Instrument Interpretive Drift¶
Core Idea¶
Instrument interpretive drift is the pattern in which a measurement instrument's interpretive calibration — the practice by which its output is produced from its input — silently shifts over time while its stated specification remains constant. The longitudinal data stream the instrument produces therefore mixes a constant rubric with a drifting practice, and observed temporal trends in the data mingle real change in the world with artefactual change in the instrument. Crucially, the drift is invisible to standard cross-sectional quality control: inter-rater agreement and test-retest reliability within a single time slice can remain high while the entire cohort of measurers (or the entire calibration of the instrument) drifts together. The pathology surfaces only by re-measuring frozen reference instances across time.
The prime is structurally distinct from drift in the world being measured and from one-off recalibration events. Here the world may be stable; the instrument's interpretive practice is the moving part, and its motion is silent because the instrument's own quality-control machinery cannot see the cohort-level shift it is itself part of. Four commitments fix the shape. First, an instrument with a stated specification — a rubric, a coding scheme, a sensor calibration curve, a diagnostic criterion, a benchmark, a rating standard. Second, a separate interpretive practice — how the specification is actually applied at any given time, which depends on cohort training, accumulated precedent, downstream feedback, and shifting external context. Third, a longitudinal data stream produced over a period during which the specification is held formally constant. Fourth, a drift mechanism in the interpretive practice — cohort turnover, internal-precedent accumulation, feedback-loop adaptation, external-context shift — that silently changes the mapping from input to output without changing the specification.
How would you explain it like I'm…
Same Test, Stricter Grading
The Ruler That Quietly Changed
Spec Frozen, Practice Drifting
Structural Signature¶
the instrument with a stated specification — the separate interpretive practice — the longitudinal data stream under a formally-held-constant spec — the silent drift mechanism in the practice — the within-slice quality-control blindness — the frozen-reference detection method
A measurement process exhibits instrument interpretive drift when each of the following holds:
- An instrument with a stated specification. There is an explicit rubric, coding scheme, calibration curve, diagnostic criterion, or rating standard that defines, on paper, how output is to be produced from input.
- A separate interpretive practice. Distinct from the written specification is how it is actually applied at any moment, which depends on cohort training, accumulated precedent, downstream feedback, and external context. Specification and practice are two different objects.
- A longitudinal data stream under a constant spec. The instrument produces measurements over a period during which the specification is held formally unchanged, so any change in output is not attributable to an acknowledged methodology change.
- A silent drift mechanism. The interpretive practice moves over time — through cohort turnover, internal-precedent accumulation, feedback-loop adaptation, or external-context shift — changing the input-to-output mapping without changing the specification. The drift is correlated across time and not zero-mean, so it differs statistically from noise and from a fixed offset.
- Within-slice quality-control blindness. Inter-rater agreement and test-retest reliability within a single time slice stay high because the whole cohort drifts together. The instrument's own quality control is blind to exactly the cohort-level shift it is part of.
- A frozen-reference detection requirement. The drift surfaces only by re-measuring reference instances that should not have changed; the gap on those frozen items is the practice drift, cleanly separated from real-world change.
These components decompose any longitudinal trend into three separable contributions — real world change, acknowledged specification change, and silent practice drift — with frozen-reference re-measurement as the apparatus that isolates the third.
What It Is Not¶
- Not
measurement_uncertainty. Noise is zero-mean and averages out; this drift is correlated across time and trends in one direction, so more measurements do not cancel it. High inter-rater agreement is consistent with severe drift, not evidence against it. - Not a static
bias. Bias is a fixed systematic offset removable by one calibration; this is a moving offset — a bias that drifts — so a single recalibration is defeated by continued drift. - Not a one-off recalibration event. Acknowledged methodology changes are easy to adjust for. This prime is the silent shift in interpretive practice while the stated specification stays formally constant and no version bump is recorded.
