Stimulus–Response Pilot¶
Empirical pilot method — instantiates Dose–Response Calibration
Runs a bounded trial across several predeclared stimulus levels to fit the shape of the input-to-response curve, with its uncertainty and its subgroup differences attached.
Stimulus–Response Pilot is the curve-learning mechanism of the family: a designed trial whose whole deliverable is the map relating intensity to response. Its defining move is to measure response at several stimulus levels simultaneously — not to optimize a single dial in place, but to recover the shape of the relationship across the tested range, complete with how confident that shape is and how it differs between subgroups. Where a spend-calibration method turns one knob and watches the margin, this pilot lays out a spread of levels at once and reads the contour they trace. Its output is an estimated response curve with a band of uncertainty around it and, where it matters, separate curves for populations that respond differently.
Example¶
An agronomy team wants to know how a wheat variety yields as nitrogen fertilizer rate rises. Rather than guess or copy a neighbor's rate, they lay out a replicated field trial: plots receiving 0, 60, 120, 180, and 240 kg N/ha (illustrative), randomized and repeated across blocks to average out field variation. At harvest they measure yield per plot — the response — and fit a curve through the five levels. It rises steeply from zero, bends over around 150 kg, and flattens after that: more nitrogen buys almost nothing past the bend, and the confidence interval widens at the high end where fewer plots inform the estimate. Splitting the plots by soil type reveals a real difference — the sandy blocks plateau earlier than the clay ones. The pilot hands back not a single "best rate" but a mapped relationship with its error band and two subgroup curves, which downstream decisions can then read against price and cost.
How it works¶
The method is defined by testing multiple predeclared levels concurrently under a design (randomization, replication, blocking) that lets field noise be separated from the signal. Response is measured at each level and a curve is fitted through the points — linear, sigmoid, or plateau-shaped as the data warrant — rather than assumed. Two things travel with that curve and distinguish this mechanism from a simple "run a test and pick a winner": an uncertainty band quantifying how well the shape is pinned down, and a stratified reading that fits separate curves where subgroups diverge instead of averaging a heterogeneous system into one misleading line.
Tuning parameters¶
- Level spacing — where the tested doses sit. Levels clustered near the suspected inflection resolve the bend sharply but leave the extremes vague; evenly spread levels map the whole range coarsely.
- Replication — how many repeats per level. More replicates tighten the uncertainty band but cost plots, subjects, or time.
- Curve family — the functional form fitted (linear, quadratic, sigmoid, plateau). A flexible form fits the data but can overfit noise; a rigid one is stable but may miss the true shape.
- Stratification depth — how finely the sample is split into subgroups. Finer strata expose real heterogeneity but shrink each stratum until its own curve turns noisy.
- Range width — how far the highest tested level reaches. A wider range catches the plateau and the downturn but risks exposing part of the sample to a wasteful or damaging level.
When it helps, and when it misleads¶
Its strength is that it refuses to assume linearity: by measuring several points at once it recovers the actual shape — the threshold, the bend, the plateau — and reports how sure it is, which is precisely what lets a decision reject "if a little helped, more will help more." Attaching subgroup curves also catches the case where an average hides one population helped and another wasted.
Its failure mode is extrapolation and overfitting. A curve fitted over a tested range says nothing reliable beyond it, yet a tidy fitted line invites reading off values the trial never probed; and a flexible form chasing noise can invent a bend that replication would erase. The classic misuse is treating the fitted curve as a law rather than a range-bound, noisy estimate — a dose-ranging study[n1] answers "what happens between these levels," not "what happens everywhere." The discipline is to carry the uncertainty band into every downstream use and to re-run when the context that produced the curve has moved.
How it implements the components¶
response_metric— it measures the intended outcome (yield) at each tested level, the raw material the curve is fitted from.calibration_curve— its central deliverable: the fitted input-to-response relationship across the tested range.uncertainty_band— confidence intervals travel with the curve, marking where the shape is well pinned down and where it is guesswork.sensitivity_strata— it fits separate curves for subgroups that respond differently rather than averaging them into one line.
It does not track the marginal return of a continuously turned dial (marginal_response_metric) or set a spend cap (escalation_stop_rule) — that is Advertising Spend Calibration, its method-type twin, which optimizes one dial in place rather than mapping the curve; and it does not itself fix the danger boundary (harm_threshold), which Policy Intensity Pilot supplies.
Related¶
- Instantiates: Dose–Response Calibration — this method is the archetype's curve-learning core: it recovers the response map that the decision rules read against.
- Sibling mechanisms: Advertising Spend Calibration · Intensity Ladder Trial · Policy Intensity Pilot · Staffing Level Experiment · Training Load Calibration · Alert Threshold Tuning · Medication Dose Calibration
Editorial Notes¶
Form Classification¶
Form family: Experiment, Test & Rehearsal
Rationale: Stimulus–Response Pilot operates as an active test, trial, simulation, drill, or rehearsal that generates evidence through a deliberate attempt or perturbation because it runs a bounded trial across several predeclared stimulus levels to fit the shape of the input-to-response curve, with its uncertainty and its subgroup differences attached.
Independent corroboration: The frozen evidence defines Stimulus–Response Pilot as 'Runs a bounded trial across several predeclared stimulus levels to fit the shape of the input-to-response curve, with its uncertainty and its subgroup differences attached', so its operative form is Experiment, Test & Rehearsal.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Statistics & Experimental Design
Origin pattern: Single lineage
Present-day reach: Multi-domain
Rationale: Predeclared input levels estimating a response curve are dose-response experimentation.
Related originating lineages:
- Data Science & Analytics — Data science, analytics, and operational monitoring supplies a parallel or contributing lineage for the mechanism's defining operation: runs a bounded trial across several predeclared stimulus levels to fit the shape of the input-to-response curve, with its uncertainty and its subgroup differences attached.
- Mathematics — Mathematical modeling, proof, and abstract-structure practice supplies a parallel or contributing lineage for the mechanism's defining operation: runs a bounded trial across several predeclared stimulus levels to fit the shape of the input-to-response curve, with its uncertainty and its subgroup differences attached.
- Medicine & Healthcare — Clinical medicine, public health, and recovery practice supplies a parallel or contributing lineage for the mechanism's defining operation: runs a bounded trial across several predeclared stimulus levels to fit the shape of the input-to-response curve, with its uncertainty and its subgroup differences attached.
- Pharmacology & Toxicology — Dose-response is foundational.
- Psychology — Behavior changes with stimulus intensity.
Review resolution: The blind reviewers agree that statistics_experimental_design is the primary origin and differ only on alternate origin disagreement, domain reach disagreement. I preserve every independently explained alternate from both records rather than imposing a numeric cap. I retain single_lineage because the combined evidence shows one traceable formative lineage. The broader reach of multi_domain records portability separately from historical provenance; encyclopedia_synthesis=false preserves the affirmative synthesis judgment where either reviewer identified one.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
[n1] A dose-ranging study is an experimental design that administers several predeclared levels of an intervention to estimate the response across a range — standard in agronomy and pharmacology alike. Its result is valid only within the tested range, which is why extrapolation is the design's characteristic misuse. ↩