Evidence-Fidelity Decay

Prime #

843

Origin domain

Epistemology And Philosophy Of Science

Subdomain

evidentiary records and documentation → Epistemology And Philosophy Of Science

Core Idea

The delay between an event and its capture lets involuntary backfill fill memory gaps from schema and expectation, and because observation, inference, and reconstruction are written in the same register, the record's attribution structure is destroyed — leaving correctness intact but mis-weightable.

How would you explain it like I'm…

Memory Filling Gaps

If you wait a long time before telling someone about your day, you forget some parts, so you fill in the gaps with what you THINK probably happened. But when you tell the story, the real parts and the made-up parts sound exactly the same. So the listener can't tell which bits really happened and which bits you just guessed.

Guesses That Look Real

Evidence-Fidelity Decay is what happens when a written record of an event gets less accurate the longer you wait to write it down. As memory fades, your brain quietly fills the holes with guesses based on what usually happens. The tricky part is that the things you actually saw, the things you figured out, and the things you reconstructed all get written in the same plain voice. So later, a reader can't tell which sentences are real observations and which are filled-in guesses. The record might still be mostly right — what's lost is the ability to know which parts to trust the most.

Lost Attribution Over Time

Evidence-Fidelity Decay is the pattern where a record's faithfulness to its source event drops as the delay between event and capture grows, and the gap from forgetting is silently filled by inference and reconstruction from prior expectations. The decisive fact is that observation, inference, and reconstruction get written in the same register — same prose, same voice, same surface markers — so a later reader can't recover which sentences carry which evidentiary weight. What's lost isn't correctness (it may be mostly right) or completeness (it may look complete) but attribution structure: which lines are direct observation, which are inference, which are reconstruction. Because the backfill is involuntary, the recorder isn't lying — they simply can't mark which sentences came from memory versus reconstruction. The two design variables that decide whether this bites are the attribution-marking convention of the capture format and the length of the delay.

 

Evidence-Fidelity Decay is the pattern in which an evidentiary record's fidelity to its source event falls with the delay between the event and its capture, and the gap created by memory decay is silently filled by inference from remembered cues and reconstruction from prior expectations. The decisive structural fact is that observation, inference, and reconstruction are written in the same register — same prose, same voice, same surface markers — so a downstream consumer cannot recover which sentences carry which evidentiary weight. What is lost is not the record's correctness (it may be substantially right) nor its completeness (it may appear complete) but its attribution structure: which sentences are direct observation, which inference, which reconstruction. Three commitments fix the pattern. The substrate is a human-mediated record of a delay-sensitive event: the longer the gap, the more reconstruction is needed. The backfill is involuntary and undetected by the recorder, who is not fabricating but cannot mark memory versus reconstruction. And the output is registered uniformly: no typographic, syntactic, or conventional signal separates the strata, so attribution information is destroyed at capture time and is unrecoverable downstream. Because the strata are fused at the surface, the consumer mis-weights the record — treating reconstructed material as observed and observed material as reconstructed — and the design variable that decides whether this happens is the attribution-marking convention of the capture format together with the length of the delay.

Broad Use

• Qualitative research: field notes written hours later blend the seen with the inferred; contemporaneous jottings counter the decay.
• Law: witness statements days after blend what was seen, cued, and reconstructed; cross-examination recovers lost attribution.
• After-action reports: weeks-late reports fuse observation, later debriefs, and narrative arc; hot-washes run immediately.
• Oral history: at decades' remove, the account is shaped by retellings and exposure to canonical versions.
• Laboratory notebooks: a bench entry is mostly observation; an end-of-day reconstruction blends strata.
• Machine-learning labelling: labels after batch fatigue blend the artefact with the labeller's hardening schema.

Clarity

Separates three failure modes a reader conflates — incorrect, incomplete, and attribution lost — and makes the third inspectable: ask of any sentence whether it is observation, inference, or reconstruction.

Manages Complexity

Treats a whole record as the output of one known transformation (event → delay → backfill → fusion), so per-sentence inspection collapses into one audit — how long the delay, how the format marked the strata — applied to the record as a unit.

Abstract Reasoning

Because backfill is schema-driven and directional, not zero-mean noise, the reasoner can predict the direction of distortion and rank two pipelines of equal delay but different attribution conventions in advance.

Knowledge Transfer

• Across substrates: four matched interventions recur — shrink the delay (bench notes, hot-washes), mark the strata, preserve a co-record, re-capture to audit.
• Law ↔ software ↔ data labelling: a contemporaneous police statement, code-review comment, and real-time label all collapse the delay to zero.
• Safety ↔ oral history ↔ ML: an investigator six weeks after a near-miss faces the same fused-strata problem and the same fix.

Example

A police witness statement taken four days after a robbery fuses what the witness saw, what the interviewer cued, and what was reconstructed for coherence — all in uniform prose, which is why cross-examination exists to recover the lost attribution.

Relationships to Other Primes

Parents (1) — more general patterns this builds on

• Evidence-Fidelity Decay is a kind of, typical Temporal Decay and Degradation — The file: 'a TIME-DEPENDENT loss whose magnitude grows with the event-to-capture delay' — a specialization of temporal decay applied to an evidentiary record's attribution structure (vs a static bias).

Path to root: Evidence-Fidelity Decay → Temporal Decay and Degradation → Entropy (Thermodynamic Sense)

Not to Be Confused With

• Evidence-Fidelity Decay is not Provenance because it concerns the record's internal evidentiary structure at capture, before any custody chain, whereas provenance tracks custody and origin after the record exists with fixed contents.
• Evidence-Fidelity Decay is not Measurement Uncertainty because its backfill is schema-driven, correlated, and directional — so averaging certifies the bias — whereas noise is zero-mean and averages out.
• Evidence-Fidelity Decay is not Narrative Construction in History because the backfill is involuntary and undetected, whereas narrative construction is the deliberate, conscious shaping of events into a story.