- Not
data_driftorconcept_drift. Those name change in the world being measured — input distributions or input-output relations shifting. Here the world may be stable; the instrument's interpretive practice is the moving part. - Not
goodharts_law. Metric-targeting is one possible drift mechanism, but the prime also covers cohort turnover, precedent accumulation, and external-context shift — drift not driven by anyone optimising the measure. - Common misclassification. Auditing the rubric text, finding it unchanged, and concluding the instrument is stable. The specification and the practice are two objects; a change in practice without a change in version is the prime's signature, invisible to any audit that inspects only the document.
Broad Use¶
The constant-spec, drifting-practice shape recurs across substrates that share no material. In manufacturing metrology, the canonical analogue, a measurement gauge needs periodic recalibration against a master standard precisely because its interpretive practice (transduction, offset, gain) drifts even while its stamped specification is unchanged, and gauge R&R programs exist to detect cohort-level drift. In machine-learning annotation, human labellers applying a constant guide produce label distributions that drift over months as the cohort turns over, internal precedent accumulates, and downstream feedback reshapes habits — while inter-annotator agreement at any one slice stays high. In medical coding, ICD conventions tighten or loosen over years so that what gets recorded as "heart failure" or "sepsis" changes without the diagnostic criteria on paper changing, a primary confound in disease time-trend analyses. The pattern recurs in judicial sentencing (the same statutory range, but practice on what counts as a "serious" instance drifts across decades), in survey-response coding (practice drifting across waves as new coders join), in performance reviews (the rubric document unchanged while the practice of "exceeds expectations" drifts under grade-inflation feedback), in academic peer review (acceptance standards drifting across editor cohorts), in astronomical photometry (detector sensitivity drifting across years even at constant stated specification), in standardised-test scoring (rubrics held constant but applied differently across scorer cohorts), and in compliance auditing (the same checklist, drifting practice on what passes). In each, within-slice agreement can stay high while the cohort drifts together, so the artefact is invisible to the instrument's own quality control and surfaces only on re-measured frozen references.
Clarity¶
Instrument interpretive drift clarifies by separating three failure modes that conflate when one has only the phrase "the data is drifting." The world is drifting is real change in the measured phenomenon. The specification was changed is an acknowledged methodology change, easy to adjust for. The specification is unchanged but the practice has silently moved is this prime — and it is its own pathology because its detection method, its corrective, and its institutional locus all differ from the other two. Naming it makes inspectable a class of longitudinal-trend artefacts that would otherwise be invisible, because the surface form of the data carries no marker distinguishing real change from practice drift.
The clarifying force is sharpest in explaining why inter-rater agreement is not sufficient to detect the drift: the cohort agrees with itself at every time slice precisely because it is drifting together, so high agreement is consistent with severe drift rather than evidence against it. This insight reframes the whole detection problem, because it shows that the standard quality-control instrument is blind to exactly the failure that matters. The prime also distinguishes itself carefully from neighbours that share surface features. It is not the world-side drift of input distributions or input–output relationships; here the instrument is the moving part and the world may be stable. It is not zero-mean noise around a stable instrument; the drift is correlated across time and not zero-mean, so it has a different statistical structure. It is not a static systematic offset; it is a moving systematic offset, a bias that drifts. And while metric-targeting can be one mechanism of practice drift, the prime covers many drift mechanisms not driven by targeting — cohort turnover, precedent accumulation, external-context shift — so it is broader than any single feedback story. Holding these apart keeps the drift from being mistaken for noise, for a fixed bias, or for world-side change with a different fix.
Manages Complexity¶
A long-baseline data stream contains, in principle, three signal components: real change in the world, real change in the specification, and silent change in the practice. The unaided analyst conflates them and reaches the wrong policy or scientific conclusion. The prime decomposes them into separately auditable quantities, each with its own detection apparatus: cross-sectional quality control for real-world change under stable practice, methodology-change documentation for specification change, and frozen-reference re-measurement for practice drift. The analyst stops asking "is the data drifting?" and starts asking "of the three components, which is drifting and how much?" — a decomposition that converts an undifferentiated trend into three independently measurable contributions.
