumbeat than durable ones. It owns the timing of the reprime — how often — while leaving the design of the reprime cue itself to another mechanism.

Example

A B2B software company trains its sales teams on a repositioned product story and wants reps sharp on it at every customer call. It refreshes with short micro-lessons, but a single company-wide cadence wastes some teams and starves others. Refresh Cadence Adaptation runs the reprime schedule as a loop. It holds a decay profile per team: the enterprise team, juggling long complex deals and heavy context-switching, forgets the new story within about ten days; the SMB team, repeating it daily, holds it for a month. The predicted cadence was "refresh every three weeks for everyone."

Observed decay contradicts that. So the controller tightens the enterprise team to a ten-day refresh and relaxes the SMB team to monthly. As later probes come in — enterprise decay slows once deals shorten — it lengthens their interval again. The output is not a cue and not a diagnosis; it is a live, per-cohort refresh frequency that tracks measured fade and avoids both under- and over-refreshing.

How it works

  • Compare realised to predicted decay. Each new probe reading is scored against the model's expectation; the gap is the control error.
  • Adjust the interval in the error's direction — faster-than-expected fade shortens the cadence, slower-than-expected lengthens it.
  • Hold a decay profile per cohort so each group's cadence tracks its own fade rather than a blended average.
  • Cap the frequency so tightening never crosses into saturation; the loop protects salience as hard as it protects freshness.

Tuning parameters

  • Gain / responsiveness — how sharply cadence reacts to a decay surprise. High gain adapts fast but chases noise; low gain is stable but sluggish.
  • Cohort granularity — how finely the population is split into separately-timed groups. Finer splits fit better but multiply overhead and thin the data per group.
  • Saturation ceiling — the minimum spacing between reprimes, below which refreshing backfires.
  • Update latency — how many fresh probes accumulate before the cadence is allowed to move.
  • Asymmetry — whether the loop tightens faster than it loosens (protective) or moves symmetrically.

When it helps, and when it misleads

Its strength is efficiency under drift: it spends reminder budget where decay is actually fast and withholds it where the state is holding, so timing stays right even as conditions change. Its failure mode is over-refreshing — a loop tuned to chase every dip drives the cadence down until reminders lose salience and are tuned out, the classic alert-fatigue trap where more cues yield less attention.[1] The classic misuse is optimising cadence purely for retention while ignoring the annoyance cost, so a technically "well-timed" stream of nudges quietly trains people to ignore all of them. The guarding discipline is a hard saturation ceiling, asymmetric tuning that loosens readily and tightens cautiously, and treating a rising ignore rate as a decay signal in its own right.

How it implements the components

  • refresh_or_reprime_protocol — governs the cadence of the protocol: how frequently a reprime fires, tightening or loosening with measured decay while capped against fatigue.
  • cohort_specific_decay_profile — maintains a per-cohort fade estimate and acts on it, assigning each group its own refresh interval.

It sets frequency, not the cue, and acts on cohorts rather than merely comparing them: it does NOT implement context_and_load_register — designing the single context-matched cue is Contextual Reactivation, its reprime-side twin — and does NOT implement competing_cue_noise_monitor — diagnosing which cohorts fade fastest and why is Decay Segment Comparison, its cohort-side twin.

Editorial Notes

Form Classification

Form family: Control, Automation & Runtime

Rationale: Refresh Cadence Adaptation operates as a live operational control that automatically routes, enforces, adapts, or responds during execution because it tunes how often a primed state is refreshed by comparing realised decay against prediction — faster fade earns earlier, more frequent reprimes, set per cohort.

Independent corroboration: The frozen evidence defines Refresh Cadence Adaptation as 'Tunes how often a primed state is refreshed by comparing realised decay against prediction — faster fade earns earlier, more frequent reprimes, set per cohort', so its operative form is Control, Automation & Runtime.

Review outcome: Independent reviewer agreement; high confidence.

Origin Attribution

Primary origin: Cognitive Science

Origin pattern: Cross-disciplinary synthesis

Present-day reach: Multi-domain

Rationale: Adaptive spacing changes presentation intervals from observed learner performance and memory decay; feedback control supplies the cadence-tuning implementation.

Related originating lineages:

  • Psychology — Priming and memory-decay research materially supplies the cohort-specific state being controlled.
  • Systems Thinking & Cybernetics — The systems_cybernetics lineage materially shaped Refresh Cadence Adaptation through feedback, state estimation, recursion, and system regulation.

Review resolution: The blind reviewers disagreed on primary lineage. Light authoritative research resolves the defining form in favor of cognitive_science: Adaptive spacing changes presentation intervals from observed learner performance and memory decay; feedback control supplies the cadence-tuning implementation. The rejected primary is retained only when it materially shaped the mechanism, and present-day breadth is recorded separately as domain_reach=multi_domain.

Review outcome: Researched adjudication after independent review; high confidence.

Sources consulted:

References

[1] Alert fatigue is the well-known erosion of responsiveness that follows too-frequent alerts or reminders: as volume rises, recipients habituate and begin ignoring even important signals. It is the natural over-refreshing failure of any cadence controller. withdrawn registry