Semantic Drift
|
CYCLE Calibration position — Active Development This page is a conceptual instrument under Permanent Beta. It declares a real calibration position, not a finished product waiting to ship. Checking continues; an edit is only required when evidence demands it. Stage: Seed to Fruit. Feedback welcome — especially clarity, failure modes, and calibration gaps. Use discussion or Contribute. |
Semantic Drift
Definition names the phenomenon. Purpose states the job. Calibration tests the map — it does not seize the term.
Meta
Semantic Drift
| Type | Calibration Standard |
|---|---|
| Functional Layer | Failure Condition / Conceptual Instrument Health |
| Application Layer | Conceptual Instruments |
| Category | Calibration Procedures |
| Version | 0.3 |
| Maturity | Seed |
| Last Calibration | 2026-07-27 |
| Status | Permanent Beta |
| Description | Standard for detecting and documenting progressive loss of alignment among a conceptual instrument’s dated definition, underlying mechanism, operational use, effects, and intended domain — distinct from lawful development and from neighboring failure conditions. |
Menu
Core Principles
- Reality gets final vote
- See the Game. Refuse the Game. Build Better.
- Permanent Beta
Navigation
Related
Definition (what phenomenon does this name?)
Semantic Drift is the progressive divergence of a conceptual instrument from a dated or otherwise recoverable baseline of its calibrated definition and underlying mechanism, relative to its actual usage and/or operational effect, within its intended domain, such that the instrument’s map is impaired in one or more of these ways:
- ambiguity increases in a way that impairs consistent application;
- predictive or discriminatory power decreases;
- reliability of domain mapping decreases (the instrument no longer reliably tracks the domain it was intended to serve).
Semantic Drift does not require intentional manipulation.
Purpose (what work must naming this enable?)
Detect and document when a conceptual instrument is losing reliable alignment among dated definition, underlying mechanism, operational use, operational effect, and intended domain, so that divergence can be:
- distinguished from mere variation, disagreement, enrichment, or lawful development;
- distinguished from neighboring failure conditions;
- addressed through a traceable Calibration Recommendation without claiming ownership of the definition.
Non-claims
- Usage change alone is not Semantic Drift.
- Disagreement over values or policy is not Semantic Drift unless progressive divergence from a baseline definition+mechanism is shown.
- Detecting a neighboring condition (e.g. overload) is not automatically detecting Semantic Drift.
- A newly preferred wording is not evidence that an earlier instrument drifted.
- Completing a checklist does not raise standing by itself.
- Early reports should narrow tested domain; broad generalization is unearned.
Intended domain
Broad: Conceptual instruments whose definitions, mechanisms, uses, boundaries, or operational effects can change over time.
Default tested scope for early reports: Defined instruments within Metrology of the Abstract / The Sovereign Games, unless a report explicitly widens scope and adds case diversity.
Five poles (alignment object)
Semantic Drift concerns alignment among:
- Definition — dated/recoverable wording
- Underlying mechanism — how inputs become classification, explanation, prediction, judgment, or action (grade operable / incompletely operationalized / missing; do not invent)
- Operational use — how the term is actually deployed
- Operational effect — decisions, exclusions, or outcomes that deployment drives
- Intended domain — stated scope (plus tested scope for the report)
A drift-relevant finding is progressive divergence of use and/or effect from baseline definition+mechanism within the intended or tested domain — not “pole X looks ugly in isolation.”
Which poles are well-evidenced is recorded in the report. Thin evidence on a pole → note it; do not invent fill.
Baseline gate (Claim: strong claims need a from-state)
A strong claim of Semantic Drift requires a dated or otherwise recoverable baseline (prior definition text, mechanism statement, and/or versioned instrument record).
Without such a baseline:
- do not issue a strong drift finding;
- record unsupported or indeterminate for strong drift;
- at most note usage instability or recommend establishing a baseline.
Adequacy of the baseline is a calibrator judgment from the cited record.
Progressive loss of alignment (operable path)
Alignment is lost over time when:
- a recoverable baseline (definition + mechanism) is fixed; and
- later usage and/or operational effect in the intended or tested domain diverges from that baseline; and
- divergence is shown across more than one stage or record, not by a single disagreement snapshot.
Discriminator: not every change is drift
Not Semantic Drift when the case is adequately accounted for as:
- Disagreement / variation — same baseline definition+mechanism; parties differ on application, values, or policy with no progressive pull of usage/effect away from that baseline;
- Enrichment — added examples, clarity, or scope notes that do not change load-bearing mechanism or purpose;
- Lawful development — dated revision that preserves declared purpose/identity with updated mechanism stated, or honest split/rename of child senses; prestige of the old label is not used to smuggle an incompatible payload.
Semantic Drift (discriminator path) when progressive baseline divergence is shown and the change is not adequately accounted for as the above.
Impairment (why drift matters)
For a full drift finding that relies on harm, the report cites at least one impairment family with concrete in-domain evidence relative to baseline purpose/mechanism:
- ambiguity ↑ (incompatible applications; unclear which sense is in force);
- predictive or discriminatory power ↓ (fails to separate cases it was supposed to separate);
- reliability of domain mapping ↓ (unstable or non-comparable “readings”; no longer tracks the intended domain).
Calibrator call from results: strength of impairment is judged from the cases in the report. Insufficient evidence → indeterminate on impairment (and withhold or soften full drift certification). Underclaim when thin.
Non-intentional causes (inclusion, not identity)
Semantic Drift does not require intentional manipulation.
It can arise from non-intentional processes, including (examples, not a closed list): gradual overload, failure to create needed distinctions, boundary erosion, or unrecorded change to the core mechanism.
These are candidate pathways. They are not identical with Semantic Drift. Pathway vs independent vs co-occurring is settled under neighbor rules and the detection sketch — calibrator call from results.
Neighbor relation rules (bounds)
Neighboring conditions are diagnosed on their own terms. Default: independent until drift tests are met.
| Condition | Default | Relation to drift |
|---|---|---|
| Conceptual overload | Independent | Pathway, co-occurring, or independent — not auto-drift |
| Boundary collapse | Independent | Same |
| Missing child instrument | Independent | Same (often pathway when one label is forced to do two jobs) |
| Missing exception | Independent | Same |
| Unrecorded mechanism change | Often direct pathway | Still requires baseline, progression, and domain framing for strong drift |
| Lawful development | Not drift | See discriminator section |
Do not treat detection of a neighbor as detection of Semantic Drift by default. When both are present, diagnose both; do not merge names. Relation type is a calibrator judgment from results.
Anti-laundering
Reject “fixes” that keep the label’s prestige while replacing the conceptual payload. That is a repair pathology — not successful calibration of Semantic Drift.
Detection sketch (report order)
For instrument I in tested domain D:
- Fix baseline (or stop at weak/indeterminate for strong drift).
- State purpose of I (job, not a restated definition) and tested domain.
- Inspect poles (definition, mechanism grade, use, effect, domain).
- Show progression (multi-stage divergence of use/effect from baseline definition+mechanism).
- Apply discriminator (not disagreement / enrichment / lawful development).
- Assign neighbor relations (independent / pathway / co-occurring).
- Judge impairment from results (≥1 family for full harm-path drift; else indeterminate).
- Issue recommendations only (tighten, split, rename, baseline, bound neighbors) — no ownership of the definition.
Primary procedure for map-level runs: Abstract Definition Mechanics. Term-level sense work: Term Decomposition Protocol.
Standing
Seed / Self-Assessed until independent runs under friction. Local reports do not generalize standing across all fields.
Permanent Beta
This standard is subject to the same diagnostics it defines.
See the Game. Refuse the Game. Build Better.