The compression is operational because each component has a matched apparatus, and the practice-drift component has a clean isolation method. Banking frozen reference instances at study inception and re-measuring them at intervals yields the drift on items that should not have changed, which is the instrument-practice drift, cleanly separated from real change. Once isolated, the drift can be subtracted from longitudinal comparisons, or the comparisons can be conservatively marked unreliable. Rotating or refreshing the measurer cohort with anchor-based retraining resets practice against the original specification. Maintaining a stable subset of measurers across waves partitions the drift into within-measurer and cohort-turnover components. And versioning the rubric separately from its application distinguishes acknowledged specification change from silent practice change. Because the practice-drift contribution can be measured directly on frozen references rather than inferred from the full data stream, the prime turns an apparently intractable confound — disentangling real change from artefact in a long time series — into a tractable measurement on a small set of unchanged-by-design reference items.
Abstract Reasoning¶
Instrument interpretive drift supports reasoning about the validity of longitudinal comparisons at the level of measurement-system design rather than data content. A statistician designing a longitudinal study can ask, in advance: what is our instrument-drift exposure? Have we banked frozen reference instances we can re-measure? Have we tracked the cohort of measurers and their training pipeline? Because these questions reference only the abstract roles — instrument, stated specification, interpretive practice, longitudinal stream, drift mechanism, within-slice QC, frozen-reference detection — they are portable: a metrology engineer's gauge-R&R discipline transfers directly to ML annotation pipelines, medical-coding adjudication panels, judicial sentencing reviews, and peer-review meta-analyses.
Several reusable moves follow. The three-component decomposition treats any longitudinal trend as the sum of world change, specification change, and practice drift, so the reasoner asks which component dominates rather than treating the trend as a single quantity. The within-slice-blindness move recognises that inter-rater agreement cannot detect cohort drift, so the reasoner refuses to take high agreement as evidence of stable practice and instead demands a frozen-reference check. The frozen-reference move converts the invisible practice drift into a measurable quantity by re-measuring items that should not have changed, so the drift becomes an observable rather than an inference. And the rubric-versioning move distinguishes acknowledged specification change from silent practice change by requiring that any change in practice be documented as a version bump even when the rubric text is unchanged — a change in practice without a change in version being the prime's diagnostic signature. The same reasoning that tells a manufacturer to re-measure master gauges tells an economics department to lock away anchor answer sets, because both are reasoning about practice drift relative to a frozen reference rather than about the data stream itself.
Knowledge Transfer¶
Across substrates the same matched intervention vocabulary recurs, and a practitioner who has internalised it in one domain can deploy it on first contact with another. Bank frozen reference instances at study inception and schedule re-measurement — master gauges in manufacturing, anchor papers in test scoring, gold-standard annotation sets in ML, reference patients in medical diagnosis, reference stars in astronomy, sentinel cases for judicial review. Re-measure those references at intervals and compute the drift on items that should not have changed, since that drift is the instrument-practice drift, cleanly separated from real change. Adjust longitudinal comparisons for the measured drift, conservatively or quantitatively. Rotate or refresh the measurer cohort with anchor-based retraining to reset practice against the original specification. Maintain panel continuity by keeping a stable subset of measurers across waves so drift can be partitioned into within-measurer and cohort-turnover components. And distinguish rubric-versioning from practice-drift by versioning the rubric explicitly and tracking its application separately from its text.
