drift concept

Definition	A change over time in what a consumer reads or in the data it reads. Chapter 8 tests a drift claim against its null and finds half of it was noise.
Example	A read operator measured at rank one moved to correlation 0.667 with its earlier self, against a null of 0.933.
Book	Data Mining as Observation, draft 0.2, commit `f3914f0`; entry id `drift`, kind concept.
Status	refutes or corrects [refuted]; refutes or corrects [void]. Corrections: 2 item(s), see below.
Defining equation	Book equation 13.1.
Assumptions and scope	A drift is a change over time in what a consumer reads or in the data it reads, and its damage is the read distortion of the drift, so it is consumer-relative. The same drift is read as its full squared size by one consumer and as nothing by another, and a drift confined to the nuisance is read as zero however large. The operator-drift claim was tested against a paired null and half of it was noise. The rank-one drift held in fourteen of sixteen cells on one model, and the refresh intervention derived from it was refuted, which the ledger carries.
Prior art	none recorded
Evidence	`geometric-observation/claims/LEDGER.md:36`, `geometric-observation/claims/LEDGER.md:48`, `readscope/CALIBRATION.md:600-660`, `readscope/calibration/records/c11c-operator-drift.json`, `lean/DataMiningAsObservation/Drift.lean`
Reviewed	not yet reviewed; generated 2026-09-10 from records at the commits on the provenance page.

The read operator moving over time, so an old certificate no longer describes the reader.

Equation

Book equation 13.1.

\[\mathrm{FC}=\Pr\big[W\ \text{refutes}\ \big|\ \mathcal C_t\ \text{clears}\big],\qquad d_O(\Delta)=\operatorname{tr}\big(P_C\,M_{\mathrm{drift}}(\Delta)\big),\qquad M_{\mathrm{drift}}(\Delta)=\mathbb E\big[\delta_\Delta\delta_\Delta^{\top}\big].\]

Conditions

A drift is a change over time in what a consumer reads or in the data it reads, and its damage is the read distortion of the drift, so it is consumer-relative. The same drift is read as its full squared size by one consumer and as nothing by another, and a drift confined to the nuisance is read as zero however large.
The operator-drift claim was tested against a paired null and half of it was noise. The rank-one drift held in fourteen of sixteen cells on one model, and the refresh intervention derived from it was refuted, which the ledger carries.

Conditions are curated in entries.toml rather than read from a record.

Ledger

refutes or corrects. OT-4 [refuted]. Operator drift predicts a real long-generation degradation and a derived refresh intervention moves it. geometric-observation/claims/LEDGER.md:36.
refutes or corrects. OT-11 [void]. Feedback-free staleness: streaming-retrieval damage tracks measured drift; derived-cadence re-allocation removes it. geometric-observation/claims/LEDGER.md:48.

First stated

Volume 14, chapter 19, geometric-observation/chapters/ch19_the_certificate_that_ages.md, DOI 10.5281/zenodo.21776291, and the operator-drift record in readscope, readscope/calibration/records/c11c-operator-drift.json.

Measurements

Where the book states it	Numbers, as the book’s sources table records them	Source
chapter 11 section 11.9	drift at rank one 0.667 vs null 0.933, sixteen cells	`readscope/CALIBRATION.md:600-660` F-24; `readscope/calibration/records/c11c-operator-drift.json`

Failures and corrections

OT-4, [refuted]. Operator drift predicts a real long-generation degradation and a derived refresh intervention moves it. geometric-observation/claims/LEDGER.md:36.
OT-11, [void]. Feedback-free staleness: streaming-retrieval damage tracks measured drift; derived-cadence re-allocation removes it. geometric-observation/claims/LEDGER.md:48.

Invariance envelope

none declared

Machine checked

lean/DataMiningAsObservation/Drift.lean, theorems damage_rank_one, same_drift_two_consumers, damage_eq_zero_of_nuisance, damage_nonneg, at observation-data-mining f3914f0; what the check covers is stated in the book’s appendix C.

Used in

Data Mining as Observation chapters 2, 8, 11, 13, 14.

coherence time; freshness; certificate; read distortion.

Status

Generated 2026-09-10 by encyclopedia/generate.py; book at observation-data-mining f3914f0; the commit of every record is listed in the encyclopedia’s provenance.

← dot productearly stopping →