The transfer is deep because these are not analogies but the same toolkit applied to one structural problem: a manufacturer's gauge-R&R session and a peer-review journal's editorial recalibration retreat are doing the same work. An economics department makes the mapping concrete. It changes nothing about its grading rubric between 2010 and 2025 — same point allocations, same example answers, same pass-mark — yet the proportion of As in introductory courses doubles, and a "grade inflation" investigation begins. Three explanations compete: students improved, the rubric changed (it did not), or the practice of applying it drifted; standard quality control at any single year shows high inter-grader agreement because the cohort drifts together, exactly the combination the prime predicts. The matched intervention vocabulary applies directly: bank frozen reference answer sets at inception (2010 anchor papers locked away), have current graders blindly re-grade them (the drift on anchor papers is the instrument drift), adjust historical comparisons by the measured drift, rotate the cohort with anchor-based retraining, maintain a small panel of graders across years, and require any practice change to be documented as a version bump even when the rubric text is unchanged. Because the toolkit is substrate-neutral, the metrologist running gauge R&R and the journal editor running an anchor-recalibration retreat are applying the same moves to the same pathology, and a practitioner who knows it in one field can design the safeguard for another on first contact.
Examples¶
Formal/abstract¶
Manufacturing metrology supplies the pattern in its sharpest, most quantified form, and it is where the three-component decomposition is provably clean. A coordinate-measuring machine on a production line has a stated specification — its certified accuracy, its calibration curve, its measurement protocol — held formally constant across a year. Its interpretive practice is the actual transduction: the offset and gain of its sensors, the thermal state of the frame, the operator's fixturing habits. Over months this practice drifts silently — sensor offset creeps, the cohort of operators turns over and develops new fixturing conventions — without a single line of the stated specification changing. The longitudinal data stream of measured part dimensions therefore mixes real change in the manufacturing process with artefactual change in the gauge. The within-slice blindness is exact and demonstrable: a gauge repeatability-and-reproducibility (R&R) study run on any single day shows excellent agreement among operators and excellent test-retest reliability, because they are all drifting together against the same slowly-moving instrument. The drift surfaces only through the frozen-reference method: re-measuring a certified master artefact — a gauge block whose true dimension cannot have changed — and reading the gap. That gap is the instrument-practice drift, cleanly separated from any real process change, because the reference is unchanged by construction. The corrective is the matched toolkit: bank the master, re-measure on a schedule, subtract the measured drift from longitudinal comparisons, and recalibrate the cohort against the master.
Mapped back: The CMM's certified spec is the stated specification, its sensor-and-operator practice is the interpretive practice, the master gauge block is the frozen reference, and the daily R&R study is the within-slice quality control blind to cohort drift — instrument interpretive drift with the decomposition into real-process change versus gauge drift made metrologically exact.
Applied/industry¶
An economics department and a hospital coding office run the identical pathology in human-judgement substrates. The department changes nothing about its grading rubric between 2010 and 2025 — same point allocations, same example answers, same pass mark — yet the proportion of As in introductory courses doubles, triggering a "grade inflation" investigation. Three explanations compete: students improved (world change), the rubric changed (specification change — it did not), or the practice of applying it drifted (this prime). Standard quality control at any single year shows high inter-grader agreement, because the grader cohort drifts together — exactly the within-slice blindness the prime predicts. The matched intervention isolates the third component: bank frozen reference answer sets from 2010, have current graders blindly re-grade them, and read the drift on papers that cannot have changed; that drift is the instrument drift, which can then be subtracted from the historical comparison. The same structure governs medical coding: ICD conventions for what gets recorded as "sepsis" tighten or loosen across years through accumulated precedent and downstream reimbursement feedback, while the diagnostic criteria on paper stay fixed — so a disease time-trend showing rising sepsis incidence confounds real epidemiological change with coding-practice drift. The corrective is the same: re-code a frozen set of archived charts under current practice and read the drift against the original coding. A grade-inflation auditor, a medical-coding adjudication panel, and a manufacturing metrologist running gauge R&R are applying one playbook — frozen references, blind re-measurement, drift subtraction, cohort recalibration — to one structural pathology.
Mapped back: The unchanged grading rubric and the fixed ICD criteria are stated specifications; the grader cohort and the coder cohort are the drifting interpretive practices; the 2010 anchor papers and the archived charts are frozen references; the high single-year inter-rater agreement is the within-slice blindness — the same prime in academic assessment and clinical coding.
Structural Tensions¶
T1 — Specification versus Practice (scopal). The prime's whole leverage rests on holding the written specification distinct from the practice that applies it — two objects that share a name and are routinely conflated. The characteristic failure is auditing the rubric text, finding it unchanged, and concluding the instrument is stable, when the practice has moved beneath a fixed spec. The diagnostic is to ask not "has the specification changed?" but "has the mapping from input to output changed while the specification stayed fixed?" A change in practice without a change in version is the prime's signature, and any audit that inspects only the document misses it entirely.
T2 — Within-Slice Agreement versus Cohort Drift (scalar / local-global). High inter-rater reliability at any single time slice is locally reassuring but globally uninformative, because a cohort that drifts together agrees with itself at every instant. The failure is taking strong agreement as evidence of stable practice — exactly inverting its meaning, since synchronous drift produces high agreement. The diagnostic is to refuse cross-sectional QC as a stability check and instead re-measure a frozen reference across time: only the longitudinal gap on unchanged items distinguishes a stable instrument from a cohort marching in lockstep away from its origin.
T3 — World Change versus Practice Drift (sign/direction). A longitudinal trend mixes real change in the measured world with artefactual change in the instrument, and the two are indistinguishable on the surface of the data — a rising sepsis rate could be epidemiology or coding drift. The failure is attributing an instrument artefact to the world (or vice versa), reaching a wrong scientific or policy conclusion. The diagnostic is the frozen reference: re-measure items that cannot have changed in the world; any movement on them is pure practice drift, which can then be subtracted to expose the genuine world-side signal underneath.
T4 — Drifting Bias versus Static Offset versus Noise (measurement). The drift is a moving systematic bias — correlated across time, not zero-mean — so it is statistically distinct from both a fixed offset (removable by one calibration) and from noise (removable by averaging). The failure is applying the wrong correction: a single recalibration assumes a static offset and is defeated by continued drift, while averaging assumes noise and leaves the trend intact. The diagnostic is to characterise the temporal structure of the error: if it trends rather than scatters and persists rather than offsets, neither a one-off calibration nor noise-reduction will fix it, and only repeated frozen-reference subtraction tracks a moving bias.
T5 — Frozen Reference Stability versus Reference Decay (temporal). Detection depends on a reference that genuinely cannot have changed — yet references can themselves degrade (a gauge block wears, an anchor answer leaks into training, archived charts are re-interpreted), at which point the instrument the method trusts is itself drifting. The failure is reading reference decay as instrument stability or vice versa, corrupting the one fixed point the method relies on. The diagnostic is to verify the reference's invariance independently and protect it from contamination (lock it away, rotate sealed copies); a reference that is not provably unchanged cannot isolate drift and silently re-introduces the confound it was meant to remove.
T6 — Cohort Turnover versus Within-Measurer Creep (coupling). Practice drift has at least two coupled sources — new measurers joining with different calibration, and continuing measurers whose own application creeps — and they call for different remedies. The failure is treating all drift as turnover (and fixing it by retraining new hires) when continuing-measurer creep is the real driver, or the reverse. The diagnostic is to maintain a stable panel across waves so the total drift can be partitioned: the gap among long-tenured measurers isolates within-measurer creep, and the residual against the full cohort isolates turnover, letting each be addressed at its actual source.
Structural–Framed Character¶
Instrument interpretive drift sits just onto the framed side of the structural–framed spectrum — the balanced-hybrid case where its framed label and aggregate of 0.5 reflect every diagnostic reading exactly mid. There is a genuine structural skeleton: an instrument with a stated specification, a separate interpretive practice that silently drifts, a longitudinal stream under a formally-held-constant spec, within-slice quality-control blindness, and a frozen-reference detection method. But the "interpretive practice" that drifts is, in its home cases, a cohort of human measurers applying a rubric, and that practice-dependence pulls each criterion halfway toward framed.
Vocabulary partly travels: the constant-spec / drifting-practice / frozen-reference shape is statable in each domain's own words — gauge drift in manufacturing metrology, annotation-guideline drift in ML labelling, coding-practice creep in ICD coding, sentencing drift in the judiciary, ratings inflation in performance reviews, standards drift in peer review — yet a metrological "specification / interpretive calibration / cohort" lexicon tends to come along (vocab_travels 0.5). It carries a mild negative valence: "drift" and "contamination" frame the pattern as a defect that corrupts longitudinal trends, something to be detected and corrected, though the underlying mapping-shift is itself value-neutral (evaluative_weight 0.5). Its origin is metrological and statistical rather than rooted in a single named institution, sitting between formal and human-practice origins (institutional_origin 0.5). It is partly human-practice-bound: a purely physical sensor's calibration curve can drift without people, but the prime's signature mechanisms — cohort turnover, internal-precedent accumulation, feedback-loop adaptation — and its defining blindness (a cohort that drifts together while inter-rater agreement stays high) presuppose human interpretive practice (human_practice_bound 0.5). And invoking it imports a modest interpretive frame — reading a longitudinal trend as a mix of real change and instrument artefact, with the matched "re-measure frozen references" remedy — while still recognising a real shift in the input-output mapping that is genuinely present (import_vs_recognize 0.5). Because every criterion lands at the midpoint, the prime is a true hybrid that the rubric places just on the framed side of center: a real measurement-system shift wrapped in a practice-bound, evaluatively-loaded metrological frame.
Substrate Independence¶
Instrument interpretive drift is a moderately substrate-independent prime — composite 3 / 5 on the substrate-independence scale. Its domain breadth is real but bounded: the constant-spec, drifting-practice shape recurs across manufacturing metrology (gauge drift and gauge R&R), machine-learning annotation, medical coding (ICD convention creep), judicial sentencing, survey-response coding, performance reviews, academic peer review, astronomical photometry, standardised-test scoring, and compliance auditing — genuinely distinct fields, but nearly all of them human interpretive practices, with only a thin reach toward purely physical instruments (a detector's sensitivity drift) where the prime's signature mechanisms do not fully apply. Its structural abstraction is mid because the signature — a fixed specification, an interpretive practice that drifts, and a frozen reference that exposes it — is portable, but its defining mechanisms (cohort turnover, internal-precedent accumulation, feedback-loop adaptation) and its diagnostic blindness (a cohort drifting together while inter-rater agreement stays high) presuppose human interpreters; a pure sensor can drift but does not drift as a cohort. Its transfer evidence earns a 4: the same artefact is documented across the substrates with the same recovery doctrine — periodic recalibration against a master standard, gauge R&R, re-measuring frozen references — and the within-slice-agreement-while-the-cohort-drifts pathology recurs intact, surfacing only on re-measured anchors in metrology, ICD coding, and ML labelling alike. The composite sits at 3 because breadth and abstraction are pinned by the human-interpretive ceiling even though cross-domain transfer within that band is concretely evidenced.
- Composite substrate independence — 3 / 5
- Domain breadth — 3 / 5
- Structural abstraction — 3 / 5
- Transfer evidence — 4 / 5
Relationships to Other Primes¶
Parents (1) — more general patterns this builds on
-
Instrument Interpretive Drift is a kind of, typical Temporal Decay and Degradation
A time-dependent, correlated (non-zero-mean) drift of a measurement instrument's interpretive practice while its stated spec stays fixed — a specialization of temporal degradation applied to interpretive calibration (vs a static bias removable by one calibration).
Path to root: Instrument Interpretive Drift → Temporal Decay and Degradation → Entropy (Thermodynamic Sense)
Neighborhood in Abstraction Space¶
Instrument Interpretive Drift sits among the more crowded primes in the catalog (38th percentile for distinctiveness): several abstractions describe nearly the same structure, so a description that fits it will tend to fit its neighbors too — transporting it usually means disambiguating within this family rather than landing on it exactly.
Family — Cue-Outcome Drift & Silent Failure (18 primes)
Nearest neighbors
- Concept Drift — 0.75
- Data Drift — 0.74
- Calibration Anomaly — 0.73
- Calibrated Rule versus Moving World — 0.71
- Measurement Uncertainty and Observational Noise — 0.71
Computed from structural-signature embeddings · 2026-06-14
Not to Be Confused With¶
The nearest confusion is with measurement_uncertainty, because both describe a measured value departing from the truth. Observational noise is zero-mean and stochastic: it scatters around the true value, is uncorrelated across measurements, and is removed by averaging or by more samples. Instrument interpretive drift is correlated across time and not zero-mean: the interpretive practice moves systematically in one direction, so its contribution to a longitudinal trend is a trend, not a scatter. This statistical difference dictates opposite remedies. Against noise, averaging more measurements helps and high inter-rater agreement is reassuring. Against drift, both move exactly the wrong way: averaging late accounts leaves the trend intact, and high within-slice inter-rater agreement is consistent with severe drift because a cohort that drifts together agrees with itself at every instant. This is the prime's most counterintuitive claim — synchronous drift produces the very agreement that a noise framing would read as evidence of stability. A practitioner who models the problem as noise will trust the agreement statistic and certify a cohort marching in lockstep away from its origin; only a frozen-reference re-measurement, which a noise framing never prescribes, exposes the drift.
A second genuine confusion is with a static bias, and the distinction is the difference between a fixed offset and a moving one. A bias is a constant systematic error — the instrument always reads two degrees high — removable by a single calibration that subtracts the offset. Instrument interpretive drift is a moving systematic offset: the bias itself changes over time as the interpretive practice creeps, so a one-time calibration that corrects today's offset is defeated by tomorrow's continued drift. The remedy structure differs accordingly. A static bias needs one calibration; a drifting bias needs repeated frozen-reference subtraction that tracks the offset as it moves. Treating drift as a static bias — recalibrating once and assuming the instrument is now correct — leaves the moving error to re-accumulate, and the trend re-corrupts after the single correction. The discriminating question is the temporal structure of the error: a fixed offset is removed by one calibration, a trending one only by ongoing re-anchoring.
A third confusion, especially in metric-laden organisations, is with goodharts_law. Goodhart's law names the specific dynamic where a measure, once it becomes a target, ceases to be a good measure because agents optimise against it. This is one mechanism by which interpretive practice can drift — graders inflate, coders shade toward reimbursement — but it is only one. Instrument interpretive drift is broader: the practice can also drift through cohort turnover (new measurers calibrate differently), internal-precedent accumulation (the cohort's habits evolve absent any target), and external-context shift, none of which involve anyone optimising the measure. Collapsing the prime into Goodhart's law restricts it to the targeting case and misses drift that has no strategic driver at all — a survey-coding panel whose conventions creep across waves simply because the personnel changed. The discriminating question is whether a target is being gamed: where it is, Goodhart is the operative mechanism within the umbrella; where the drift proceeds with no metric under optimisation pressure, the prime still applies and a Goodhart-style anti-gaming fix would miss the cause.
These distinctions matter because each points to a different corrective: averaging or replication for noise, a single calibration for static bias, anti-gaming design for Goodhart — none of which detects or removes the silent, correlated, cohort-level practice drift that this prime names, which is isolated only by banking frozen references and re-measuring them across time.
Solution Archetypes¶
No catalogued solution archetypes reference this prime yet.